Machine Learning and Deep Learning in Medical Imaging

Journal Name	Impact Factor	CiteScore	Launched Year	First Decision (median)	APC
BioMed biomed	-	-	2021	28.1 Days	CHF 1000	Submit
Cancers cancers	4.4	8.8	2009	19.1 Days	CHF 2900	Submit
Diagnostics diagnostics	3.3	5.9	2011	21.6 Days	CHF 2600	Submit
Journal of Clinical Medicine jcm	2.9	5.2	2012	18.5 Days	CHF 2600	Submit
Journal of Imaging jimaging	3.3	6.7	2015	18 Days	CHF 1800	Submit

37 pages, 1793 KB

Open AccessSystematic Review

The Role of Artificial Intelligence in Prognosis, Recurrence Prediction, and Treatment Outcomes in Laryngeal Cancer: A Systematic Review

by Hadi Afandi Al-Hakami, Ismail A. Abdullah, Nora S. Almutairi, Rimaz R. Aldawsari, Ghadah Ali Alluqmani, Halah Ahmed Fallatah, Yara Saud Alsulami, Elyas Mohammed Alasiri, Rahaf D. Alsufyani, Raghad Ayman Alorabi and Reffal Mohammad Aldainiy

Cancers 2026, 18(8), 1257; https://doi.org/10.3390/cancers18081257 - 16 Apr 2026

Viewed by 573

Abstract

Background: Laryngeal cancer (LC), a common subtype of head and neck cancers (HNC), is most frequently represented by laryngeal squamous cell carcinoma (LSCC). Prognosis largely depends on early detection; however, traditional prognostic tools, including tumor-node-metastasis (TNM) staging, often show limited predictive accuracy. Artificial [...] Read more.

Background: Laryngeal cancer (LC), a common subtype of head and neck cancers (HNC), is most frequently represented by laryngeal squamous cell carcinoma (LSCC). Prognosis largely depends on early detection; however, traditional prognostic tools, including tumor-node-metastasis (TNM) staging, often show limited predictive accuracy. Artificial intelligence (AI), including machine learning (ML), natural language processing, and deep learning (DL), has emerged as a promising approach to improving cancer diagnosis, prognosis, and treatment planning by analyzing clinical data and medical imaging. Objective: This systematic review assesses the role of AI in prognosis, recurrence prediction, and treatment outcomes in LC. Methods: PubMed, MEDLINE, Scopus, Web of Science, IEEE Xplore, and ScienceDirect were searched up to January 2025. A total of 1062 records were identified; after title/abstract screening and full-text assessment, 29 studies were included. Eligible studies involved adult patients with LC and applied AI to diagnose, prognose, predict recurrence, or assess treatment outcomes using human datasets. Study quality and risk of bias were evaluated using the QUADAS-2 and QUIPS. Results: The 29 included studies were mostly retrospective, with sample sizes ranging from 10 to 63,000 patients. Most focused on LSCC, with a higher prevalence in males. The studies utilized various AI techniques, including deep learning models such as convolutional neural networks (CNNs) and DeepSurv, as well as ML algorithms like random survival forest, gradient boosting machines, random forest, k-nearest neighbors, naïve Bayes, and decision trees. AI models demonstrated strong prognostic performance, surpassing Cox regression and TNM staging in predicting survival and recurrence. Several studies reported outcomes related to treatment, such as chemotherapy response, occult lymph node metastasis, and the need for salvage surgery. Methodological quality varied, with biases related to patient selection and confounding factors. Conclusions: AI has the potential to improve prognosis estimation, recurrence prediction, and treatment outcome assessment in LC. However, although AI can be a helpful addition to clinical decision-making, more prospective studies, external validation, and standardized evaluation are necessary before these technologies can be confidently adopted in everyday clinical practice. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

15 pages, 2099 KB

Open AccessReview

Current Trends and Future Prospects of Radiomics and Machine Learning (ML) Models in Spinal Tumors—A Narrative Review

by Vivek Sanker, Suhrud Panchawgh, Anmol Kaur, Vinay Suresh, Dhanya Mahesh, Eeman Ahmad, Srinath Hariharan, Dhiraj Pangal, Maria Jose Cavgnaro, Mirabela Rusu, John Ratliff and Atman Desai

J. Imaging 2026, 12(3), 138; https://doi.org/10.3390/jimaging12030138 - 19 Mar 2026

Viewed by 541

Abstract

The intersection between radiomics, the computational analysis of imaging data, and machine learning (ML) may lead to new developments in the diagnosis, prognosis, and management of diseases. For spinal tumors specifically, applications of these fields appear promising. In this educational narrative review, we [...] Read more.

The intersection between radiomics, the computational analysis of imaging data, and machine learning (ML) may lead to new developments in the diagnosis, prognosis, and management of diseases. For spinal tumors specifically, applications of these fields appear promising. In this educational narrative review, we provide a summary of the current advancements in radiomics and artificial intelligence (AI), as well as applications of both fields in the diagnosis and management of spinal tumors. We also provide a suggested workflow of radiomics and machine learning analysis of spinal tumors for researchers, including a list and description of commonly used radiomic features. Future directions in the field of radiomics and machine learning applications to spinal tumors may involve validating already proposed algorithms with larger datasets, ensuring that all computational applications to patient care maintain high ethical standards, and continuing work in developing novel and highly accurate computational techniques to enhance patient outcomes. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

13 pages, 2003 KB

Open AccessArticle

External Validation of an Open-Source Model for Automated Muscle Segmentation in CT Imaging of Cancer Patients

by Hendrik Erenstein, Jona Van den Broeck, Annemieke van der Heij-Meijer, Wim P. Krijnen, Aldo Scafoglieri, Harriët Jager-Wittenaar, Martine Sealy and Peter van Ooijen

J. Imaging 2026, 12(3), 135; https://doi.org/10.3390/jimaging12030135 - 18 Mar 2026

Viewed by 475

Abstract

Computed tomography (CT) at the third lumbar vertebra (L3) is widely used for muscle quantification, but manual segmentation is labor intensive. This study externally validates an AI model, trained on a public dataset, for automated L3 muscle segmentation using an independent cohort, including [...] Read more.

Computed tomography (CT) at the third lumbar vertebra (L3) is widely used for muscle quantification, but manual segmentation is labor intensive. This study externally validates an AI model, trained on a public dataset, for automated L3 muscle segmentation using an independent cohort, including a subgroup analysis of subject characteristics (e.g., age and a history of cancer). The AI model was trained on 900 CT scans with expert annotations from a publicly available repository. Validation was performed on 232 PET CT scans from the University Hospital Brussels, each manually segmented by an expert. Segmentation post-processing employed a density-based clustering algorithm to discard arm muscles and Hounsfield unit (HU) thresholding to refine the muscle segmentation. Performance was assessed using the Dice Similarity Coefficient (DSC) and Segmentation Surface Error (SSE). The model achieved a median DSC of 0.978 and a median SSE of 3.863 cm² across the validation set. At lower BMI values, the model was more prone to overestimation of muscle surface area. Most segmentation errors occurred in the abdominal wall muscles. Analysis showed no significant difference between arm positioning above the head and alongside the body, indicating robustness to minor artifacts from arm positioning. The AI model delivers accurate, automated L3 muscle segmentation, supporting larger-scale body composition studies. However, diminished accuracy at low BMI values and limited demographic diversity of the data highlight the need for broader validation. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

21 pages, 4219 KB

Open AccessArticle

3D-StyleGAN2-ADA: Volumetric Synthesis of Realistic Prostate T2W MRI

by Claudia Giardina and Verónica Vilaplana

J. Imaging 2026, 12(3), 130; https://doi.org/10.3390/jimaging12030130 - 14 Mar 2026

Viewed by 496

Abstract

This work investigates the extension of StyleGAN2-ADA to three-dimensional prostate T2-weighted (T2W) MRI generation. The architecture is adapted to operate on 3D anisotropic volumes, enabling stable training at a clinically relevant resolution of

256 \times 256 \times 24

, where a baseline 3D-StyleGAN [...] Read more.

This work investigates the extension of StyleGAN2-ADA to three-dimensional prostate T2-weighted (T2W) MRI generation. The architecture is adapted to operate on 3D anisotropic volumes, enabling stable training at a clinically relevant resolution of

256 \times 256 \times 24

, where a baseline 3D-StyleGAN fails to converge. Quantitative evaluation using Fréchet Inception Distance (FID), Kernel Inception Distance (KID), and generative Precision–Recall metrics demonstrates substantial improvements over a 3D-StyleGAN baseline. Specifically, FID decreased from 114.2 to 27.3, while generative Precision increased from 0.22 to 0.82, indicating markedly improved fidelity and alignment with the real data distribution. Beyond generative metrics, the synthetic volumes were evaluated through radiomic feature analysis and downstream prostate segmentation. Synthetic data augmentation resulted in segmentation performance comparable to real-data training, supporting that volumetric generation preserves anatomically relevant structures, while multivariate radiomic analyses showed strong global feature alignment between real and synthetic volumes. These findings indicate that a 3D extension of StyleGAN2-ADA enables stable high-resolution volumetric prostate MRI synthesis while preserving anatomically coherent structure and global radiomic characteristics. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

18 pages, 2234 KB

Open AccessArticle

A Gated Attention-Based Multiple Instance Learning and Test-Time Augmentation Approach for Diagnosing Active Sacroiliitis in Sacroiliac Joint MRI Scans

by Zeynep Keskin, Onur İnan, Ömer Özberk, Reyhan Bilici, Sema Servi, Selma Özlem Çelikdelen and Mehmet Yıldırım

J. Clin. Med. 2026, 15(6), 2101; https://doi.org/10.3390/jcm15062101 - 10 Mar 2026

Viewed by 415

Abstract

Background and Objective: Axial spondyloarthritis (axSpA) is a group of chronic inflammatory diseases that primarily affect the sacroiliac joints. Early diagnosis is crucial for preventing irreversible structural damage. Magnetic Resonance Imaging (MRI) is the gold standard for detecting early inflammatory changes such as [...] Read more.

Background and Objective: Axial spondyloarthritis (axSpA) is a group of chronic inflammatory diseases that primarily affect the sacroiliac joints. Early diagnosis is crucial for preventing irreversible structural damage. Magnetic Resonance Imaging (MRI) is the gold standard for detecting early inflammatory changes such as sacroiliitis. However, conventional MRI interpretation is inherently subjective and susceptible to both intra- and inter-observer variability. Therefore, artificial intelligence (AI)-driven diagnostic solutions are increasingly being explored. Among them, the Gated Attention Multiple Instance Learning (MIL) framework holds strong potential in modeling heterogeneous inflammatory distributions, thanks to its slice-level attention mechanism. This study aims to evaluate the diagnostic performance of a deep learning model based on Gated Attention MIL for automated sacroiliitis detection. Furthermore, its results are compared with a baseline deep learning architecture (standard ResNet-18), and its consistency with radiologist annotations is analyzed. Materials and Methods: The dataset included 554 subjects, comprising 276 patients diagnosed with axSpA and 278 healthy controls. All MRI data were derived from axial T2-weighted fat-suppressed (T2_TSE_TRA_FS) sequences. Patient-wise data splitting was employed to construct training, validation, and independent test sets. The proposed model architecture integrates ResNet-18-based feature extraction, a gated attention mechanism for instance-level weighting, and bag-level classification. Additionally, Test-Time Augmentation (TTA) was implemented to enhance robustness during inference. Results: On the independent test set, the model achieved an accuracy of 85.88%, sensitivity of 92.86%, specificity of 79.07%, and an F1-score of 86.67%. Attention heatmaps generated by the MIL module showed strong spatial overlap with bone marrow edema regions annotated by expert radiologists. Implementation of TTA led to an approximate 10% improvement in overall classification accuracy. Conclusions: The Gated Attention MIL framework demonstrated high diagnostic performance for sacroiliitis detection, indicating its value as a reliable decision support tool for early axSpA diagnosis. Validation on larger, multi-center datasets is warranted to ensure generalizability and to support clinical integration in routine radiology workflows. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Graphical abstract

28 pages, 56643 KB

Open AccessArticle

Endo-DET: A Domain-Specific Detection Framework for Multi-Class Endoscopic Disease Detection

by Yijie Lu, Yixiang Zhao, Qiang Yu, Wei Shao and Renbin Shen

J. Imaging 2026, 12(3), 112; https://doi.org/10.3390/jimaging12030112 - 6 Mar 2026

Viewed by 593

Abstract

Gastrointestinal cancers account for roughly a quarter of global cancer incidence, and early detection through endoscopy has proven effective in reducing mortality. Multi-class endoscopic disease detection, however, faces three persistent challenges: feature redundancy from non-pathological content, severe illumination inconsistency across imaging modalities, and [...] Read more.

Gastrointestinal cancers account for roughly a quarter of global cancer incidence, and early detection through endoscopy has proven effective in reducing mortality. Multi-class endoscopic disease detection, however, faces three persistent challenges: feature redundancy from non-pathological content, severe illumination inconsistency across imaging modalities, and extreme scale variability with blurry boundaries. This paper introduces Endo-DET, a domain-specific detection framework addressing these challenges through three synergistic components. The Adaptive Lesion-Discriminative Filtering (ALDF) module achieves lesion-focused attention via sparse simplex projection, reducing complexity from

O (N^{2})

to

O (α N^{2})

. The Global–Local Illumination Modulation Neck (GLIM-Neck) enables illumination-aware multi-scale fusion through four cooperative mechanisms, maintaining stable performance across white-light endoscopy, narrow-band imaging, and chromoendoscopy. The Lesion-aware Unified Calibration and Illumination-robust Discrimination (LUCID) module uses dual-stream reciprocal modulation to integrate boundary-sensitive textures with global semantics while suppressing instrument artifacts. Experiments on EDD2020, Kvasir-SEG, PolypGen2021, and CVC-ClinicDB show that Endo-DET improves mAP50-95 over the DEIM baseline by 5.8, 10.8, 4.1, and 10.1 percentage points respectively, with mAP75 gains of 6.1, 10.3, 6.8, and 9.3 points, and Recall50-95 improvements of 10.9, 12.1, 11.1, and 11.5 points. Running at 330 FPS with TensorRT FP16 optimization, Endo-DET achieves consistent cross-dataset improvements while maintaining real-time capability, providing a methodological foundation for clinical computer-aided diagnosis. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

25 pages, 1420 KB

Open AccessArticle

Identification of Retinal Diseases Using Light Convolutional Neural Networks and Intrinsic Mode Function Technique

by Preethi Kulkarni and Konda Srinivasa Reddy

Diagnostics 2026, 16(5), 773; https://doi.org/10.3390/diagnostics16050773 - 4 Mar 2026

Cited by 1 | Viewed by 406

Abstract

Background/Objectives: Fundus imaging provides a detailed view of the interior surface of the eye and plays a crucial role in the early diagnosis of retinal diseases. However, automated interpretation of fundus images remains challenging due to variations in illumination, noise, and structural [...] Read more.

Background/Objectives: Fundus imaging provides a detailed view of the interior surface of the eye and plays a crucial role in the early diagnosis of retinal diseases. However, automated interpretation of fundus images remains challenging due to variations in illumination, noise, and structural complexity. Methods: A novel hybrid model that integrates the Intrinsic Mode Function (IMF) filter, derived from Empirical Mode Decomposition (EMD), with a Light Convolutional Neural Network (LightCNN) for enhanced fundus image classification was proposed. The IMF filter effectively decomposes the input signal into intrinsic components, isolating high-frequency noise and preserving critical retinal patterns. These refined components are subsequently processed by the LightCNN architecture, which offers lightweight yet highly discriminative feature extraction and classification capabilities. Results: Experimental results on DIARETDB fundus datasets demonstrate that the proposed IMF + LightCNN model achieves 99.4% accuracy, 99.1% precision, 98.87% recall, and a 98.31 F1-score, significantly outperforming conventional CNN and ResNet-based models. Conclusions: Integrating advanced signal processing with lightweight deep learning improves both diagnostic accuracy and computational efficiency. This hybrid framework establishes a promising pathway for reliable and real-time clinical screening of retinal diseases. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

15 pages, 1404 KB

Open AccessArticle

A Deep Learning-Based Decision Support System for Cholelithiasis in MRI Data

by Ebru Hasbay, Caglar Cengizler, Mahmut Ucar, Nagihan Durgun, Hayriye Ulkucan Disli and Deniz Bolat

J. Clin. Med. 2026, 15(5), 1891; https://doi.org/10.3390/jcm15051891 - 2 Mar 2026

Viewed by 411

Abstract

Background: Cholelithiasis can lead to significant complications if not diagnosed and treated promptly. Recent advances in deep learning and the improved ability of computer systems to detect clinically significant textural and morphological patterns in magnetic resonance imaging (MRI) can help reduce the time [...] Read more.

Background: Cholelithiasis can lead to significant complications if not diagnosed and treated promptly. Recent advances in deep learning and the improved ability of computer systems to detect clinically significant textural and morphological patterns in magnetic resonance imaging (MRI) can help reduce the time and resources required for the radiological evaluation of the gallbladder and cholelithiasis. Objective: To detect cholelithiasis, a support system with a graphical user interface for magnetic resonance (MR) images of the gallbladder was implemented to reduce the manual effort and time required to identify gallstones. Method: A commonly used deep learning model for pixel-level mask generation and instance segmentation, Mask Region Based Convolutional Neural Network (Mask R-CNN), was modified, trained, and evaluated to provide a robust pipeline for automated analysis. The primary aim was to automatically locate and label the gallbladder in T2-weighted axial MR images to detect gallstones and highlight the visual characteristics of the target region, thereby supporting radiologists. All automation was designed to operate on a single optimal slice instead of the entire volume. While this approach limits generalisability, it offers a practical starting point for method development. This setup reflects a feasibility-oriented design, rather than a comprehensive diagnostic capability. The dataset included 788 axial MR images from different patients. Each image was labeled and segmented by an experienced radiologist to train and test the models at the image level. Results: The proposed model with squeeze and excitation (SE) modification improved classification accuracy, and at the image level, stone detection improved in terms of accuracy, precision, and specificity, although recall and F1 scores slightly decreased. Conclusions: The results show that the modified Mask R-CNN model can detect gallstones with up to 0.89 accuracy, supporting the clinical applicability of the proposed method. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

26 pages, 1041 KB

Open AccessReview

Artificial Intelligence in Orthopaedics: Clinical Performance, Limitations, and Translational Readiness—A Review

by Wojciech Michał Glinkowski, Antonina Spalińska, Agnieszka Wołk and Krzysztof Wołk

J. Clin. Med. 2026, 15(5), 1751; https://doi.org/10.3390/jcm15051751 - 25 Feb 2026

Cited by 1 | Viewed by 1464

Abstract

Background/Objectives: Musculoskeletal disorders and their surgical treatment significantly affect global disability, healthcare utilization, and costs. Artificial intelligence (AI) is a key enabler of data-driven musculoskeletal care. Their applications include diagnostic imaging, surgical planning, risk prediction, rehabilitation, and digital health ecosystems. This narrative review [...] Read more.

Background/Objectives: Musculoskeletal disorders and their surgical treatment significantly affect global disability, healthcare utilization, and costs. Artificial intelligence (AI) is a key enabler of data-driven musculoskeletal care. Their applications include diagnostic imaging, surgical planning, risk prediction, rehabilitation, and digital health ecosystems. This narrative review synthesizes current evidence on the use of AI in orthopaedics and musculoskeletal care across five areas: diagnostic imaging, surgical planning and intraoperative augmentation, predictive analytics and patient-reported outcomes, rehabilitation intelligence and teleorthopaedics, and system-level management. An additional task is to identify translational gaps and priorities for safe, ethical, and equitable implementation of AI. Methods: A structured narrative review was conducted using targeted searches in PubMed, Scopus, and Web of Science supplemented by semantic and citation-based explorations in Semantic Scholar, OpenAlex, and Google Scholar. The main search period was January 2019 to December 2025. The retrieved peer-reviewed articles were analyzed for clinical relevance to human musculoskeletal care, quantitative outcomes, and the translational implications of the results. From the broader pool of eligible publications, 40 clinically relevant studies were selected for detailed synthesis covering imaging, surgical planning, predictive modeling, rehabilitation, and system-level applications. Owing to the significant heterogeneity in the model architectures, datasets, and endpoints, the results were organized into five predefined thematic areas. Results: The most mature evidence is for AI-assisted detection of bone fractures on radiographs, identification of implants, and use of sizing templates in preoperative planning for arthroplasty, where deep learning systems have achieved expert-level diagnostic performance (e.g., fracture detection sensitivity of approximately 90% and specificity of approximately 92% and implant identification accuracy of 97–99%) and improved the accuracy of preoperative planning compared to conventional templating. AI-based planning increases the likelihood of reducing intraoperative corrections, shortening surgery time, reducing blood loss, and improving the final functional outcomes. Predictive models can support the stratification of risk for complications, rehospitalizations, and patient-reported outcomes, although external validation remains limited and is often single-center at this stage of research. Emerging applications in rehabilitation and teleorthopaedics, including sensor-based monitoring and learning systems integrated with Patient-Reported Outcome Measures (PROMs), are conceptually promising, but are mainly limited to feasibility or pilot studies. Conclusions: AI is beginning to influence musculoskeletal care, moving beyond pattern recognition toward integrated, patient-centered decision support throughout the perioperative and rehabilitation periods. Its widespread use remains constrained by limited multicenter validation, dataset bias, algorithmic opacity, and immature regulatory and governance frameworks. Future work should prioritize prospective multicenter impact studies, repeatable revalidation of local models, integration of PROM and teleorthopedic data with health learning systems, and adaptation to changing regulatory requirements to enable safe, ethical, effective, and equitable implementation in routine orthopedic practice. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

21 pages, 896 KB

Open AccessArticle

Adaptive Bandelet Transform and Transfer Learning for Geometry-Aware Thyroid Cancer Ultrasound Classification

by Yassine Habchi, Hamza Kheddar, Mohamed Chahine Ghanem and Jamal Hwaidi

Diagnostics 2026, 16(4), 554; https://doi.org/10.3390/diagnostics16040554 - 13 Feb 2026

Viewed by 517

Abstract

Background and Objectives: Classification of thyroid nodules (TN) in ultrasound remains challenging due to limited labelled data and the limited capacity of conventional feature representations to capture complex, multi-directional textures. This work aims to improve data-efficient TN classification by integrating a geometry-adaptive Bandelet [...] Read more.

Background and Objectives: Classification of thyroid nodules (TN) in ultrasound remains challenging due to limited labelled data and the limited capacity of conventional feature representations to capture complex, multi-directional textures. This work aims to improve data-efficient TN classification by integrating a geometry-adaptive Bandelet Transform (BT) with transfer learning (TL) to enhance feature representation and generalisation. Methods: The proposed pipeline first applies BT to strengthen directional and structural encoding in ultrasound images via quadtree-driven geometric adaptation. It then mitigates class imbalance using SMOTE and increases data diversity through targeted data augmentation. The resulting representations are classified using multiple ImageNet-pretrained architectures, where VGG19 yields the most consistent performance. Results: Experiments on the publicly available DDTI dataset show that BT-based preprocessing consistently improves performance over classical wavelet representations across multiple quadtree thresholds, with the best results obtained at

T = 30

. Under this setting, the proposed BT+TL (VGG19) model achieves 98.91% accuracy, 98.11% sensitivity, 97.31% specificity, and a 98.89% F1-score, outperforming comparable approaches reported in the literature. Conclusions: Coupling geometry-adaptive transforms with modern TL backbones provides a robust and data-efficient strategy for ultrasound TN classification, particularly under limited annotation and challenging texture variability. The complete project is publicly available. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

14 pages, 1705 KB

Open AccessArticle

A Multiphase CT-Based Integrated Deep Learning Framework for Rectal Cancer Detection, Segmentation, and Staging: Performance Comparison with Radiologist Assessment

by Tzu-Hsueh Tsai, Jia-Hui Lin, Yen-Te Liu, Jhing-Fa Wang, Chien-Hung Lee and Chiao-Yun Chen

J. Imaging 2026, 12(2), 76; https://doi.org/10.3390/jimaging12020076 - 10 Feb 2026

Viewed by 559

Abstract

Accurate staging of rectal cancer is crucial for treatment planning; however, computed tomography (CT) interpretation remains challenging and highly dependent on radiologist expertise. This study aimed to develop and evaluate an AI-assisted system for rectal cancer detection and staging using CT images. The [...] Read more.

Accurate staging of rectal cancer is crucial for treatment planning; however, computed tomography (CT) interpretation remains challenging and highly dependent on radiologist expertise. This study aimed to develop and evaluate an AI-assisted system for rectal cancer detection and staging using CT images. The proposed framework integrates three components—a convolutional neural network (RCD-CNN) for lesion detection, a U-Net model for rectal contour delineation and tumor localization, and a 3D convolutional network (RCS-3DCNN) for staging prediction. CT scans from 223 rectal cancer patients at Kaohsiung Medical University Chung-Ho Memorial Hospital were retrospectively analyzed, including both non-contrast and contrast-enhanced studies. RCD-CNN achieved an accuracy of 0.976, recall of 0.975, and precision of 0.976. U-Net yielded Dice scores of 0.897 (rectal contours) and 0.856 (tumor localization). Radiologist-based clinical staging had 82.6% concordance with pathology, while AI-based staging achieved 80.4%. McNemar’s test showed no significant difference between the AI and radiologist staging results (p = 1.0). The proposed AI-assisted system achieved staging accuracy comparable to that of radiologists and demonstrated feasibility as a decision-support tool in rectal cancer management. This study introduces a novel three-stage, dual-phase CT-based AI framework that integrates lesion detection, segmentation, and staging within a unified workflow. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

19 pages, 2617 KB

Open AccessArticle

Topic-Modeling Guided Semantic Clustering for Enhancing CNN-Based Image Classification Using Scale-Invariant Feature Transform and Block Gabor Filtering

by Natthaphong Suthamno and Jessada Tanthanuch

J. Imaging 2026, 12(2), 70; https://doi.org/10.3390/jimaging12020070 - 9 Feb 2026

Viewed by 471

Abstract

This study proposes a topic-modeling guided framework that enhances image classification by introducing semantic clustering prior to CNN training. Images are processed through two key-point extraction pipelines: Scale-Invariant Feature Transform (SIFT) with Sobel edge detection and Block Gabor Filtering (BGF), to obtain local [...] Read more.

This study proposes a topic-modeling guided framework that enhances image classification by introducing semantic clustering prior to CNN training. Images are processed through two key-point extraction pipelines: Scale-Invariant Feature Transform (SIFT) with Sobel edge detection and Block Gabor Filtering (BGF), to obtain local feature descriptors. These descriptors are clustered using K-means to build a visual vocabulary. Bag of Words histograms then represent each image as a visual document. Latent Dirichlet Allocation is applied to uncover latent semantic topics, generating coherent image clusters. Cluster-specific CNN models, including AlexNet, GoogLeNet, and several ResNet variants, are trained under identical conditions to identify the most suitable architecture for each cluster. Two topic guided integration strategies, the Maximum Proportion Topic (MPT) and the Weight Proportion Topic (WPT), are then used to assign test images to the corresponding specialized model. Experimental results show that both the SIFT-based and BGF-based pipelines outperform non-clustered CNN models and a baseline method using Incremental PCA, K-means, Same-Cluster Prediction, and unweighted Ensemble Voting. The SIFT pipeline achieves the highest accuracy of 95.24% with the MPT strategy, while the BGF pipeline achieves 93.76% with the WPT strategy. These findings confirm that semantic structure introduced through topic modeling substantially improves CNN classification performance. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

34 pages, 1241 KB

Open AccessReview

Advanced Microwave Imaging Techniques for Early Detection of Breast Cancer: A Review and Future Perspectives

by Areej Safdar, Behnaz Sohani, Faiz Iqbal, Roohollah Barzamini, Amir Rahmani and Aliyu Aliyu

BioMed 2026, 6(1), 6; https://doi.org/10.3390/biomed6010006 - 3 Feb 2026

Viewed by 1599

Abstract

Breast cancer remains the most frequently diagnosed cancer in women worldwide, with outcomes strongly dependent on stage at detection. Conventional imaging modalities such as mammography, ultrasound and MRI are limited by reduced sensitivity in dense breasts, radiation exposure, high cost and restricted availability [...] Read more.

Breast cancer remains the most frequently diagnosed cancer in women worldwide, with outcomes strongly dependent on stage at detection. Conventional imaging modalities such as mammography, ultrasound and MRI are limited by reduced sensitivity in dense breasts, radiation exposure, high cost and restricted availability in low-resource settings. This review critically examines microwave imaging (MWI) as a non-invasive, radiation-free and an emerging resource-efficient breast imaging modality that exploits dielectric contrast between healthy and malignant breast tissues. We first summarise experimental and clinical evidence on breast dielectric properties and their implications for numerical phantoms and device design. We then review passive, active (tomographic and radar-based) and hybrid MWI systems, including key clinical prototypes such as SAFE, MammoWave, MARIA and Wavelia, and analyse associated image-reconstruction algorithms from classical inverse scattering to advanced beamforming, Huygens-based methods and AI based reconstruction. Finally, we discuss outstanding challenges—tissue heterogeneity, calibration, hardware constraints and computational complexity—and identify future directions including AI-assisted reconstruction, multimodal hybrid imaging and large-scale clinical validation needed to translate MWI into routine breast cancer screening and diagnosis. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

11 pages, 1590 KB

Open AccessArticle

Radiomic Analysis for Ki-67 Classification in Small Bowel Neuroendocrine Tumors

by Filippo Checchin, Davide Malerba, Alessandro Gambella, Aurora Rita Puleri, Virginia Sambuceti, Alessandro Vanoli, Federica Grillo, Lorenzo Preda and Chandra Bortolotto

Cancers 2026, 18(3), 463; https://doi.org/10.3390/cancers18030463 - 30 Jan 2026

Viewed by 587

Abstract

Objective: To analyze radiomic features extracted from CT images of small bowel neuroendocrine tumors and evaluate their association with Ki-67 expression. Methods: 128 small bowel NET primary and secondary lesions from 34 patients were analyzed. Manual segmentation of the lesions was [...] Read more.

Objective: To analyze radiomic features extracted from CT images of small bowel neuroendocrine tumors and evaluate their association with Ki-67 expression. Methods: 128 small bowel NET primary and secondary lesions from 34 patients were analyzed. Manual segmentation of the lesions was conducted on portal-phase CT images using ITK-SNAP v. 4.0^®, and 107 radiomic features were extracted using the PyRadiomics library. The lesions were categorized into two groups based on their Ki-67 index expression (≤1% and >1%). Correlation filtering reduced the set of 107 to 41 radiomic features. Inferential statistical analyses (t-test and Mann–Whitney U, following Shapiro–Wilk and Levene’s tests) identified 19 significant features (p < 0.05) that were predominantly texture related. A ranking procedure further reduced these to eight top-performing variables across multiple selection methods (Information Gain, Gini, ANOVA, χ²). Five supervised Machine Learning models (Logistic Regression, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), XGBoost, and Random Forest) were trained and validated using 5-fold cross-validation. The evaluation metrics employed included AUC, accuracy, precision, recall, F1 score, and a confusion matrix. Results: Random Forest exhibited the best overall performance (AUC = 0.80; F1 score = 0.813; Recall = 0.847). The model’s low false negative rate (15.3%) suggests potential clinical utility in minimizing the risk of underestimating more aggressive lesions. Conclusions: Radiomics represents a promising frontier to identify patterns associated with histopathological markers. This study highlights its potential for non-invasive assessment of proliferative rate in small bowel neuroendocrine tumors, confirming the performance in the literature, and posing an interesting prospect for future research. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

35 pages, 1699 KB

Open AccessReview

Will AI Replace Physicians in the Near Future? AI Adoption Barriers in Medicine

by Rafał Obuchowicz, Adam Piórkowski, Karolina Nurzyńska, Barbara Obuchowicz, Michał Strzelecki and Marzena Bielecka

Diagnostics 2026, 16(3), 396; https://doi.org/10.3390/diagnostics16030396 - 26 Jan 2026

Cited by 8 | Viewed by 2611

Abstract

Objectives: This study aims to evaluate whether contemporary artificial intelligence (AI), including convolutional neural networks (CNNs) for medical imaging and large language models (LLMs) for language processing, could replace physicians in the near future and to identify the principal clinical, technical, and [...] Read more.

Objectives: This study aims to evaluate whether contemporary artificial intelligence (AI), including convolutional neural networks (CNNs) for medical imaging and large language models (LLMs) for language processing, could replace physicians in the near future and to identify the principal clinical, technical, and regulatory barriers. Methods: A narrative review is conducted on the scientific literature addressing AI performance and reproducibility in medical imaging, LLM competence in medical knowledge assessment and patient communication, limitations in out-of-distribution generalization, absence of physical examination and sensory inputs, and current regulatory and legal frameworks, particularly within the European Union. Results: AI systems demonstrate high accuracy and reproducibility in narrowly defined tasks, such as image interpretation, lesion measurement, triage, documentation support, and written communication. These capabilities reduce interobserver variability and support workflow efficiency. However, major obstacles to physician replacement persist, including limited generalization beyond training distributions, inability to perform physical examination or procedural tasks, susceptibility of LLMs to hallucinations and overconfidence, unresolved issues of legal liability at higher levels of autonomy, and the continued requirement for clinician oversight. Conclusions: In the foreseeable future, AI will augment rather than replace physicians. The most realistic trajectory involves automation of well-defined tasks under human supervision, while clinical integration, physical examination, procedural performance, ethical judgment, and accountability remain physician-dependent. Future adoption should prioritize robust clinical validation, uncertainty management, escalation pathways to clinicians, and clear regulatory and legal frameworks. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

17 pages, 2764 KB

Open AccessArticle

Radiomics as a Decision Support Tool for Detecting Occult Periapical Lesions on Intraoral Radiographs

by Barbara Obuchowicz, Joanna Zarzecka, Marzena Jakubowska, Rafał Obuchowicz, Michał Strzelecki, Adam Piórkowski, Joanna Gołda, Karolina Nurzynska and Julia Lasek

J. Clin. Med. 2026, 15(3), 971; https://doi.org/10.3390/jcm15030971 - 25 Jan 2026

Viewed by 521

Abstract

Background: Periapical lesions are common consequences of pulp necrosis but may remain undetectable on conventional intraoral radiographs, becoming evident only on cone-beam computed tomography (CBCT). Improving lesion recognition on plain radiographs is therefore of high clinical relevance. Methods: This retrospective, single-center study analyzed [...] Read more.

Background: Periapical lesions are common consequences of pulp necrosis but may remain undetectable on conventional intraoral radiographs, becoming evident only on cone-beam computed tomography (CBCT). Improving lesion recognition on plain radiographs is therefore of high clinical relevance. Methods: This retrospective, single-center study analyzed 56 matched pairs of intraoral periapical radiographs (RVG) and CBCT scans. A total of 109 regions of interest (ROIs) were included, which were classified as CBCT-positive/RVG-negative (onlyCBCT, n = 64) or true negative (noLesion, n = 45). Radiomic texture features were extracted from circular ROIs on RVG images using PyRadiomics. Feature distributions were compared using Mann–Whitney U tests with false discovery rate correction, and classification was performed using a logistic regression model with nested cross-validation. Results: Forty-four radiomic texture features showed statistically significant differences between onlyCBCT and noLesion ROIs, predominantly with small to medium effect sizes. For a 40-pixel ROI radius, the classifier achieved a mean area under the ROC curve of 0.71, mean accuracy of 68%, and mean sensitivity of 73%. Smaller ROIs (20–40 pixels) yielded higher AUCs and substantially better accuracy than larger sampling regions (≥60 pixels). Conclusions: Quantifiable radiomic signatures of periapical pathology are present on conventional radiographs even when lesions are visually occult. Radiomics may serve as a complementary decision support tool for identifying CBCT-only periapical lesions in routine clinical imaging. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

24 pages, 2918 KB

Open AccessArticle

Quantifying Explainability in OCT Segmentation of Macular Holes and Cysts: A SHAP-Based Coverage and Factor Contribution Analysis

by İlknur Tuncer Fırat, Murat Fırat and Taner Tuncer

Diagnostics 2026, 16(1), 97; https://doi.org/10.3390/diagnostics16010097 - 27 Dec 2025

Viewed by 662

Abstract

Background: Optical coherence tomography (OCT) can quantify the morphology and dimensions of a macular hole for diagnosis and treatment planning. Objective: The aim of this study was to perform automatic segmentation of macular holes (MHs) and cysts from OCT macular volumes using [...] Read more.

Background: Optical coherence tomography (OCT) can quantify the morphology and dimensions of a macular hole for diagnosis and treatment planning. Objective: The aim of this study was to perform automatic segmentation of macular holes (MHs) and cysts from OCT macular volumes using a deep learning-based model and to quantitatively evaluate decision reliability using the model’s focus regions and GradientSHAP-based explainability. Methods: In this study, we automatically segmented MHs and cysts in OCT images from the open-access OIMHS dataset. The dataset comprises 125 eyes from 119 patients and 3859 OCT B-scans. OCT B-scan slices were input to a UNet-48-based model with a 2.5D stacking strategy. Performance was evaluated using Dice and intersection-over-union (IoU), boundary accuracy was evaluated using the 95th-percentile Hausdorff distance (HD95), and calibration was evaluated using the expected calibration error (ECE). Explainability was quantified from GradientSHAP maps using lesion coverage and spatial focus metrics: Attribution Precision in Lesion (APILτ), which is the proportion of attributions (SHAP contributions) falling inside the lesion; Attribution Recall in Lesion (ARILτ), which is the proportion of the true lesion covered by the attributions; and leakage (Leakτ = 1 − APILτ), which is the proportion of attributions falling outside the lesion. Spatial focus was monitored using the center-of-mass distance (COM-dist), which is the Euclidean distance between the attribution center and the segmentation center. All metrics were calculated using the top τ% of the pixels with the highest SHAP values. SHAP features were clustered using PCA and k-means. Explanations were calculated using the clinical mask in ground truth (GT) mode and the model segmentation in prediction (Pred) mode. Results: The Dice/IoU values for holes and cysts were 0.94/0.91 and 0.87/0.81, respectively. Across lesion classes, HD95 = 6 px and ECE = 0.008, indicating good boundary accuracy and calibration. In GT mode (τ = 20), three regimes were observed: (i) retina-dominant: high ARIL (hole: 0.659; cyst: 0.654), high Leak (hole: 0.983; cyst: 0.988), and low COM-dist (hole: 7.84 px; cyst: 6.91 px), with the focus lying within the retina and largely confined to the retinal tissue; (ii) peri-lesional: highest ARIL (hole: 0.684; cyst: 0.719), relatively lower Leak (hole: 0.917; cyst: 0.940), and medium/high COM-dist (hole: 16.22 px; cyst: 10.17 px), with the focus located around the lesion; (iii) narrow-coverage: primarily seen for cysts in GT mode (ARIL: 0.494; Leak: 1.000; COM-dist: 52.02 px), with markedly reduced coverage. In Pred mode, the ARIL20 for holes increased in the retina-dominant cluster (0.758) and COM-dist decreased (6.24 px), indicating better agreement with the model segmentation. Conclusions: The model exhibited high accuracy and good calibration for MH and cyst segmentation in OCT images. Quantitative characterization of SHAP validated the model results. In the clinic, peri-lesion and narrow-coverage conditions are the key situations that require careful interpretation. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

26 pages, 3269 KB

Open AccessArticle

DiagNeXt: A Two-Stage Attention-Guided ConvNeXt Framework for Kidney Pathology Segmentation and Classification

by Hilal Tekin, Şafak Kılıç and Yahya Doğan

J. Imaging 2025, 11(12), 433; https://doi.org/10.3390/jimaging11120433 - 4 Dec 2025

Cited by 2 | Viewed by 899

Abstract

Accurate segmentation and classification of kidney pathologies from medical images remain a major challenge in computer-aided diagnosis due to complex morphological variations, small lesion sizes, and severe class imbalance. This study introduces DiagNeXt, a novel two-stage deep learning framework designed to overcome these [...] Read more.

Accurate segmentation and classification of kidney pathologies from medical images remain a major challenge in computer-aided diagnosis due to complex morphological variations, small lesion sizes, and severe class imbalance. This study introduces DiagNeXt, a novel two-stage deep learning framework designed to overcome these challenges through an integrated use of attention-enhanced ConvNeXt architectures for both segmentation and classification. In the first stage, DiagNeXt-Seg employs a U-Net-based design incorporating Enhanced Convolutional Blocks (ECBs) with spatial attention gates and Atrous Spatial Pyramid Pooling (ASPP) to achieve precise multi-class kidney segmentation. In the second stage, DiagNeXt-Cls utilizes the segmented regions of interest (ROIs) for pathology classification through a hierarchical multi-resolution strategy enhanced by Context-Aware Feature Fusion (CAFF) and Evidential Deep Learning (EDL) for uncertainty estimation. The main contributions of this work include: (1) enhanced ConvNeXt blocks with large-kernel depthwise convolutions optimized for 3D medical imaging, (2) a boundary-aware compound loss combining Dice, cross-entropy, focal, and distance transform terms to improve segmentation precision, (3) attention-guided skip connections preserving fine-grained spatial details, (4) hierarchical multi-scale feature modeling for robust pathology recognition, and (5) a confidence-modulated classification approach integrating segmentation quality metrics for reliable decision-making. Extensive experiments on a large kidney CT dataset comprising 3847 patients demonstrate that DiagNeXt achieves 98.9% classification accuracy, outperforming state-of-the-art approaches by 6.8%. The framework attains near-perfect AUC scores across all pathology classes (Normal: 1.000, Tumor: 1.000, Cyst: 0.999, Stone: 0.994) while offering clinically interpretable uncertainty maps and attention visualizations. The superior diagnostic accuracy, computational efficiency (6.2× faster inference), and interpretability of DiagNeXt make it a strong candidate for real-world integration into clinical kidney disease diagnosis and treatment planning systems. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

13 pages, 2760 KB

Open AccessArticle

Deep Learning for Sex Estimation from Whole-Foot X-Rays: Benchmarking CNNs for Rapid Forensic Identification

by Rukiye Çiftçi, İpek Atik, Özgür Eken and Monira I. Aldhahi

Diagnostics 2025, 15(22), 2923; https://doi.org/10.3390/diagnostics15222923 - 19 Nov 2025

Viewed by 1178

Abstract

Background: Accurate sex estimation is crucial in forensic identification when DNA analysis is impractical or remains are fragmented. Traditional anthropometric approaches often rely on single bone measurements and yield moderate levels of accuracy. Objective: This study aimed to evaluate deep convolutional neural networks [...] Read more.

Background: Accurate sex estimation is crucial in forensic identification when DNA analysis is impractical or remains are fragmented. Traditional anthropometric approaches often rely on single bone measurements and yield moderate levels of accuracy. Objective: This study aimed to evaluate deep convolutional neural networks (CNNs) for automated sex estimation using entire foot radiographs, an approach rarely explored. Methods: Digital foot radiographs from 471 adults (238 men, 233 women, aged 18–65 years) without deformities or prior surgery were retrospectively collected at a single tertiary center. Six CNN architectures (AlexNet, ResNet-18, ResNet-50, ShuffleNet, GoogleNet, and InceptionV3) were trained using transfer learning (70/15/15 train–validation–test split, data augmentation). The model performance was assessed using accuracy, sensitivity, specificity, precision, and F1-score. Results: InceptionV3 achieved the highest accuracy (97.1%), surpassing previously reported methods (typically 72–89%), with balanced sensitivity (97.5%) and specificity (96.8%). ResNet-50 followed closely (95.7%), whereas simpler networks, such as AlexNet, underperformed (90%). Conclusions: Deep learning applied to whole-foot radiographs delivers state-of-the-art accuracy for sex estimation, enabling rapid, reproducible, and cost-effective forensic identification when DNA analysis is delayed or unavailable, such as in mass disasters or clinical emergency settings. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

21 pages, 14940 KB

Open AccessArticle

Clinical Application of Vision Transformers for Melanoma Classification: A Multi-Dataset Evaluation Study

by Antony Garcia, Jixing Zhou, Gabriela Pinero-Crespo, Thomas Beachkofsky and Xinming Huang

Cancers 2025, 17(21), 3447; https://doi.org/10.3390/cancers17213447 - 28 Oct 2025

Cited by 1 | Viewed by 1914

Abstract

Background: Melanoma is one of the most lethal skin cancers, with survival rates largely dependent on early detection, yet diagnosis remains difficult because of its visual similarity to benign nevi. Convolutional neural networks have achieved strong performance in dermoscopic analysis but often [...] Read more.

Background: Melanoma is one of the most lethal skin cancers, with survival rates largely dependent on early detection, yet diagnosis remains difficult because of its visual similarity to benign nevi. Convolutional neural networks have achieved strong performance in dermoscopic analysis but often depend on fixed input sizes and local features, which can limit generalization. Vision Transformers, which capture global image relationships through self-attention, offer a promising alternative. Methods: A ViT-L/16 model was fine-tuned using the ISIC 2019 dataset containing more than 25,000 dermoscopic images. To expand the dataset and balance class representation, synthetic melanoma and nevus images were produced with StyleGAN2-ADA, retaining only high-confidence outputs. Model performance was evaluated on an external biopsy-confirmed dataset (MN187) and compared with CNN baselines (ResNet-152, DenseNet-201, EfficientNet-B7, ConvNeXt-XL, ViT-B/16) and the commercial MoleAnalyzer Pro system using ROC-AUC and DeLong’s test. Results: The ViT-L/16 model reached a baseline ROC-AUC of 0.902 on MN187, surpassing all CNN baselines and the MoleAnalyzer Pro system, though the difference was not statistically significant (p = 0.07). After adding 46,000 confidence-filtered GAN-generated images, the ROC-AUC increased to 0.915, giving a statistically significant improvement over the commercial MoleAnalyzer Pro system (p = 0.032). Conclusions: Vision Transformers show strong potential for melanoma classification, especially when combined with GAN-based augmentation, offering advantages in global feature representation and data expansion that support the development of dependable AI-driven clinical decision-support systems. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

20 pages, 5063 KB

Open AccessArticle

AI Diffusion Models Generate Realistic Synthetic Dental Radiographs Using a Limited Dataset

by Brian Kirkwood, Byeong Yeob Choi, James Bynum and Jose Salinas

J. Imaging 2025, 11(10), 356; https://doi.org/10.3390/jimaging11100356 - 11 Oct 2025

Viewed by 1970

Abstract

Generative Artificial Intelligence (AI) has the potential to address the limited availability of dental radiographs for the development of Dental AI systems by creating clinically realistic synthetic dental radiographs (SDRs). Evaluation of artificially generated images requires both expert review and objective measures of [...] Read more.

Generative Artificial Intelligence (AI) has the potential to address the limited availability of dental radiographs for the development of Dental AI systems by creating clinically realistic synthetic dental radiographs (SDRs). Evaluation of artificially generated images requires both expert review and objective measures of fidelity. A stepwise approach was used to processing 10,000 dental radiographs. First, a single dentist screened images to determine if specific image selection criterion was met; this identified 225 images. From these, 200 images were randomly selected for training an AI image generation model. Second, 100 images were randomly selected from the previous training dataset and evaluated by four dentists; the expert review identified 57 images that met image selection criteria to refine training for two additional AI models. The three models were used to generate 500 SDRs each and the clinical realism of the SDRs was assessed through expert review. In addition, the SDRs generated by each model were objectively evaluated using quantitative metrics: Fréchet Inception Distance (FID) and Kernel Inception Distance (KID). Evaluation of the SDR by a dentist determined that expert-informed curation improved SDR realism, and refinement of model architecture produced further gains. FID and KID analysis confirmed that expert input and technical refinement improve image fidelity. The convergence of subjective and objective assessments strengthens confidence that the refined model architecture can serve as a foundation for SDR image generation, while highlighting the importance of expert-informed data curation and domain-specific evaluation metrics. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

13 pages, 1023 KB

Open AccessArticle

Validation of an Artificial Intelligence Model for Breast Cancer Molecular Subtyping Using Hematoxylin and Eosin-Stained Whole-Slide Images in a Population-Based Cohort

by Umay Kiraz, Claudio Fernandez-Martin, Emma Rewcastle, Einar G. Gudlaugsson, Ivar Skaland, Valery Naranjo, Sandra Morales-Martinez and Emiel A. M. Janssen

Cancers 2025, 17(19), 3234; https://doi.org/10.3390/cancers17193234 - 5 Oct 2025

Cited by 3 | Viewed by 1511

Abstract

Background/Objectives: Breast cancer (BC) is the most commonly diagnosed cancer in women and the leading cause of cancer-related deaths globally. Molecular subtyping is crucial for prognosis and treatment planning, with immunohistochemistry (IHC) being the most commonly used method. However, IHC has limitations, [...] Read more.

Background/Objectives: Breast cancer (BC) is the most commonly diagnosed cancer in women and the leading cause of cancer-related deaths globally. Molecular subtyping is crucial for prognosis and treatment planning, with immunohistochemistry (IHC) being the most commonly used method. However, IHC has limitations, including observer variability, a lack of standardization, and a lack of reproducibility. Gene expression profiling is considered the ground truth for molecular subtyping; unfortunately, this is expensive and inaccessible to many institutions. This study investigates the potential of an artificial intelligence (AI) model to predict BC molecular subtypes directly from hematoxylin and eosin (H&E)-stained whole-slide images (WSIs). Methods: A pretrained deep learning framework based on multiple-instance learning (MIL) was validated on the Stavanger Breast Cancer (SBC) dataset, consisting of 538 BC cases. Three classification tasks were assessed, including two-class [triple negative BC (TNBC) vs. non-TNBC], three-class (luminal vs. HER2-positive vs. TNBC), and four-class (luminal A vs. luminal B vs. HER2-positive vs. TNBC) groups. Performance metrics were used for the evaluation of the AI model. Results: The AI model demonstrated strong performance in distinguishing TNBC from non-TNBC (AUC = 0.823, accuracy = 0.833, F1-score = 0.824). However, performance declined with an increasing number of classes. Conclusions: The study highlights the potential of AI in BC molecular subtyping from H&E WSIs, offering an easily applicable and standardized method to IHC. Future improvements should focus on optimizing multi-class classification and validating AI-based methods against gene expression analyses for enhanced clinical applicability. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

15 pages, 2112 KB

Open AccessArticle

Radiomics-Based Preoperative Assessment of Muscle-Invasive Bladder Cancer Using Combined T2 and ADC MRI: A Multicohort Validation Study

by Dmitry Kabanov, Natalia Rubtsova, Aleksandra Golbits, Andrey Kaprin, Valentin Sinitsyn and Mikhail Potievskiy

J. Imaging 2025, 11(10), 342; https://doi.org/10.3390/jimaging11100342 - 1 Oct 2025

Cited by 2 | Viewed by 1292

Abstract

Accurate preoperative staging of bladder cancer on MRI remains challenging because visual reads vary across observers. We investigated a multiparametric MRI (mpMRI) radiomics approach to predict muscle invasion (≥T2) and prospectively tested it on a validation cohort. Eighty-four patients with urothelial carcinoma underwent [...] Read more.

Accurate preoperative staging of bladder cancer on MRI remains challenging because visual reads vary across observers. We investigated a multiparametric MRI (mpMRI) radiomics approach to predict muscle invasion (≥T2) and prospectively tested it on a validation cohort. Eighty-four patients with urothelial carcinoma underwent 1.5-T mpMRI per VI-RADS (T2-weighted imaging and DWI-derived ADC maps). Two blinded radiologists performed 3D tumor segmentation; 37 features per sequence were extracted (LifeX) using absolute resampling. In the training cohort (n = 40), features that differed between non-muscle-invasive and muscle-invasive tumors (Mann–Whitney p < 0.05) underwent ROC analysis with cut-offs defined by the Youden index. A compact descriptor combining GLRLM-LRLGE from T2 and GLRLM-SRLGE from ADC was then fixed and applied without re-selection to a prospective validation cohort (n = 44). Histopathology within 6 weeks—TURBT or cystectomy—served as the reference. Eleven T2-based and fifteen ADC-based features pointed to invasion; DWI texture features were not informative. The descriptor yielded AUCs of 0.934 (training) and 0.871 (validation) with 85.7% sensitivity and 96.2% specificity in validation. Collectively, these findings indicate that combined T2/ADC radiomics can provide high diagnostic accuracy and may serve as a useful decision support tool, after multicenter, multi-vendor validation. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

13 pages, 1587 KB

Open AccessArticle

Glioma Grading by Integrating Radiomic Features from Peritumoral Edema in Fused MRI Images and Automated Machine Learning

by Amir Khorasani

J. Imaging 2025, 11(10), 336; https://doi.org/10.3390/jimaging11100336 - 27 Sep 2025

Cited by 1 | Viewed by 1326

Abstract

We aimed to investigate the utility of peritumoral edema-derived radiomic features from magnetic resonance imaging (MRI) image weights and fused MRI sequences for enhancing the performance of machine learning-based glioma grading. The present study utilized the Multimodal Brain Tumor Segmentation Challenge 2023 (BraTS [...] Read more.

We aimed to investigate the utility of peritumoral edema-derived radiomic features from magnetic resonance imaging (MRI) image weights and fused MRI sequences for enhancing the performance of machine learning-based glioma grading. The present study utilized the Multimodal Brain Tumor Segmentation Challenge 2023 (BraTS 2023) dataset. Laplacian Re-decomposition (LRD) was employed to fuse multimodal MRI sequences. The fused image quality was evaluated using the Entropy, standard deviation (STD), peak signal-to-noise ratio (PSNR), and structural similarity index measure (SSIM) metrics. A comprehensive set of radiomic features was subsequently extracted from peritumoral edema regions using PyRadiomics. The Boruta algorithm was applied for feature selection, and an optimized classification pipeline was developed using the Tree-based Pipeline Optimization Tool (TPOT). Model performance for glioma grade classification was evaluated based on accuracy, precision, recall, F1-score, and area under the curve (AUC) parameters. Analysis of fused image quality metrics confirmed that the LRD method produces high-quality fused images. From 851 radiomic features extracted from peritumoral edema regions, the Boruta algorithm selected different sets of informative features in both standard MRI and fused images. Subsequent TPOT automated machine learning optimization analysis identified a fine-tuned Stochastic Gradient Descent (SGD) classifier, trained on features from T₁Gd+FLAIR fused images, as the top-performing model. This model achieved superior performance in glioma grade classification (Accuracy = 0.96, Precision = 1.0, Recall = 0.94, F1-Score = 0.96, AUC = 1.0). Radiomic features derived from peritumoral edema in fused MRI images using the LRD method demonstrated distinct, grade-specific patterns and can be utilized as a non-invasive, accurate, and rapid glioma grade classification method. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

24 pages, 2959 KB

Open AccessArticle

From Detection to Diagnosis: An Advanced Transfer Learning Pipeline Using YOLO11 with Morphological Post-Processing for Brain Tumor Analysis for MRI Images

by Ikram Chourib

J. Imaging 2025, 11(8), 282; https://doi.org/10.3390/jimaging11080282 - 21 Aug 2025

Cited by 4 | Viewed by 3819

Abstract

Accurate and timely detection of brain tumors from magnetic resonance imaging (MRI) scans is critical for improving patient outcomes and informing therapeutic decision-making. However, the complex heterogeneity of tumor morphology, scarcity of annotated medical data, and computational demands of deep learning models present [...] Read more.

Accurate and timely detection of brain tumors from magnetic resonance imaging (MRI) scans is critical for improving patient outcomes and informing therapeutic decision-making. However, the complex heterogeneity of tumor morphology, scarcity of annotated medical data, and computational demands of deep learning models present substantial challenges for developing reliable automated diagnostic systems. In this study, we propose a robust and scalable deep learning framework for brain tumor detection and classification, built upon an enhanced YOLO-v11 architecture combined with a two-stage transfer learning strategy. The first stage involves training a base model on a large, diverse MRI dataset. Upon achieving a mean Average Precision (mAP) exceeding 90%, this model is designated as the Brain Tumor Detection Model (BTDM). In the second stage, the BTDM is fine-tuned on a structurally similar but smaller dataset to form Brain Tumor Detection and Segmentation (BTDS), effectively leveraging domain transfer to maintain performance despite limited data. The model is further optimized through domain-specific data augmentation—including geometric transformations—to improve generalization and robustness. Experimental evaluations on publicly available datasets show that the framework achieves high mAP@0.5 scores (up to 93.5% for the BTDM and 91% for BTDS) and consistently outperforms existing state-of-the-art methods across multiple tumor types, including glioma, meningioma, and pituitary tumors. In addition, a post-processing module enhances interpretability by generating segmentation masks and extracting clinically relevant metrics such as tumor size and severity level. These results underscore the potential of our approach as a high-performance, interpretable, and deployable clinical decision-support tool, contributing to the advancement of intelligent real-time neuro-oncological diagnostics. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

23 pages, 5155 KB

Open AccessArticle

Enhancing Early Detection of Diabetic Foot Ulcers Using Deep Neural Networks

by A. Sharaf Eldin, Asmaa S. Ahmoud, Hanaa M. Hamza and Hanin Ardah

Diagnostics 2025, 15(16), 1996; https://doi.org/10.3390/diagnostics15161996 - 9 Aug 2025

Cited by 4 | Viewed by 2822

Abstract

Background/Objectives: Diabetic foot ulcers (DFUs) remain a critical complication of diabetes, with high rates of amputation when not diagnosed early. Despite advancements in medical imaging, current DFU detection methods are often limited by their computational complexity, poor generalizability, and delayed diagnostic performance. [...] Read more.

Background/Objectives: Diabetic foot ulcers (DFUs) remain a critical complication of diabetes, with high rates of amputation when not diagnosed early. Despite advancements in medical imaging, current DFU detection methods are often limited by their computational complexity, poor generalizability, and delayed diagnostic performance. This study presents a novel hybrid diagnostic framework that integrates traditional feature extraction methods with deep learning (DL) to improve the early real-time computer-aided detection (CAD) of DFUs. Methods: The proposed model leverages plantar thermograms to detect early thermal asymmetries associated with DFUs. It uniquely combines the oriented FAST and rotated BRIEF (ORB) algorithm with the Bag of Features (BOF) method to extract robust handcrafted features while also incorporating deep features from pretrained convolutional neural networks (ResNet50, AlexNet, and EfficientNet). These features were fused and input into a lightweight deep neural network (DNN) classifier designed for binary classification. Results: Our model demonstrated an accuracy of 98.51%, precision of 100%, sensitivity of 98.98%, and AUC of 1.00 in a publicly available plantar thermogram dataset (n = 1670 images). An ablation study confirmed the superiority of ORB + DL fusion over standalone approaches. Unlike previous DFU detection models that rely solely on either handcrafted or deep features, our study presents the first lightweight hybrid framework that integrates ORB-based descriptors with deep CNN representations (e.g., ResNet50 and EfficientNet). Compared with recent state-of-the-art models, such as DFU_VIRNet and DFU_QUTNet, our approach achieved a higher diagnostic performance (accuracy = 98.51%, AUC = 1.00) while maintaining real-time capability and a lower computational overhead, making it highly suitable for clinical deployment. Conclusions: This study proposes the first integration of ORB-based handcrafted features with deep neural representations for DFU detection from thermal images. The model delivers high accuracy, robustness to noise, and real-time capabilities, outperforming existing state-of-the-art approaches and demonstrating strong potential for clinical deployment. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

24 pages, 4892 KB

Open AccessArticle

Diffusion Model-Based Augmentation Using Asymmetric Attention Mechanisms for Cardiac MRI Images

by Mertcan Özdemir and Osman Eroğul

Diagnostics 2025, 15(16), 1985; https://doi.org/10.3390/diagnostics15161985 - 8 Aug 2025

Cited by 1 | Viewed by 1393

Abstract

Background: The limited availability of cardiac MRI data significantly constrains deep learning applications in cardiovascular imaging, necessitating innovative approaches to address data scarcity while preserving critical cardiac anatomical features. Methods: We developed a specialized denoising diffusion probabilistic model incorporating an attention-enhanced UNet architecture [...] Read more.

Background: The limited availability of cardiac MRI data significantly constrains deep learning applications in cardiovascular imaging, necessitating innovative approaches to address data scarcity while preserving critical cardiac anatomical features. Methods: We developed a specialized denoising diffusion probabilistic model incorporating an attention-enhanced UNet architecture with strategically placed attention blocks across five hierarchical levels. The model was trained and evaluated on the OCMR dataset and compared against state-of-the-art generative approaches including StyleGAN2-ADA, WGAN-GP, and VAE baselines. Results: Our approach achieved superior image quality with a Fréchet Inception Distance of 77.78, significantly outperforming StyleGAN2-ADA (117.70), WGAN-GP (227.98), and VAE (325.26). Structural similarity metrics demonstrated excellent performance (SSIM: 0.720 ± 0.143; MS-SSIM: 0.925 ± 0.069). Clinical validation by cardiac radiologists yielded discrimination accuracy of only 60.0%, indicating near-realistic image quality that is challenging for experts to distinguish from real images. Comprehensive anatomical analysis revealed that 13 of 20 cardiac metrics showed no significant differences between real and synthetic images, with particularly strong preservation of left ventricular features. Discussion: The generated synthetic images demonstrate high anatomical fidelity with expert-level quality, as evidenced by the difficulty radiologists experienced in distinguishing synthetic from real images. The strong preservation of cardiac anatomical features, particularly left ventricular characteristics, indicates the model’s potential for medical image analysis applications. Conclusions: This work establishes diffusion models as a robust solution for cardiac MRI data augmentation, successfully generating anatomically accurate synthetic images that enhance downstream clinical applications while maintaining diagnostic fidelity. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

22 pages, 4079 KB

Open AccessArticle

Breast Cancer Classification with Various Optimized Deep Learning Methods

by Mustafa Güler, Gamze Sart, Ömer Algorabi, Ayse Nur Adıguzel Tuylu and Yusuf Sait Türkan

Diagnostics 2025, 15(14), 1751; https://doi.org/10.3390/diagnostics15141751 - 10 Jul 2025

Cited by 7 | Viewed by 2297

Abstract

Background/Objectives: In recent years, there has been a significant increase in the number of women with breast cancer. Breast cancer prediction is defined as a medical data analysis and image processing problem. Experts may need artificial intelligence technologies to distinguish between benign and [...] Read more.

Background/Objectives: In recent years, there has been a significant increase in the number of women with breast cancer. Breast cancer prediction is defined as a medical data analysis and image processing problem. Experts may need artificial intelligence technologies to distinguish between benign and malignant tumors in order to make decisions. When the studies in the literature are examined, it can be seen that applications of deep learning algorithms in the field of medicine have achieved very successful results. Methods: In this study, 11 different deep learning algorithms (Vanilla, ResNet50, ResNet152, VGG16, DenseNet152, MobileNetv2, EfficientB1, NasNet, DenseNet201, ensemble, and Tuned Model) were used. Images of pathological specimens from breast biopsies consisting of two classes, benign and malignant, were used for classification analysis. To limit the computational time and speed up the analysis process, 10,000 images, 6172 IDC-negative and 3828 IDC-positive, were selected. Of the images, 80% were used for training, 10% were used for validation, and 10% were used for testing the trained model. Results: The results demonstrate that DenseNet201 achieved the highest classification accuracy of 89.4%, with a precision of 88.2%, a recall of 84.1%, an F1 score of 86.1%, and an AUC score of 95.8%. Conclusions: In conclusion, this study highlights the potential of deep learning algorithms in breast cancer classification. Future research should focus on integrating multi-modal imaging data, refining ensemble learning methodologies, and expanding dataset diversity to further improve the classification accuracy and real-world clinical applicability. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

12 pages, 2782 KB

Open AccessArticle

Platelets Image Classification Through Data Augmentation: A Comparative Study of Traditional Imaging Augmentation and GAN-Based Synthetic Data Generation Techniques Using CNNs

by Itunuoluwa Abidoye, Frances Ikeji, Charlie A. Coupland, Simon D. J. Calaminus, Nick Sander and Eva Sousa

J. Imaging 2025, 11(6), 183; https://doi.org/10.3390/jimaging11060183 - 4 Jun 2025

Cited by 5 | Viewed by 2636

Abstract

Platelets play a crucial role in diagnosing and detecting various diseases, influencing the progression of conditions and guiding treatment options. Accurate identification and classification of platelets are essential for these purposes. The present study aims to create a synthetic database of platelet images [...] Read more.

Platelets play a crucial role in diagnosing and detecting various diseases, influencing the progression of conditions and guiding treatment options. Accurate identification and classification of platelets are essential for these purposes. The present study aims to create a synthetic database of platelet images using Generative Adversarial Networks (GANs) and validate its effectiveness by comparing it with datasets of increasing sizes generated through traditional augmentation techniques. Starting from an initial dataset of 71 platelet images, the dataset was expanded to 141 images (Level 1) using random oversampling and basic transformations and further to 1463 images (Level 2) through extensive augmentation (rotation, shear, zoom). Additionally, a synthetic dataset of 300 images was generated using a Wasserstein GAN with Gradient Penalty (WGAN-GP). Eight pre-trained deep learning models (DenseNet121, DenseNet169, DenseNet201, VGG16, VGG19, InceptionV3, InceptionResNetV2, and AlexNet) and two custom CNNs were evaluated across these datasets. Performance was measured using accuracy, precision, recall, and F1-score. On the extensively augmented dataset (Level 2), InceptionV3 and InceptionResNetV2 reached 99% accuracy and 99% precision/recall/F1-score, while DenseNet201 closely followed, with 98% accuracy, precision, recall and F1-score. GAN-augmented data further improved DenseNet’s performance, demonstrating the potential of GAN-generated images in enhancing platelet classification, especially where data are limited. These findings highlight the benefits of combining traditional and GAN-based augmentation techniques to improve classification performance in medical imaging tasks. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

18 pages, 3703 KB

Open AccessArticle

The Value of PET/CT-Based Radiomics in Predicting Adrenal Metastases in Patients with Cancer

by Qiujun He, Xiangxing Kong, Xiangxi Meng, Xiuling Shen and Nan Li

Diagnostics 2025, 15(11), 1356; https://doi.org/10.3390/diagnostics15111356 - 28 May 2025

Cited by 2 | Viewed by 1876

Abstract

Objectives: Differentiation of adrenal incidentalomas (AIs) remains a challenge in the oncological setting. The aim of the study was to explore the diagnostic efficacy of [18F]Fluorodeoxyglucose (FDG) positron emission tomography combined with computed tomography (PET/CT)-based radiomics in identifying adrenal metastases and to compare [...] Read more.

Objectives: Differentiation of adrenal incidentalomas (AIs) remains a challenge in the oncological setting. The aim of the study was to explore the diagnostic efficacy of [18F]Fluorodeoxyglucose (FDG) positron emission tomography combined with computed tomography (PET/CT)-based radiomics in identifying adrenal metastases and to compare it with that of conventional PET/CT parameters. Materials: Retrospective analysis was performed on 195 AIs for model construction, nomogram drawing, and internal validation. An additional 30 AIs were collected for external validation of the radiomics model and nomogram. Logistic regression analysis was employed to build models based on clinical and PET/CT routine parameters. The open-source software Python (version 3.7.11) was utilized to process the regions of interest (ROI) delineated by ITK-SNAP, extracting radiomic features. Least absolute shrinkage and selection operator (LASSO) regression analysis was applied for feature selection. Based on the selected features, the optimal model was chosen from ten machine learning algorithms, and the nomogram was constructed. Results: The area under the curve (AUC), sensitivity, specificity, and accuracy of conventional parameters of PET/CT were 0.919, 0.849, 0.892, and 0.844, respectively. XGBoost demonstrated superior diagnostic efficiency among the radiomics models, outperforming those constructed using independent predictors. The AUC, accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of XGBoost’s internal and external validation were 0.945, 0.932, 0.930, 0.960, 0.970, 0.890 and 0.910, 0.900, 0.860, 1, 1, 0.750. The accuracy, sensitivity, specificity, PPV, and NPV of the nomogram in external validation were 0.870, 0.952, 0.667, 0.870, and 0.857. Conclusions: The radiomics model and conventional PET/CT parameters both showed high diagnostic performance (AUC p > 0.05) in discriminating adrenal metastases from benign lesions, offering a practical, non-invasive approach for clinical assessment. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

14 pages, 4088 KB

Open AccessArticle

Is a 3-Minute Knee MRI Protocol Sufficient for Daily Clinical Practice? A SuperResolution Reconstruction Approach Using AI and Compressed Sensing

by Robert Hahnfeldt, Robert Terzis, Thomas Dratsch, Lajos Maximilian Basten, Philip Rauen, Johannes Oppermann, David Grevenstein, Jan Paul Janßen, Nour El-Hoda Abou Zeid, Kristina Sonnabend, Christoph Katemann, Stephan Skornitzke, David Maintz, Jonathan Kottlors, Grischa Bratke and Andra-Iza Iuga

Diagnostics 2025, 15(10), 1206; https://doi.org/10.3390/diagnostics15101206 - 9 May 2025

Cited by 5 | Viewed by 2512

Abstract

Objectives: The purpose of this study was to assess whether a 3-min 2D knee protocol can meet the needs for clinical application if using a SuperResolution reconstruction approach. Methods: In this prospective study, a total of 20 volunteers underwent imaging of the knee [...] Read more.

Objectives: The purpose of this study was to assess whether a 3-min 2D knee protocol can meet the needs for clinical application if using a SuperResolution reconstruction approach. Methods: In this prospective study, a total of 20 volunteers underwent imaging of the knee using a 3T MRI scanner (Philips Ingenia Elition X 3.0T, Philips). The imaging protocol, consisting of a fat-saturated 2D proton density sequence in coronal, sagittal, and transverse orientations, as well as a sagittal T1-weighted sequence, was acquired with standard and ultra-low resolution. The standard sequences were reconstructed using an AI-assisted Compressed SENSE method (SmartSpeed). The ultra-low-resolution sequences have been reconstructed using a vendor-provided prototype. Four experienced readers (two radiologists and two orthopedic surgeons) evaluated the sequences for image quality, anatomical structures, and incidental pathologies. The consensus evaluation of two different experienced radiologists specialized in musculoskeletal imaging served as the gold standard. Results: The acquisition time for the entire protocol was 11:01 min for standard resolution and 03:36 min for ultra-low resolution. In the overall assessment, CS-SuperRes-reconstructed sequences showed slightly improved accuracy and increased specificity compared to the standard CS-AI method (0.87 vs. 0.86 and 0.9 vs. 0.87, respectively), while the standard method exhibited a higher sensitivity (0.73 vs. 0.57). Overall, 24 out of 40 pathologies were detected in the ultra-low-resolution images compared to 26 in the standard images. Conclusions: The CS-SuperRes method enables a 2D knee protocol to be completed in 3 min, with improved accuracy compared to the clinical standard. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

14 pages, 2257 KB

Open AccessArticle

Automatic Analysis of Ultrasound Images to Estimate Subcutaneous and Visceral Fat and Muscle Tissue in Patients with Suspected Malnutrition

by Antonio Cuesta-Vargas, José María Arjona-Caballero, Gabriel Olveira, Daniel de Luis Román, Diego Bellido-Guerrero and Jose Manuel García-Almeida

Diagnostics 2025, 15(8), 988; https://doi.org/10.3390/diagnostics15080988 - 13 Apr 2025

Cited by 2 | Viewed by 2031

Abstract

Background: Malnutrition is a prevalent condition associated with adverse health outcomes, requiring the accurate assessment of muscle composition and fat distribution. Methods: This study presents a novel method for the automatic analysis of ultrasound images to estimate subcutaneous and visceral fat, as well [...] Read more.

Background: Malnutrition is a prevalent condition associated with adverse health outcomes, requiring the accurate assessment of muscle composition and fat distribution. Methods: This study presents a novel method for the automatic analysis of ultrasound images to estimate subcutaneous and visceral fat, as well as muscle, in patients with suspected malnutrition. The proposed system utilizes computer vision techniques to segment regions of interest (ROIs), calculate relevant variables, and store data for clinical evaluation. Unlike traditional segmentation methods that rely solely on thresholding or pre-defined masks, our method employs an iterative hierarchical approach to refine contour detection and improve localization accuracy. A dataset of abdominal and leg ultrasound images, captured in both longitudinal and transversal planes, was analyzed. Results: Results showed higher precision for longitudinal scans compared to transversal scans, particularly for length-related variables, with the Y-axis Vastus intermediate achieving a precision of 92.87%. However, area-based measurements demonstrated lower precision due to differences between manual adjustments by experts and automatic geometric approximations. Conclusions: These findings highlight the system’s potential for clinical use while emphasizing the need for further algorithmic refinements to improve precision in area calculations. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

25 pages, 8439 KB

Open AccessArticle

Validation of Replicable Pipeline 3D Surface Reconstruction for Patient-Specific Abdominal Aortic Lumen Diagnostics

by Edoardo Ugolini, Giorgio La Civita, Moad Al Aidroos, Samuele Salti, Giuseppe Lisanti, Emanuele Ghedini, Gianluca Faggioli, Mauro Gargiulo and Giovanni Rossi

BioMed 2025, 5(2), 9; https://doi.org/10.3390/biomed5020009 - 25 Mar 2025

Cited by 1 | Viewed by 2542

Abstract

Background: Accurate prognoses are challenging in high-risk vascular conditions, such as abdominal aortic aneurysms, and limited diagnostic standards, decision-making criteria, and data semantics often hinder clinical reliability and impede diagnostics’ digital transition. This study aims to evaluate the performance, robustness, and usability of [...] Read more.

Background: Accurate prognoses are challenging in high-risk vascular conditions, such as abdominal aortic aneurysms, and limited diagnostic standards, decision-making criteria, and data semantics often hinder clinical reliability and impede diagnostics’ digital transition. This study aims to evaluate the performance, robustness, and usability of an automatic, replicable pipeline for aortic lumen surface reconstruction for pathological vessels. The goal is to provide a solid tool for geometric reconstruction to a more complex enhanced diagnostic framework. Methods: A U-Net convolutional neural network is trained using preoperative CTA scans, with 101 for model training and 14 for model testing, covering a wide anatomical and aortoiliac pathology spectrum. Validation included segmentation metric, robustness, reliability, and usability assessments. Performances are investigated by means of the test set’s prediction metrics for several instances of the model’s input. Clinical reliability is evaluated based on manual measurements performed by a vascular surgeon on the obtained 3D aortic lumen surfaces. Results: The test set is selected to cover a wide portion of aortoiliac pathologies. The algorithm demonstrated robustness with an average F1-score of 0.850 ± 0.120 and an intersection over union score of 0.760 ± 0.150 in the test set. Clinical reliability is assessed using the mean absolute errors for diameter and length measurements, respectively, of 1.73 mm and 2.27 mm. The 3D surface reconstruction demonstrated reliability, low processing times, and clinically valid reconstructions. Conclusions: The proposed algorithm can correctly reconstruct pathological vessels. Secondary aortoiliac pathologies are detected properly for challenging anatomies. To conclude, the proposed 3D reconstruction application to a digital, patient-specific diagnostic tool is, therefore, possible. Automatic replicable pipelines ensured the usability of the model’s outputs. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Graphical abstract

27 pages, 721 KB

Open AccessReview

Lights and Shadows on Artificial Intelligence in Glaucoma: Transforming Screening, Monitoring, and Prognosis

by Alessio Martucci, Gabriele Gallo Afflitto, Giulio Pocobelli, Francesco Aiello, Raffaele Mancino and Carlo Nucci

J. Clin. Med. 2025, 14(7), 2139; https://doi.org/10.3390/jcm14072139 - 21 Mar 2025

Cited by 9 | Viewed by 2975

Abstract

Background/Objectives: Artificial intelligence (AI) is increasingly being integrated into medicine, including ophthalmology, owing to its strong capabilities in image recognition. Methods: This review focuses on the most recent key applications of AI in the diagnosis and management of, as well as [...] Read more.

Background/Objectives: Artificial intelligence (AI) is increasingly being integrated into medicine, including ophthalmology, owing to its strong capabilities in image recognition. Methods: This review focuses on the most recent key applications of AI in the diagnosis and management of, as well as research on, glaucoma by performing a systematic review of the latest papers in the literature. Results: In glaucoma, AI can help analyze large amounts of data from diagnostic tools, such as fundus images, optical coherence tomography scans, and visual field tests. Conclusions: AI technologies can enhance the accuracy of glaucoma diagnoses and could provide significant economic benefits by automating routine tasks, improving diagnostic accuracy, and enhancing access to care, especially in underserved areas. However, despite these promising results, challenges persist, including limited dataset size and diversity, class imbalance, the need to optimize models for early detection, and the integration of multimodal data into clinical practice. Currently, ophthalmologists are expected to continue playing a leading role in managing glaucomatous eyes and overseeing the development and validation of AI tools. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

18 pages, 7165 KB

Open AccessArticle

Novel Preprocessing-Based Sequence for Comparative MR Cervical Lymph Node Segmentation

by Elif Ayten Tarakçı, Metin Çeliker, Mehmet Birinci, Tuğba Yemiş, Oğuz Gül, Enes Faruk Oğuz, Merve Solak, Esat Kaba, Fatma Beyazal Çeliker, Zerrin Özergin Coşkun, Ahmet Alkan and Özlem Çelebi Erdivanlı

J. Clin. Med. 2025, 14(6), 1802; https://doi.org/10.3390/jcm14061802 - 7 Mar 2025

Viewed by 1359

Abstract

Background and Objective: This study aims to utilize deep learning methods for the automatic segmentation of cervical lymph nodes in magnetic resonance images (MRIs), enhancing the speed and accuracy of diagnosing pathological masses in the neck and improving patient treatment processes. Materials [...] Read more.

Background and Objective: This study aims to utilize deep learning methods for the automatic segmentation of cervical lymph nodes in magnetic resonance images (MRIs), enhancing the speed and accuracy of diagnosing pathological masses in the neck and improving patient treatment processes. Materials and Methods: This study included 1346 MRI slices from 64 patients undergoing cervical lymph node dissection, biopsy, and preoperative contrast-enhanced neck MRI. A preprocessing model was used to crop and highlight lymph nodes, along with a method for automatic re-cropping. Two datasets were created from the cropped images—one with augmentation and one without—divided into 90% training and 10% validation sets. After preprocessing, the ResNet-50 images in the DeepLabv3+ encoder block were automatically segmented. Results: According to the results of the validation set, the mean IoU values for the DWI, T2, T1, T1+C, and ADC sequences in the dataset without augmentation created for cervical lymph node segmentation were 0.89, 0.88, 0.81, 0.85, and 0.80, respectively. In the augmented dataset, the average IoU values for all sequences were 0.91, 0.89, 0.85, 0.88, and 0.84. The DWI sequence showed the highest performance in the datasets with and without augmentation. Conclusions: Our preprocessing-based deep learning architectures successfully segmented cervical lymph nodes with high accuracy. This study is the first to explore automatic segmentation of the cervical lymph nodes using comprehensive neck MRI sequences. The proposed model can streamline the detection process, reducing the need for radiology expertise. Additionally, it offers a promising alternative to manual segmentation in radiotherapy, potentially enhancing treatment effectiveness. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

27 pages, 23884 KB

Open AccessArticle

GM-CBAM-ResNet: A Lightweight Deep Learning Network for Diagnosis of COVID-19

by Junjiang Zhu, Yihui Zhang, Cheng Ma, Jiaming Wu, Xuchen Wang and Dongdong Kong

J. Imaging 2025, 11(3), 76; https://doi.org/10.3390/jimaging11030076 - 3 Mar 2025

Cited by 2 | Viewed by 3839

Abstract

COVID-19 can cause acute infectious diseases of the respiratory system, and may probably lead to heart damage, which will seriously threaten human health. Electrocardiograms (ECGs) have the advantages of being low cost, non-invasive, and radiation free, and is widely used for evaluating heart [...] Read more.

COVID-19 can cause acute infectious diseases of the respiratory system, and may probably lead to heart damage, which will seriously threaten human health. Electrocardiograms (ECGs) have the advantages of being low cost, non-invasive, and radiation free, and is widely used for evaluating heart health status. In this work, a lightweight deep learning network named GM-CBAM-ResNet is proposed for diagnosing COVID-19 based on ECG images. GM-CBAM-ResNet is constructed by replacing the convolution module with the Ghost module (GM) and adding the convolutional block attention module (CBAM) in the residual module of ResNet. To reveal the superiority of GM-CBAM-ResNet, the other three methods (ResNet, GM-ResNet, and CBAM-ResNet) are also analyzed from the following aspects: model performance, complexity, and interpretability. The model performance is evaluated by using the open ‘ECG Images dataset of Cardiac and COVID-19 Patients’. The complexity is reflected by comparing the number of model parameters. The interpretability is analyzed by utilizing Gradient-weighted Class Activation Mapping (Grad-CAM). Parameter statistics indicate that, on the basis of ResNet19, the number of model parameters of GM-CBAM-ResNet19 is reduced by 45.4%. Experimental results show that, under less model complexity, GM-CBAM-ResNet19 improves the diagnostic accuracy by approximately 5% in comparison with ResNet19. Additionally, the interpretability analysis shows that CBAM can suppress the interference of grid backgrounds and ensure higher diagnostic accuracy under lower model complexity. This work provides a lightweight solution for the rapid and accurate diagnosing of COVD-19 based on ECG images, which holds significant practical deployment value. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

12 pages, 6563 KB

Open AccessArticle

Assessing Image Quality in Multiplexed Sensitivity-Encoding Diffusion-Weighted Imaging with Deep Learning-Based Reconstruction in Bladder MRI

by Seung Ha Cha, Yeo Eun Han, Na Yeon Han, Min Ju Kim, Beom Jin Park, Ki Choon Sim, Deuk Jae Sung, Seulki Yoo, Patricia Lan and Arnaud Guidon

Diagnostics 2025, 15(5), 595; https://doi.org/10.3390/diagnostics15050595 - 28 Feb 2025

Cited by 2 | Viewed by 1465

Abstract

Background/Objectives: This study compared the image quality of conventional multiplexed sensitivity-encoding diffusion-weighted imaging (MUSE-DWI) and deep learning MUSE-DWI with that of vendor-specific deep learning (DL) reconstruction applied to bladder MRI. Methods: This retrospective study included 57 patients with a visible bladder mass. DWI [...] Read more.

Background/Objectives: This study compared the image quality of conventional multiplexed sensitivity-encoding diffusion-weighted imaging (MUSE-DWI) and deep learning MUSE-DWI with that of vendor-specific deep learning (DL) reconstruction applied to bladder MRI. Methods: This retrospective study included 57 patients with a visible bladder mass. DWI images were reconstructed using a vendor-provided DL algorithm (AIR^TM Recon DL; GE Healthcare)—a CNN-based algorithm that reduces noise and enhances image quality—applied here as a prototype for MUSE-DWI. Two radiologists independently assessed qualitative features using a 4-point scale. For the quantitative analysis, signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), signal intensity ratio (SIR), and apparent diffusion coefficient (ADC) of the bladder lesions were recorded by two radiologists. The weighted kappa test and intraclass correlation were used to evaluate the interobserver agreement in the qualitative and quantitative analyses, respectively. Wilcoxon signed-rank test was used to compare the image quality of the two sequences. Results: DL MUSE-DWI demonstrated significantly improved qualitative image quality, with superior sharpness and lesion conspicuity. There were no significant differences in the distortion or artifacts. The qualitative analysis of the images by the two radiologists was in good to excellent agreement (κ ≥ 0.61). Quantitative analysis revealed higher SNR, CNR, and SIR in DL MUSE-DWI than in MUSE-DWI. The ADC values were significantly higher in DL MUSE-DWI. Interobserver agreement was poor (ICC ≤ 0.32) for SNR and CNR and excellent (ICC ≥ 0.85) for SIR and ADC values in both DL MUSE-DWI and MUSE-DWI. Conclusions: DL MUSE-DWI significantly enhanced the image quality in terms of lesion sharpness, conspicuity, SNR, CNR, and SIR, making it a promising tool for clinical imaging. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

29 pages, 704 KB

Open AccessSystematic Review

Predicting Surgical Difficulty in Rectal Cancer Surgery: A Systematic Review of Artificial Intelligence Models Applied to Pre-Operative MRI

by Conor Hardacre, Thomas Hibbs, Matthew Fok, Rebecca Wiles, Nada Bashar, Shakil Ahmed, Miguel Mascarenhas Saraiva, Yalin Zheng and Muhammad Ahsan Javed

Cancers 2025, 17(5), 812; https://doi.org/10.3390/cancers17050812 - 26 Feb 2025

Cited by 2 | Viewed by 2536

Abstract

Introduction: Following the rapid advances in minimally invasive surgery, there are a multitude of surgical modalities available for resecting rectal cancers. Robotic resections represent the current pinnacle of surgical approaches. Currently, decisions on the surgical modality depend on local resources and the expertise [...] Read more.

Introduction: Following the rapid advances in minimally invasive surgery, there are a multitude of surgical modalities available for resecting rectal cancers. Robotic resections represent the current pinnacle of surgical approaches. Currently, decisions on the surgical modality depend on local resources and the expertise of the surgical team. Given limited access to robotic surgery, developing tools based on pre-operative data that can predict the difficulty of surgery would streamline the efficient utilisation of resources. This systematic review aims to appraise the existing literature on artificial intelligence (AI)-driven preoperative MRI analysis for surgical difficulty prediction to identify knowledge gaps and promising models warranting further clinical evaluation. Methods: A systematic review and narrative synthesis were undertaken in accordance with PRISMA and SWiM guidelines. Systematic searches were performed on Medline, Embase, and the CENTRAL Trials register. Studies published between 2012 and 2024 were included where AI was applied to preoperative MRI imaging of adult rectal cancer patients undergoing surgeries, of any approach, for the purpose of stratifying surgical difficulty. Data were extracted according to a pre-specified protocol to capture study characteristics and AI design; the objectives and performance outcome metrics were summarised. Results: Systematic database searches returned 568 articles, 40 ultimately included in this review. AI to support preoperative difficulty assessments were identified across eight domains (direct surgical difficulty grading, extramural vascular invasion (EMVI), lymph node metastasis (LNM), lymphovascular invasion (LVI), perineural invasion (PNI), T staging, and the requirement for multiple linear stapler firings. For each, at least one model was identified with very good performance (AUC scores of >0.80), with several showing excellent performance considerably above this threshold. Conclusions: AI tools applied to preoperative rectal MRI to support preoperative difficulty assessment for rectal cancer surgeries are emerging, with the progressing development and strong performance of many promising models. These warrant further clinical evaluation, which can aid personalised surgical approaches and ensure the adequate utilisation of limited resources. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

27 pages, 13630 KB

Open AccessArticle

SADASNet: A Selective and Adaptive Deep Architecture Search Network with Hyperparameter Optimization for Robust Skin Cancer Classification

by Günay İlker and İnik Özkan

Diagnostics 2025, 15(5), 541; https://doi.org/10.3390/diagnostics15050541 - 24 Feb 2025

Cited by 4 | Viewed by 1643

Abstract

Background/Objectives: Skin cancer is a major public health concern, where early diagnosis and effective treatment are essential for prevention. To enhance diagnostic accuracy, researchers have increasingly utilized computer vision systems, with deep learning-based approaches becoming the primary focus in recent studies. Nevertheless, there [...] Read more.

Background/Objectives: Skin cancer is a major public health concern, where early diagnosis and effective treatment are essential for prevention. To enhance diagnostic accuracy, researchers have increasingly utilized computer vision systems, with deep learning-based approaches becoming the primary focus in recent studies. Nevertheless, there is a notable research gap in the effective optimization of hyperparameters to design optimal deep learning architectures, given the need for high accuracy and lower computational complexity. Methods: This paper puts forth a robust metaheuristic optimization-based approach to develop novel deep learning architectures for multi-class skin cancer classification. This method, designated as the SADASNet (Selective and Adaptive Deep Architecture Search Network by Hyperparameter Optimization) algorithm, is developed based on the Particle Swarm Optimization (PSO) technique. The SADASNet method is adapted to the HAM10000 dataset. Innovative data augmentation techniques are applied to overcome class imbalance issues and enhance the performance of the model. The SADASNet method has been developed to accommodate a range of image sizes, and six different original deep learning models have been produced as a result. Results: The models achieved the following highest performance metrics: 99.31% accuracy, 97.58% F1 score, 97.57% recall, 97.64% precision, and 99.59% specificity. Compared to the most advanced competitors reported in the literature, the proposed method demonstrates superior performance in terms of accuracy and computational complexity. Furthermore, it maintains a broad solution space during parameter optimization. Conclusions: With these outcomes, this method aims to enhance the classification of skin cancer and contribute to the advancement of deep learning. Full article

(This article belongs to the Topic Machine Learning and Deep Learning in Medical Imaging)

► Show Figures

Figure 1

Topic Menu

Topic Editors

Machine Learning and Deep Learning in Medical Imaging

Topic Information

Keywords

Participating Journals

Published Papers (39 papers)

Further Information

Guidelines

MDPI Initiatives

Follow MDPI

Topic Menu

Topic Editors

Machine Learning and Deep Learning in Medical Imaging

Topic Information

Keywords

Participating Journals

Related Topic

Published Papers (39 papers)

Further Information

Guidelines

MDPI Initiatives

Follow MDPI