Skip to Content
CancersCancers
  • Systematic Review
  • Open Access

20 April 2026

AI-Driven Breast Cancer Diagnosis: A Systematic Review of Imaging Modalities, Deep Learning, and Explainability

,
,
,
,
,
,
and
1
Information Systems Department, Assiut University, Assiut 71515, Egypt
2
Bioengineering Department, J.B. Speed School of Engineering, University of Louisville, Louisville, KY 40292, USA
3
Pathology Department, Faculty of Medicine, Mansoura University, Mansoura 35516, Egypt
4
Department of Pathology and Laboratory Medicine, University of Louisville, Louisville, KY 40292, USA

Simple Summary

Breast cancer remains a leading cause of cancer-related mortality among women worldwide. Early and accurate diagnosis significantly improves patient outcomes. This systematic review examines how artificial intelligence (AI) and deep-learning technologies are transforming breast cancer diagnosis across multiple imaging modalities, including mammography, ultrasound, MRI, molecular breast imaging, PET, and histopathology. We analyzed 65 peer-reviewed studies published between 2018 and 2024, focusing on convolutional neural networks, vision transformers, graph neural networks, and explainable AI methods. Our findings indicate that AI models can achieve diagnostic accuracies exceeding 96% in certain contexts, supporting radiologists in detecting subtle abnormalities and reducing false positives. However, challenges remain regarding dataset standardization, model generalizability, and clinical integration. We emphasize the importance of explainable AI techniques to foster clinician trust and highlight future directions for translating these innovations into routine clinical practice.

Abstract

Background: This article provides a comprehensive overview of recent advancements in artificial intelligence (AI) and deep-learning technologies for breast cancer (BC) diagnosis across various imaging modalities. Methods: A systematic review was conducted in strict adherence to the PRISMA guidelines, incorporating a comparative analysis of 65 peer-reviewed studies published between 2018 and 2024. The evaluation focused on diagnostic performance, architectural developments, and clinical integration strategies. Results: The review synthesizes primary findings on convolutional neural networks (CNNs), emerging architectures including graph neural networks, and hybrid models, with diagnostic accuracy, risk prediction, and personalized screening strategies identified as the leading research domains. Notable achievements include CNNs attaining up to 98.5% accuracy in mammography and Vision Transformers reaching 96% in histopathological analysis. Furthermore, the implementation of explainable AI methodologies, such as SHAP, LIME, and Grad-CAM, is emphasized for maintaining transparency, trust, and accountability in clinical decision-making. Conclusions: AI constitutes a pivotal factor in facilitating early BC diagnosis and optimizing treatment outcomes. Nevertheless, significant challenges persist, including dataset heterogeneity, model generalizability, standardization of imaging protocols, computational resource limitations, and the seamless integration of these technologies into established clinical workflows. Future research must prioritize robust multi-dataset validation and standardized implementation frameworks to overcome existing limitations and advance successful BC diagnostic practices.

1. Introduction

Breast cancer (BC) continues to be one of the most common and lethal cancers among women globally. Lowering mortality and increasing the survival of patients relies greatly on prompt diagnosis and accurate prognosis. In recent years, major developments in imaging techniques and AI have changed the landscape of BC diagnosis such that we can achieve more accurate, rapid, and personalized diagnostic tools [1].
Mammography and ultrasound remain the basis for the screening of BC using conventional imaging methods. Mammography, through the use of low-dose X-rays, is a powerful technique to detect small tumors at an early stage, and it has progressed with contrast-enhanced mammography (CEM), digital breast tomosynthesis (DBT), and digital mammography (DM) to increase its diagnostic precision with a better image quality, as well as 3D views [2]. Ultrasound is particularly useful for dense breast tissue, provides real-time imaging, and advanced techniques such as elastography, which enhances lesion characterization [3].
Magnetic resonance imaging (MRI) is essential for identifying invasive tumors, evaluating the effectiveness of treatment, and performing high-resolution and thorough tissue characterization, combining diffusion-weighted imaging (DWI) and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), enabling personalized medicine by identifying molecular subtypes [4]. Molecular breast imaging (MBI) and positron emission tomography (PET) highlight metabolic activity, allowing early detection, especially in dense tissue [5]. Histopathology, the gold standard for definitive diagnosis, benefits from digital pathology and AI, enhancing precision and reducing variability among pathologists by identifying subtle patterns and biomarkers [6].
Nonetheless, a few technical and medical challenges are still present when talking about standardization and large-scale clinical usage of the method. Differences in imaging protocols, types of equipment, and the requirement for AI to be integrated in the clinical-pathway are some of the issues that have to be resolved by subsequent research and development [7].
The novelty of this survey is multifaceted, extending beyond current limitations to address critical challenges in BC diagnosis. This publication is a review from 2018 to 2024 of all the advances in different imaging modalities, AI technologies, and their impact on the diagnostic accuracy and efficiency of imaging. This review, through the lens of various imaging modalities such as mammography, ultrasound, MRI, MBI, and PET, combined with AI-enabled diagnosis and explainability approaches, highlights how these technologies empower early detection, favorable patient outcomes, and tailored medicine. The article also examines the difficulties in standardizing imaging protocols, the need for specially designed equipment, and the integration of AI into the clinical workflow. Acknowledging these factors, this survey aims to serve as a strategic instrument for shaping research and development directions, such as overcoming the current limitations and being able to improve the existing BC diagnostic protocols.
The rest of this survey is organized as follows: Section 2 reviews various BC diagnosis modalities; Section 3 discusses BC datasets; Section 4 demonstrates the diagnostic and explainability framework; and Section 5 presents the related AI-based systems; Section 6 presents the limitations of current research; and finally Section 7 concludes the manuscript and suggests future directions.

2. Breast Cancer Diagnosing Modalities

Over the past few years, BC diagnosis has notably advanced in terms of imaging modality and AI integration, which have been critical in increasing detection and diagnosing accuracy [8,9]. Conventional imaging modalities, such as mammography and ultrasound, still continue to serve as the cornerstone in detecting BC, but more recent modalities, including digital breast tomosynthesis (DBT), contrast-enhanced mammography (CEM), magnetic resonance imaging (MRI), and molecular breast imaging (MBI), have increased diagnostic potential [5]. These methods provide better resolution and tissue contrast, which is essential for treatment planning and early diagnosis. Figure 1 gives a detailed summary of different BC diagnostic methods/modalities. These modalities are described in the following subsections.
Figure 1. A comprehensive overview of various diagnostic techniques/modalities used for BC detection.

2.1. Mammography

Mammography is a breast imaging modality that generates high-resolution radiographic images using low-dose ionizing radiation [10]. The first and foremost aim of mammography is to help diagnose BC early, sometimes even before signs and symptoms develop, to detect these changes in the tissue of the breast [11]. It requires flattening the breast between two plates in order to distribute the tissue evenly so that clear images can be captured. These images are subsequently evaluated for any evidence of cancer, including lumps, calcifications, or other abnormal changes [12,13].
There is a substantial benefit of mammography in terms of early diagnosis and treatment of BC. It is through mammography that tumors too small to be felt can be identified, allowing an early diagnosis and treatment, thus improving the overall survival rates [14]. It has been found to reduce BC mortality [15]. In addition, technological developments such as digital mammography, tomosynthesis, and AI have only increased both diagnostic accuracy and workflow efficiency. Advanced mammography is widely used for BC diagnosis, including several techniques:
-
Digital Mammography (DM) uses computer-aided detection, replacing conventional film with electronic devices that capture images of the breast, which are stored directly in a computer. This development offers improved aesthetic quality, enhanced image handling, and better storage and sharing capabilities for second opinions and remote consultations. It has been scientifically proven that digital mammography increases diagnostic accuracy, particularly among women with dense breasts [10,16].
-
Digital Breast Tomosynthesis (DBT), also known as 3D mammography, is an emerging imaging add-on technique to take multiple X-ray images of the breast from a set of angles [17]. Subsequently, these images are reconstructed into a 3D volume, providing a more comprehensive visualization of the breast tissue, as opposed to traditional 2D mammography. This approach improves human BC detection and decreases false-positive diagnoses by providing a clearer and more detailed inside visualization of the breast, compared with standard mammography, which enables the radiologists to better recognize the abnormalities [2].
-
Contrast-Enhanced Mammography (CEM) is performed by injecting a contrast in the bloodstream before mammography. The contrast regions with higher blood volume that are often associated with malignant lesions, and thus increase detection rates, especially in women with dense breasts or when standard mammography is inconclusive [18].
Figure 2 provides a graphical visualization of the three techniques in the mammography modality: DM, DBT, and CEM [19,20].

2.2. Ultrasound

Ultrasound is an important diagnostic method in BC diagnosis and therapy, especially to study the lesions identified in mammography or in women with dense breasts. Ultrasounds use high-frequency sound waves instead of X-rays to produce detailed images of breast tissue, as opposed to mammography. This modality is particularly suitable for differentiating solid masses from cysts and is also useful in guiding needle biopsies [3,21]. Recent developments, including high-frequency transducers, microvasculature imaging, elastography, and contrast-enhanced ultrasound, have greatly enhanced the diagnostic performance.
Figure 2. Graphical visualization of the three techniques in the mammography modality: DM, DBT, and CEM [19,20].
There has been a significant increase in the spatial resolution and tissue characterization of ultrasound images with the advancement of technologies. Higher frequency transducers afford better differentiation between low grayscale tones and features of lesions, while microvasculature imaging and elastography allow measurement of blood flow and tissue stiffness, respectively [22]. Contrast-enhanced ultrasound also helps to better visualize tumor vasculature. This is important for accurate diagnosis and to formulate treatment strategies. Ultrasound is non-ionizing radiation, which allows it to be used repeatedly and also to transmit images in a compact, relatively inexpensive package for the benefit of clinical treatment. There are a variety of advanced forms of ultrasound utilized for BC diagnosis, including:
-
High-Frequency Transducers work at frequencies typically greater than 15 MHz, giving images with detailed superficial structures due to higher spatial resolution. The increased resolution results in more visualization of fine anatomical details, improving diagnostic sensitivity for small lesions and structural abnormalities [23,24].
-
Microvasculature Imaging visualizes tiny blood vessels within tissues using advanced Doppler technology to capture slow and small blood flow. It is critical in oncology for identifying tumor angiogenesis and is used in assessing microcirculation in organs and tissues. It provides insights into vascular structures and blood flow patterns, improving the capacity to distinguish between vascularity-based benign and malignant tumors [24].
-
Elastography evaluates tissue stiffness by measuring the response to an external force and has two major methods: strain elastography and shear-wave elastography [25]. It allows non-invasive assessment of tissue elasticity and fibrosis, enhancing the diagnostic confidence with the additional data concerning the composition of tissue [26].
Figure 3 provides a graphical visualization of the three techniques in the ultrasound modality [27,28,29].
Figure 3. Graphical visualization of the three techniques in the ultrasound modality [27,28,29].

2.3. Magnetic Resonance Imaging (MRI)

Magnetic resonance imaging (MRI) excels in providing high-contrast images of soft tissues, making it particularly effective in detecting and characterizing breast lesions that might be missed by other modalities such as mammography and ultrasound. This makes it especially valuable for screening high-risk populations, such as women with a familial predisposition to BC or those presenting with dense breast tissue, where traditional methods might be less effective [24].
Techniques like diffusion-weighted imaging (DWI) and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) enhance diagnostic accuracy by providing detailed insights into the vascular properties and cellular integrity of lesions. In addition, MRI functional imaging is valuable for assessing tumor biology and the prediction of treatment responses, aiding personalized treatment plans [30]. There are several advanced types of MRI used for BC diagnosis, including:
-
Dynamic Contrast-Enhanced MRI (DCE-MRI) involves the use of contrast agents to enhance the visualization of blood flow and vascular properties within breast tissues. This technique helps in distinguishing benign from malignant lesions by analyzing how these tissues uptake and wash out the contrast agent. Studies have shown that DCE-MRI can detect BCs with a sensitivity of up to 87%, significantly higher than mammography. DCE-MRI’s ability to assess tumor angiogenesis and microvascular density makes it a crucial tool in evaluating tumor aggressiveness and planning treatment strategies. Dong et al. [31] demonstrated that DCE-MRI has a pooled diagnostic sensitivity of 0.87 and specificity of 0.74, respectively, with a diagnostic odds ratio (DOR) of 18.83, highlighting its effectiveness in BC detection.
-
Diffusion-Weighted Imaging (DWI) measures the flow of water molecules through tissues and indicates information on tissue cellularity and structural integrity. It is of particular value in malignancy detection because lesions are more likely to have restricted diffusion in the malignancy relative to benign lesions. DWI improves the diagnostic performance of MRI by differentiating high cellular density, which is usually malignant. This method is important for early discovery and evaluation of the effect of treatment, since changes in tissue diffusivity may reflect therapeutic effects. Rodriguez-Soto et al. [32] also reported that advanced DWI techniques, such as restriction spectrum imaging (RSI), improved the tumor conspicuity significantly when compared with conventional methods.
-
Magnetic Resonance Spectroscopy (MRS) examines the chemical content of breast tissue, providing more specific biochemical information, which is a valuable adjunct to conventional MRI images. MRS can detect particular metabolites related to malignant transformation, for example, heightened choline levels being a signal of an increase in cellular proliferation [33]. This subtype improves the specificity in differentiating various breast lesions and also provides additional information suggestive of the diagnosis and management of BC. MRS has been shown to help improve the accuracy of diagnosis and characterization of BC when combined with other MRI techniques [34].
-
Magnetic Resonance Elastography (MRE) is capable of evaluating the mechanical properties of breast tissues, with tissue stiffness playing an important role. The stiffness of malignant tumors is usually higher than that of benign lesions because of cellularity and fibrosis [35]. MRE allows for a non-invasive assessment of these mechanical differences, which can help differentiate benign from malignant tumors and may be used to enhance BC identification and characterization in challenging clinical situations [36].
Figure 4 provides a graphical visualization of the four techniques in the MRI modality [37,38,39,40].
Figure 4. Graphical visualization of the four techniques in the MRI modality: DCE-MRI, DWI, MRS, and MRE.

2.4. Molecular Breast Imaging (MBI)

MBI is an advanced nuclear medicine method for detecting BC. It involves the use of a radiotracer, typically Technetium-99m sestamibi, which is injected into the bloodstream and absorbed by cancer cells as a result of their increased metabolic activity. The radiotracer emits gamma rays, which are captured by a specialized gamma camera to produce in-depth images of the breast tissue. MBI is particularly beneficial for women with dense breast tissue, where traditional mammography may have reduced sensitivity [41].
One of the primary strengths of MBI is its ability to detect cancers that may be missed by mammography, especially in dense breast tissue. MBI has demonstrated a higher sensitivity for detecting small and otherwise occult BCs, with studies showing incremental cancer detection rates of up to 9.3 per 1000 exams when used alongside mammography. Additionally, MBI offers a high negative predictive value and is less affected by breast density, making it a reliable supplementary screening tool. It also provides functional imaging, which can help in assessing the response to neoadjuvant therapy and evaluating the extent of disease [42].
MBI encompasses several advanced imaging techniques, primarily including breast-specific gamma imaging (BSGI) and positron emission mammography (PEM) [43]:
-
Breast-specific gamma imaging (BSGI) utilizes a gamma camera to detect gamma rays emitted by a radiotracer, commonly Technetium-99m sestamibi, injected into the patient. Cancer cells, due to their higher metabolic activity, absorb more of the radiotracer, allowing for the creation of high-resolution images that highlight areas of concern. BSGI is particularly effective in dense breast tissue where mammography might be less sensitive [44].
-
Positron emission mammography (PEM) involves the use of a PET scanner to detect positrons emitted by a radiotracer such as Fluorodeoxyglucose (FDG). This method highlights areas with elevated glucose metabolism, a characteristic feature of cancer cells. PEM provides high-resolution images and is useful for identifying and characterizing breast lesions, especially in dense breast tissue [45].
MBI has some notable limitations despite its benefits. A major concern is the radiation exposure required, although recent advancements have reduced the administered dose. In addition, MBI is not available to many recipients because it requires special equipment and nuclear medicine facilities, which are not as easily available in some regions [46]. The procedure also involves longer imaging times compared to traditional mammography, posing logistical challenges for both patients and healthcare providers. Additionally, MBI may have trouble clearly displaying lesions close to the axillary lymph nodes and chest wall due to limitations in the detector’s field of view [47].

2.5. Positron Emission Tomography (PET)

PET allows clinicians to visualize and measure metabolic processes within the body using radiotracers such as FDG. The radiotracer builds up in tissues with high metabolic activity, including cancer cells, after being injected into the patient’s circulation. The PET scanner detects the gamma rays emitted by the radiotracer, creating detailed images that reflect the functional activity of organs and tissues [48].
One of the primary strengths of PET is its ability to detect abnormalities at the molecular level, often before structural changes become apparent on other imaging modalities like CT or MRI. This high sensitivity makes PET particularly valuable in oncology for early cancer detection, staging, and monitoring treatment response. In addition, in a single session, PET can be used in conjunction with CT or magnetic resonance imaging to offer anatomical and metabolic information, improving diagnostic accuracy. PET/CT, specifically using 18F-FDG, is instrumental in identifying malignant lesions, assessing the extent of disease spread, and evaluating the effectiveness of treatment [49]. PET has seen several recent technological advancements:
-
Standard PET is the most commonly used imaging tool for assessing the body’s metabolic activity in a clinical setting [50].
-
PET/CT combines the functional imaging power of PET with the anatomical detail of CT. This imaging modality, as a hybrid system, allows the precise location of metabolism-related anomalies within the body, improving diagnostic accuracy. In oncology, PET/CT is especially useful for cancer staging, monitoring a patient’s response to therapy, and identifying recurrent disease [51].
-
PET/MRI overlays high spatial resolution anatomic details imparted by MRI on the metabolic information available from PET, thus it can potentially be considered to be an ideal imaging modality for soft tissues. This combination is particularly advantageous for neurology and oncology imaging, where high-quality soft tissue contrast is paramount [50].
-
Total-body PET advanced imaging modality enables imaging the whole body simultaneously. This may be of great utility in the study of systemic diseases, as well as for dynamic imaging of physiological processes. Total-body PET can be used to provide a complete picture of disease spread, to assess response to therapy, and to monitor the response to therapies on multiple organs [52].
Nevertheless, PET also comes with several disadvantages. PET scanners and radiotracers may not be widely available, especially in small or remote healthcare centers. The method is rather costly, since production and manipulation of radioactive tracers are very expensive [50]. In addition, the use of radioactive agents, even when used properly, carries a risk of radiation exposure to patients and medical workers. In addition, the spatial resolution of PET scans is relatively inferior to that of both CT and MRI, making it difficult to precisely localize abnormalities. Another limitation is the possibility of false positives and false negatives; therefore, PET images have to be carefully interpreted in the context of other diagnostic tests and clinical information.

2.6. Histopathology

Histopathology refers to the microscopic examination of breast tissue to look for cancer cells. Usually, samples are taken via a biopsy, in which a piece of tissue is removed and examined by pathologists. Regarding the characteristics of the tumor, the histopathological study is essential since it evaluates the type of cell, degrees, and the presence of various indicators, such as HER2, progesterone receptor (PR), and estrogen receptor (ER). Such detailed knowledge is essential to personalize treatment approaches and predict outcomes. This detailed knowledge is crucial for personalized treatment strategies and predicting response [53,54,55].
Rapid progress of technology and growing developments of histopathology for BC diagnosis have been observed in recent years, where modern techniques have been implemented for diagnostic accuracy, efficiency, and individualization of the treatment. Some of the popular advanced techniques include:
-
Digital Pathology and Whole Slide Imaging (WSI), which is the use of digital slide scanners to convert entire glass slides into high-resolution digital images. This system also enables remote consultation and telepathology, and expert opinions are available irrespective of geographic restrictions. The validation of WSI for diagnostic purposes has been stressed to ensure clinical applicability and reliability [56].
-
Multiplex Immunohistochemistry (mIHC) and Immunofluorescence (mIF), where it can be used for detecting multiple markers simultaneously on one section of tissue. This multiplexing functionality offers the remarkable opportunity to analyze the tumor microenvironment both in its spatial comprehensiveness and for cell-cell relationships and protein expression. These methods are also useful for dissecting tumor heterogeneity, and the interplay between cancer cells and the immune system [57].
-
Next-Generation Sequencing (NGS) and Molecular Profiling permits the identification and analysis of genetic mutations and alterations in BC. Such technology identifies actionable mutations and supports targeted therapy, and may be beneficial in personalizing treatment for BC. Molecular profiling using NGS has also been more routinely incorporated into pathology practice, providing a richer landscape of the molecular basis of BC [58].
-
3D Histopathology and Optical Coherence Tomography (OCT) with the potential to provide a 3D reconstruction of the tissue samples, can offer a complete image of tumoral architecture, and its relationship with neighboring tissues. OCT can be used to achieve high-resolution images of tissue microarchitecture and can assist in the detailed assessment of tumor margins and invasive characteristics [59].
-
Automated Tissue Microarrays (TMA) enables the screening of several tissues in a single slide. Such a high-throughput approach also serves for biomarker validation and studies on a large scale. Automated TMA systems improve the productivity of histopathological studies by facilitating the simultaneous analysis of large quantities of samples in a highly standardized manner [60].
Figure 5 provides a graphical visualization of the techniques in the MBI and Histopathology modalities/approaches [61,62,63].
Figure 5. Graphical visualization of the techniques in the MBI (Left) and Histopathology (Right) modalities/approaches [61,62,63].
Table 1 compares between the different modalities regarding (1) Compatibility with dense breast tissue, (2) Cost, (3) Scan time, (4) Invasive or Non-invasive, (5) Safe procedure, (6) Radiation exposure, (7) When to use, (8) Compatibility with cancer, (9) False Positives, and (10) False Negatives.

2.7. Clinical and Technical Limitations

Despite modality-specific advancements, AI-driven BC diagnostics face cross-cutting limitations that impede clinical translation. First, data heterogeneity and domain shift remain pervasive; models trained on curated public datasets (e.g., DDSM, BreakHis) frequently degrade by 8.2–14.6% when deployed on institution-specific data due to variations in acquisition protocols, scanner vendors, and staining techniques [64]. Second, computational and infrastructural barriers restrict deployment, particularly for ViT and hybrid architectures that demand GPU-accelerated environments incompatible with legacy PACS systems. Third, clinical validation gaps persist, with only 18.5% of reviewed studies employing external, multi-institutional validation, limiting regulatory approval and real-world generalizability. Finally, modality-specific constraints compound these issues: mammography suffers from dense breast masking and radiation exposure; ultrasound remains highly operator-dependent with limited microcalcification visibility; MRI faces prohibitive costs and contrast contraindications; MBI/PET involves radiotracer exposure and restricted spatial resolution; and histopathology is constrained by biopsy invasiveness, sampling errors, and inter-observer variability. Addressing these barriers requires federated learning frameworks, lightweight model distillation, standardized imaging protocols, and prospective multicenter trials aligned with FDA SaMD and EU MDR guidelines.
Table 1. Comparative analysis of breast imaging modalities across clinical and operational criteria.

3. Breast Cancer Datasets

Breast cancer research benefits greatly from the availability of diverse and comprehensive datasets (see Table 2), which are essential for developing and validating diagnostic models. The Digital Database for Screening Mammography (DDSM) and its curated version, CBIS-DDSM, are important databases that provide extensive mammographic images used for detecting abnormalities [65]. The INbreast dataset offers full-field digital mammograms crucial for developing and testing computer-aided detection (CAD) systems [66]. Histopathological images, such as those in the BreakHis dataset, include high-resolution microscopic images that aid in classifying BC at the cellular level. Additionally, MRI-based datasets used in studies involving dynamic contrast-enhanced MRI (DCE-MRI) provide volumetric data that enhance the detection and characterization of breast tumors [64].
Table 2. Summary of public and private datasets utilized in breast cancer research.

4. Breast Cancer Diagnosing and Explainability Framework

The BC diagnosis and explainability framework (refer to Figure 6) is an intricate chain of steps aimed at not only increasing the diagnostic accuracy but also making the process transparent. Initially, it includes preprocessing techniques like image normalization, noise reduction, and contrast enhancement that basically make raw data ready for analysis [67]. After this, deep and machine learning techniques for BC segmentation are used; by employing methods such as convolutional neural networks (CNNs) and U-Net to segment BC regions in the images [68]. Subsequently, ensemble methods and models like support vector machines (SVMs) are implemented to detect BC from the segmented data [69]. Model performance evaluation metrics are then used by the framework to measure the effectiveness of diagnostic models in terms of accuracy, precision, recall, and F1, and ROC-AUC [64]. Finally, AI Models’ Explainability techniques like SHAP, LIME, and Grad-CAM can serve as tools to understand AI models’ choices, thereby proving that they are explainable and reliable. Thus, trust can be enhanced, and clinical practitioners can be facilitated in making better decisions [69].
Figure 6. Graphical visualization of the standard CAD system for BC diagnosis.

4.1. Preprocessing Techniques

Preprocessing methods are fundamental in the entire workflow for the BC diagnosis, so that the data input to the ML model is tidy, coherent, and ready for analysis. For general preprocessing techniques, data normalization and standardization (this rescales pixel values into a common range (e.g., 0 to 1) to stabilize and improve model behavior). Noise reduction is another important process, and can be performed using filters, such as the median filter, which smooths images while maintaining significant features [70,71]. Second, some conventional image enhancement methods, such as contrast improvement, are also adopted to enhance the tumor visibility, which is convenient for computer algorithms to detect the abnormal regions. Moreover, the preparation of the segmentation includes the use of the regions of interest (ROIs) to concentrate the analysis on the relevant regions, which achieves a decrease in computational cost while increasing accuracy [72].

4.2. Deep and Machine Learning Techniques for BC Segmentation

Deep learning (DL) and machine learning (ML) methods have brought a new revolution to BC segmentation by considerably improving the accuracy and efficiency of the tumor detection process on medical images.
CNNs have revolutionized BC segmentation by learning abstract features automatically, and they have enhanced the tumor localization accuracy in medical images. CNNs work well in dealing with the complexity and diversity of BC imaging data. Studies from 2020 to 2023 show the effectiveness of CNN-based models for segmenting tumors in mammograms, MRI, and ultrasound images. These models, and deeper architectures like ResNet and Inception, achieved high accuracy and detailed tumor boundaries [71].
U-Net Architecture was originally proposed for biomedical image segmentation and has been very popular as a key step in most of the BC segmentation tasks due to its capability to capture fine details as well as context information. U-Net consists of an expanding path for localization and a symmetric contracting path for context. U-Net and its variants, such as U-Net++, ResUNet, and Attention U-Net, were proven to be more accurate and reliable than the classical methods in terms of the results of tumor segmentation. A study by Alam et al. demonstrated the performance of U-Net3+ in segmenting the breast ultrasound images using an average accuracy of 82.53% and global accuracy of 90.99% [72]). Furthermore, a study presented R2U-Net that uses recurrent and residual connections, and achieved 95.6% accuracy [73]. In other studies, attention mechanisms and federated learning were utilized to supplement U-Net models and increase segmentation accuracy and generality for diverse BC imaging modalities [74,75].
Transfer Learning is an effective approach for BC segmentation, especially for the case of having a small number of labeled data. This practice consists of the employment of pre-trained models on generic data and then fine-tuning them on particular BC segmentation tasks. The benefit of using pre-trained features for a segmentation model has been well established in studies as transfer learning.
Multiscale and Multilevel Feature Extraction has been included in BC segmentation models to make them robust and accurate. These methods allow for the models to automatically perceive tumors at different scales and resolutions, which is crucial for accurately capturing the boundaries of even small and subtle tumor regions. The efficacy of these approaches has been supported by research. For instance, Swin-Net employs transformer and CNN models to enhance the segmentation accuracy by utilizing a hierarchical multiscale feature fusion module, and reports an absolute gain of 1.4–1.8% in Dice scores, respectively [76]. Another approach proposed is called the Multiscale Parallel Convolution Structure (MSPCF) that can extract feature information in multiple scales for capturing fine details and edges more effectively in the CNN [77]. In addition, multiscale and dual-adaptive attention mechanisms are used in the MDAA network, which effectively cope with the variation of tumor size and appearance and show better performance on the BC segmentation task [78].
Advanced Segmentation Models such as encoder-decoder networks, attention mechanisms, and generative models have also extended the frontier of BC segmentation. As an extension, encoder-decoder models, including those based on U-Net architecture, were empowered with attention mechanisms to highlight regions of the image that contribute most to improving segmentation. Generative models such as GANs have been employed for synthetic training data generation to overcome the problem of insufficient labeled data. Studies reported that such advanced models are highly accurate and efficient in segmenting BC images, which will further help accurate and efficient diagnoses [79,80].

4.3. Architectural Paradigms: Comparative Analysis of DL Models

While numerous architectures have been applied to BC diagnosis, their underlying inductive biases dictate distinct strengths and limitations. Convolutional Neural Networks (CNNs) remain the most widely adopted due to their translation equivariance, parameter efficiency, and proven efficacy in capturing local morphological features (e.g., microcalcifications, lesion boundaries). However, CNNs struggle with long-range contextual dependencies and global tissue architecture, which are critical in histopathology and dense breast screening. Vision Transformers (ViTs) address this limitation through self-attention mechanisms, enabling global receptive fields and superior modeling of complex spatial relationships. ViTs consistently outperform CNNs on high-resolution WSIs and multimodal fusion tasks, but require substantially larger datasets and computational resources, making them less feasible in low-resource clinical environments.
Graph Neural Networks (GNNs) excel in modeling non-Euclidean data structures, such as spatial relationships between cell nuclei, vascular networks, or radiomic feature graphs. GNNs have demonstrated exceptional performance in ultrasound and multiparametric MRI analysis, where topological feature integration is paramount. Nevertheless, GNNs require explicit graph construction, which introduces preprocessing complexity and limits end-to-end optimization. Hybrid architectures (e.g., CNN-Transformer or CNN-GNN fusion) attempt to balance local feature extraction with global context modeling, achieving state-of-the-art accuracy in several reviewed studies. However, their increased architectural depth and hyperparameter sensitivity complicate clinical deployment and explainability. Ultimately, the choice of architecture must align with the imaging modality, dataset scale, computational constraints, and the specific clinical task (e.g., screening vs. prognostic stratification).

4.4. Deep and Machine Learning Techniques for BC Diagnoses

CNNs have greatly promoted the diagnosis of BC by automatic feature extraction and enhancing the accuracy of tumor detection and classification in medical images. A study combined CNNs with long-short-term memory (LSTM) networks, achieving a 91% accuracy rate in classifying BC histopathological images, demonstrating high diagnostic performance [71]. Furthermore, CNN-based BC detection algorithms perform better when multimodal data, like imaging and clinical data, are integrated, according to a 2024 study, achieving diagnostic accuracies ranging from 74% to 98.02% depending on the dataset and diagnostic context [81].
Transfer Learning has demonstrated significant potential in the diagnosis of BC. For instance, a study by Arooj et al. reported the utilization of a customized AlexNet model, achieving 99.35% accuracy for benign and normal classes, and 100% for malignant in dataset A, 96.66% in dataset B, 99.11% in dataset C, and 100% in dataset A2 [82]. Another study by Azevedo et al. on hybrid classical-quantum models reported that the classical ResNet model with transfer learning achieved 84% accuracy, 95% precision, and 73% recall. The hybrid classical-quantum model reached similar accuracy at 84%, with 100% precision and 69% recall, demonstrating the potential for efficient and precise BC diagnosis [83].
Vision Transformers (ViT) have shown great promise in BC diagnosis due to their ability to capture global context in images, surpassing traditional CNNs. For example, BUViTNet, a ViT-based model, achieved superior performance in breast ultrasound detection with an AUC of 0.968, outperforming both CNNs and other ViT models [84]. Another study demonstrated the effectiveness of ViTs in classifying mammograms across multiple diagnostic modalities, enhancing diagnostic accuracy and capturing intricate image details [85].
Multiscale and Multilevel Feature Extraction was utilized to improve the accuracy and efficiency of medical image analysis significantly. For instance, a study by Zhang et al. utilized the MS-GWNN model that demonstrated an accuracy of 93.75% on the BACH dataset and 99.67% on the BreakHis dataset [86].

4.5. Model Performance Evaluation Metrics

Evaluating the performance of models in BC diagnosis involves a variety of metrics (see Table 3) to ensure a comprehensive assessment. For instance, Accuracy quantifies the percentage of true findings (both true positives and true negatives) among all instances studied. High accuracy indicates that the model is generally correct. Precision (Positive Predictive Value) indicates the proportion of true positive results in all positive results predicted by the model. It highlights the model’s ability to avoid false positives.
Recall (Sensitivity or True Positive Rate) quantifies the percentage of real positives that the model accurately detected. It emphasizes the model’s ability to find all relevant cases. F1 Score is the harmonic mean of precision and recall, and it has a balance between these two. It comes very handy in the case of an imbalanced dataset.
Specificity (True Negative Rate) measures the percentage of true negatives that the model correctly detects. It is important to understand the model’s ability against false-negative results. Area Under the Receiver Operating Characteristic Curve (AUC-ROC) provides an aggregated measure across all classification thresholds. A larger AUC means a better overall performance.
Collectively, these metrics are important for understanding a model’s various performance characteristics, making sure a model not only accurately detects true positives but also minimizes the number of false positives and false negatives, which will make a diagnostic test more dependable. Table 3 lists the performance metrics that can be used to evaluate an AI-based model.
Table 3. Performance metrics for classification model assessment.

4.6. AI Models Explainability

Explainability of AI models, particularly in sensitive domains such as BC diagnosis, is important for trust building, clinical decision support, and accountability. For example, Local Interpretable Model-agnostic Explanations (LIME) produces at the local explanation level surrogate models that mimic predictions from black-box AI models. LIME facilitates an understanding of the decision-making processes of complex models for individual instances by perturbing the input data and learning a simple model that is interpretable. This approach is especially valuable in describing tissue density and concerning lesions, helping to build trust around AI systems [87].
A study by Prananda et al. [88] showed that the LIME performed very well in explaining important features associated with BC severity, consistent with doctors’ visual decisions, thereby improving model interpretability and clinician trust. Similarly, Ahmed et al. [89] applied LIME to CNN predictions to highlight important image regions that affected the model’s predictions, which increased the reliability of AI-based diagnostics. Additionally, Hakkoum et al. [90] used LIME on a multilayer perceptron; LIME was demonstrated to be capable of giving real-time interpretability of predictions and guidance that could assist doctors in differentiating between benign and malignant cases. Cumulatively, these studies highlight that LIME helps to make the AI model more transparent and more reliable in clinical applications.
SHapley Additive exPlanations (SHAP) uses values that can be used to explain the contribution of each feature to the model prediction, which can be used for local and global explanability. This method has been extensively used to interpret intricate models, e.g., deep neural networks, revealing tumor size, shape, and density as the factors affecting diagnosis results. SHAP values offer explanations by allocating the prediction to the input features; thus, the impact of each feature is explained, and its quantification is made with respect to the entire model decision [91].
One study by Jansen et al. [92] gives global insights through quantifying the importance of each of the features in making predictions by the model. SHAP could successfully capture important factors such as age and tumor size. Another study by Zhang et al. [93] applied SHAP to explain the predictions of the XGBoost model. SHAP effectively visualized the impact of ultrasound features like suspicious lymph nodes and microcalcifications.
Gradient-weighted Class Activation Mapping (Grad-CAM) is a visual explanation technique that highlights important regions in an image that are crucial for the model’s prediction. This method has been particularly useful for interpreting image-based models in BC diagnosis. Grad-CAM works by using the gradients of the target class flowing into the final convolutional layer to produce a coarse localization map, which helps in understanding where the model is focusing [94]. Grad-CAM was utilized in different studies to highlight important regions in mammographic images crucial for the model’s prediction [95].
Despite their widespread adoption, post-hoc explainability techniques exhibit documented limitations in clinical contexts. Grad-CAM heatmaps often suffer from gradient saturation and coarse localization, while LIME and SHAP attributions can be sensitive to input perturbations and background correlations, potentially generating misleading clinical rationales. Emerging literature emphasizes the need for clinically grounded XAI, wherein explanations align with established radiological or pathological criteria (e.g., BI-RADS descriptors, nuclear atypia grading). Inherently interpretable architectures, such as prototype-based networks and attention-regularized models, offer promising alternatives by constraining latent representations to anatomically plausible features, thereby improving clinician trust and regulatory acceptance. Table 4 compares these techniques, detailing their working mechanisms and clinical applications.
Table 4. Comparative analysis of AI explainability techniques utilized in breast cancer diagnosis.

6. Limitations

Despite the great strides that have been made in AI-powered BC disease diagnostics, the limitations of those technologies prevent them from being widely used in clinical settings. Variations in imaging methods from one institution to another lead to differences in diagnostic accuracy; the need for standardization is evident. In addition to that, due to the high price of advanced imaging technologies (MRI, MBI, PET), access to them is limited (particularly in areas with low resources), where diagnostic capacity may also be low, thus causing disparities in diagnostic capacity.
The integration of AI into clinical workflows is hindered by the requirement for large, diverse, and annotated datasets. Besides that, there are also privacy and security concerns that make data sharing more complicated. The fusion “black box” of AI models causes less clinician trust, which indicates the necessity for explainability methods such as SHAP, LIME, and Grad-CAM [156]. Besides that, a large number of AI models are not validated prospectively in real-world clinical settings, which limits their trustworthiness in clinical practice [157].
Furthermore, most reviewed models suffer from domain shift, where performance degrades significantly when applied to data from different institutions or scanner vendors. Future work must prioritize federated learning frameworks to train robust models across distributed datasets without compromising patient privacy. Additionally, the computational cost of Transformer-based models (ViT) remains a barrier for real-time clinical deployment on edge devices.
Moreover, the regulatory pathway for AI-based diagnostic tools remains fragmented across jurisdictions (e.g., FDA 510(k) vs. EU MDR), creating uncertainty for clinical adoption. Ethical considerations (including algorithmic bias against underrepresented populations, data sovereignty in federated learning, and liability for AI-assisted misdiagnosis) require multidisciplinary frameworks before widespread deployment.

7. Conclusions and Future Research Directions

Artificial intelligence-based BC diagnostics have evolved dramatically, now featuring a fusion of different imaging modalities and DL methods for improved early detection and classification. The integration of mammography, ultrasound, MRI, MBI, PET, and histopathology with AI models has been very accurate diagnostically in a large number of studies. CNNs have been able to identify cancer correctly in mammography images with a rate of up to 98.5%, while ViTs have registered 96% accuracy in histopathological classification. The use of explainability techniques, such as SHAP, LIME, and Grad-CAM, has helped in revealing the black-box nature of AI, thus enabling clinicians to obtain a better understanding of the AI decisions. These breakthroughs signal the expanded use of AI in BC screening, diagnosis, and therapy selection. Future research must prioritize enhancing AI interpretability, standardizing imaging protocols, and increasing data diversity to enable robust model training. Furthermore, reducing computational complexity and making AI-driven diagnostics affordable will, in fact, be very important for a large-scale acceptance of the practice. To translate these advancements into clinical practice, we advocate for: (1) prospective, multicenter trials validating AI systems across diverse healthcare settings; (2) development of standardized reporting guidelines (e.g., CONSORT-AI extension) for AI diagnostic studies; (3) establishment of regulatory sandboxes for iterative AI model approval; and (4) creation of equitable data-sharing frameworks that protect patient privacy while enabling model generalizability. Ultimately, this technology can radically change the way BC diagnostics are done, making it possible to detect the disease at earlier stages, devise treatment plans specific to individual patients, and thus, increase the therapeutic success rate.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/cancers18081305/s1. Reference [158] is cited in supplementary material.

Author Contributions

Conceptualization, M.S., H.M.B. and A.E.-B.; methodology, M.S., H.M.B. and A.E.-B.; validation, K.M.A., A.M., D.G., M.G., T.H.A.S. and A.E.-B.; formal analysis, M.S., H.M.B., A.M. and A.E.-B.; investigation, M.S., H.M.B., D.G. and T.H.A.S.; writing—original draft preparation, M.S., H.M.B., K.M.A. and A.M.; writing—review and editing, M.S., H.M.B., K.M.A., A.M., D.G., M.G., T.H.A.S. and A.E.-B.; visualization, M.S., H.M.B., K.M.A. and A.M.; supervision, D.G., M.G., T.H.A.S. and A.E.-B.; project administration, D.G., T.H.A.S. and A.E.-B. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

Not applicable.

Data Availability Statement

No new data were created or analyzed in this study. Data sharing is not applicable to this article as it is a systematic review of previously published studies. All datasets referenced in this review are publicly available or accessible through institutional agreements as cited in Table 2.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Jones, M.A.; Islam, W.; Faiz, R.; Chen, X.; Zheng, B. Applying artificial intelligence technology to assist with breast cancer diagnosis and prognosis prediction. Front. Oncol. 2022, 12, 980793. [Google Scholar] [CrossRef] [Scilit]
  2. Wang, L. Mammography with deep learning for breast cancer detection. Front. Oncol. 2024, 14, 1281922. [Google Scholar] [CrossRef] [Scilit]
  3. Catalano, O.; Fusco, R.; De Muzio, F.; Simonetti, I.; Palumbo, P.; Bruno, F.; Borgheresi, A.; Agostini, A.; Gabelloni, M.; Varelli, C.; et al. Recent advances in ultrasound breast imaging: From industry to clinical practice. Diagnostics 2023, 13, 980. [Google Scholar] [CrossRef] [Scilit]
  4. Galati, F.; Rizzo, V.; Trimboli, R.M.; Kripa, E.; Maroncelli, R.; Pediconi, F. MRI as a biomarker for breast cancer diagnosis and prognosis. BJR|Open 2022, 4, 20220002. [Google Scholar] [CrossRef] [Scilit]
  5. Elhatw, A.; Chung, H.L.; Kamal, R.M.; De Jesus, C.; Jean, S.; Vishwanath, V.; Ferreira Dalla Pria, H.R.; Patel, M.M.; Guirguis, M.S.; Moseley, T.W. Advanced Breast Imaging Modalities—DBT, CEM, MBI, PEM, MRI, AI. Curr. Breast Cancer Rep. 2023, 15, 108–113. [Google Scholar] [CrossRef] [Scilit]
  6. Dexter, A.; Tsikritsis, D.; Belsey, N.A.; Thomas, S.A.; Venton, J.; Bunch, J.; Romanchikova, M. Next generation digital pathology: Emerging trends and measurement challenges for molecular pathology. J. Mol. Pathol. 2022, 3, 168–181. [Google Scholar] [CrossRef] [Scilit]
  7. Najjar, R. Redefining radiology: A review of artificial intelligence integration in medical imaging. Diagnostics 2023, 13, 2760. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  8. Zhang, J.; Wu, J.; Zhou, X.S.; Shi, F.; Shen, D. Recent advancements in artificial intelligence for breast cancer: Image augmentation, segmentation, diagnosis, and prognosis approaches. In Proceedings of the Seminars in Cancer Biology; Elsevier: Amsterdam, The Netherlands, 2023. [Google Scholar]
  9. Sabry, M.; Balaha, H.M.; Ali, K.M.; Soliman, T.H.A.; Gondim, D.; Ghazal, M.; Tahtouh, T.; El-Baz, A. A vision transformer approach for breast cancer classification in histopathology. In Proceedings of the 2024 IEEE International Symposium on Biomedical Imaging (ISBI); IEEE: New York, NY, USA, 2024; pp. 1–4. [Google Scholar]
  10. Nicosia, L.; Gnocchi, G.; Gorini, I.; Venturini, M.; Fontana, F.; Pesapane, F.; Abiuso, I.; Bozzini, A.C.; Pizzamiglio, M.; Latronico, A.; et al. History of mammography: Analysis of breast imaging diagnostic achievements over the last century. Healthcare 2023, 11, 1596. [Google Scholar] [CrossRef] [Scilit]
  11. Li, H.; Zhang, S.; Wang, Q.; Zhu, R. Clinical value of mammography in diagnosis and identification of breast mass. Pak. J. Med. Sci. 2016, 32, 1020. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  12. Bhushan, A.; Gonsalves, A.; Menon, J.U. Current state of breast cancer diagnosis, treatment, and theranostics. Pharmaceutics 2021, 13, 723. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  13. Lång, K.; Josefsson, V.; Larsson, A.M.; Larsson, S.; Högberg, C.; Sartor, H.; Hofvind, S.; Andersson, I.; Rosso, A. Artificial intelligence-supported screen reading versus standard double reading in the Mammography Screening with Artificial Intelligence trial (MASAI): A clinical safety analysis of a randomised, controlled, non-inferiority, single-blinded, screening accuracy study. Lancet Oncol. 2023, 24, 936–944. [Google Scholar]
  14. Jafari, Z.; Karami, E. Breast cancer detection in mammography images: A CNN-based approach with feature selection. Information 2023, 14, 410. [Google Scholar] [CrossRef] [Scilit]
  15. Hussain, S.; Ali, M.; Naseem, U.; Nezhadmoghadam, F.; Jatoi, M.A.; Gulliver, T.A.; Tamez-Peña, J.G. Breast cancer risk prediction using machine learning: A systematic review. Front. Oncol. 2024, 14, 1343627. [Google Scholar] [CrossRef] [Scilit]
  16. Rabie Ewis El-Sayed, A.; Ali Motawe, M.; Mohammed Faheem, Y.; Abo El-Naga Mohammed, M. ROLE OF 3D DIGITAL MAMMOGRAPHY IN STUDYING BREAST LESIONS IN ADULT FEMALES OF DENSE BREAST. Al-Azhar Med. J. 2021, 50, 3097–3112. [Google Scholar]
  17. Helvie, M.A. Digital mammography imaging: Breast tomosynthesis and advanced applications. Radiol. Clin. 2010, 48, 917–929. [Google Scholar] [CrossRef] [Scilit]
  18. Coffey, K.; Jochelson, M.S. Contrast-enhanced mammography in breast cancer screening. Eur. J. Radiol. 2022, 156, 110513. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  19. Hannsun, G.; Saponaro, S.; Sylvan, P.; Elmi, A. Contrast-enhanced mammography: Technique, indications, and review of current literature. Curr. Radiol. Rep. 2021, 9, 12. [Google Scholar] [CrossRef] [Scilit]
  20. International, M.I. Synthetic DBT Images Could Replace Digital Mammograms. 2020. Available online: https://www.medimaging.net/general-imaging/articles/294784957/synthetic-dbt-images-could-replace-digital-mammograms.html (accessed on 12 March 2026).
  21. Afrin, H.; Larson, N.B.; Fatemi, M.; Alizad, A. Deep learning in different ultrasound methods for breast cancer, from diagnosis to prognosis: Current trends, challenges, and an analysis. Cancers 2023, 15, 3139. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  22. Zhu, J.Y.; He, H.L.; Jiang, X.C.; Bao, H.W.; Chen, F. Multimodal ultrasound features of breast cancers: Correlation with molecular subtypes. BMC Med. Imaging 2023, 23, 57. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  23. Ricci, V.; Cocco, G.; Donati, D.; Farì, G.; Chang, K.V.; Özçakar, L. From Histopathology to High-Resolution Ultrasound Imaging of Skin Scars. Diagnostics 2023, 13, 3629. [Google Scholar] [CrossRef] [Scilit]
  24. Chen, J.; Fei, C.; Lin, D.; Gao, P.; Zhang, J.; Quan, Y.; Chen, D.; Li, D.; Yang, Y. A review of ultrahigh frequency ultrasonic transducers. Front. Mater. 2022, 8, 733358. [Google Scholar] [CrossRef] [Scilit]
  25. Ozturk, A.; Grajo, J.R.; Dhyani, M.; Anthony, B.W.; Samir, A.E. Principles of ultrasound elastography. Abdom. Radiol. 2018, 43, 773–785. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  26. Yamashita, Y.; Kitano, M. Role of contrast-enhanced harmonic endoscopic ultrasonography (EUS) and EUS elastography in pancreatic lesions. Clin. Endosc. 2024, 57, 164–174. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  27. Hansen, K.L.; Nielsen, M.B.; Ewertsen, C. Ultrasonography of the kidney: A pictorial review. Diagnostics 2015, 6, 2. [Google Scholar] [CrossRef] [Scilit]
  28. Corvino, A.; Varelli, C.; Catalano, F.; Cocco, G.; Delli Pizzi, A.; Boccatonda, A.; Corvino, F.; Basile, L.; Catalano, O. Use of high-frequency transducers in breast sonography. J. Pers. Med. 2022, 12, 1960. [Google Scholar] [CrossRef] [Scilit]
  29. Cleveland Clinic. Breast Ultrasound. 2024. Available online: https://my.clevelandclinic.org/health/diagnostics/21496-breast-ultrasound (accessed on 12 March 2026).
  30. Ming, W.; Zhu, Y.; Bai, Y.; Gu, W.; Li, F.; Hu, Z.; Xia, T.; Dai, Z.; Yu, X.; Li, H.; et al. Radiogenomics analysis reveals the associations of dynamic contrast-enhanced–MRI features with gene expression characteristics, PAM50 subtypes, and prognosis of breast cancer. Front. Oncol. 2022, 12, 943326. [Google Scholar] [CrossRef] [Scilit]
  31. Dong, H.; Kang, L.; Cheng, S.; Zhang, R. Diagnostic performance of dynamic contrast-enhanced magnetic resonance imaging for breast cancer detection: An update meta-analysis. Thorac. Cancer 2021, 12, 3201–3207. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  32. Rodríguez-Soto, A.E.; Meriwether, C.H.; Park, A.J.; Adams, D.W., Jr.; Wallace, A.; Ojeda-Fournier, H.; Dale, A.M.; Rakow-Penner, R. Noncontrast MRI with advanced diffusion weighted imaging for breast cancer detection in a lactating woman. Radiol. Case Rep. 2020, 15, 2358–2361. [Google Scholar] [CrossRef] [Scilit]
  33. Fardanesh, R.; Marino, M.A.; Avendano, D.; Leithner, D.; Pinker, K.; Thakur, S.B. Proton MR spectroscopy in the breast: Technical innovations and clinical applications. J. Magn. Reson. Imaging 2019, 50, 1033–1046. [Google Scholar] [CrossRef] [Scilit]
  34. Galati, F.; Trimboli, R.M.; Pediconi, F. Special Issue “Advances in Breast MRI”. Diagnostics 2021, 11, 2297. [Google Scholar] [CrossRef] [Scilit]
  35. Yang, H.; Xu, Y.; Zhao, Y.; Yin, J.; Chen, Z.; Huang, P. The role of tissue elasticity in the differential diagnosis of benign and malignant breast lesions using shear wave elastography. BMC Cancer 2020, 20, 930. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  36. Abdelgawad, M.S.; Elseady, B.A.; ELabd, O.L.; Kohla, M.S.; Samea, M.E.S.A. Comparison of magnetic resonance elastography and diffusion-weighted imaging for differentiating benign and malignant liver lesions. Egypt. J. Radiol. Nucl. Med. 2024, 55, 38. [Google Scholar] [CrossRef] [Scilit]
  37. Patel, B.K.; Samreen, N.; Zhou, Y.; Chen, J.; Brandt, K.; Ehman, R.; Pepin, K. MR elastography of the breast: Evolution of technique, case examples, and future directions. Clin. Breast Cancer 2021, 21, e102–e111. [Google Scholar] [CrossRef] [Scilit]
  38. Tafreshi, N.K.; Kumar, V.; Morse, D.L.; Gatenby, R.A. Molecular and functional imaging of breast cancer. Cancer Control 2010, 17, 143–155. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  39. Azab, E.A.; Ibrahim, M.E. Diffusion weighted (DW) MRI role in characterization of breast lesions using absolute and normalized ADC values. Egypt. J. Radiol. Nucl. Med. 2018, 49, 564–570. [Google Scholar] [CrossRef] [Scilit]
  40. Fusco, R.; Granata, V.; Maio, F.; Sansone, M.; Petrillo, A. Textural radiomic features and time-intensity curve data analysis by dynamic contrast-enhanced MRI for early prediction of breast cancer therapy response: Preliminary data. Eur. Radiol. Exp. 2020, 4, 8. [Google Scholar] [CrossRef] [Scilit]
  41. Hunt, K.N. Molecular breast imaging: A scientific review. J. Breast Imaging 2021, 3, 416–426. [Google Scholar] [CrossRef] [Scilit]
  42. Covington, M.F.; Parent, E.E.; Dibble, E.H.; Rauch, G.M.; Fowler, A.M. Advances and future directions in molecular breast imaging. J. Nucl. Med. 2022, 63, 17–21. [Google Scholar] [CrossRef] [Scilit]
  43. Muzahir, S. Molecular breast cancer imaging in the era of precision medicine. Am. J. Roentgenol. 2020, 215, 1512–1519. [Google Scholar] [CrossRef] [Scilit]
  44. Adrada, B.E.; Moseley, T.; Kappadath, S.C.; Whitman, G.J.; Rauch, G.M. Molecular breast imaging-guided percutaneous biopsy of breast lesions: A new frontier on breast intervention. J. Breast Imaging 2020, 2, 484–491. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  45. Hruska, C.B.; Corion, C.; de Geus-Oei, L.F.; Adrada, B.E.; Fowler, A.M.; Hunt, K.N.; Kappadath, S.C.; Pilkington, P.; Arias-Bouda, L.M.P.; Rauch, G.M. SNMMI procedure standard/EANM practice guideline for molecular breast imaging with dedicated γ-cameras. J. Nucl. Med. Technol. 2022, 50, 103–110. [Google Scholar] [CrossRef] [Scilit]
  46. van Loevezijn, A.A.; Corion, C.L.; Zeillemaker, A.M.; Wijers, L.M.; Smithuis, R.H.; Olmos, R.A.V.; van der Hage, J.A.; de Geus-Oei, L.F.; Benard, M.; Arias-Bouda, L.M.P. Clinical impact of molecular breast imaging as adjunct diagnostic modality in evaluation of indeterminate breast abnormalities and unresolved diagnostic concerns. Nucl. Med. Commun. 2023, 44, 417–426. [Google Scholar] [CrossRef] [Scilit]
  47. Pereslucha, A.M.; Wenger, D.M.; Morris, M.F.; Aydi, Z.B. Invasive lobular carcinoma: A review of imaging modalities with special focus on pathology concordance. Healthcare 2023, 11, 746. [Google Scholar] [CrossRef] [Scilit]
  48. Robson, N.; Thekkinkattil, D.K. Current Role and Future Prospects of Positron Emission Tomography (PET)/Computed Tomography (CT) in the Management of Breast Cancer. Medicina 2024, 60, 321. [Google Scholar] [CrossRef] [Scilit]
  49. Ko, H.; Baghdadi, Y.; Love, C.; Sparano, J.A. Clinical utility of 18F-FDG PET/CT in staging localized breast cancer before initiating preoperative systemic therapy. J. Natl. Compr. Cancer Netw. 2020, 18, 1240–1246. [Google Scholar] [CrossRef] [Scilit]
  50. Failla, M.; Floresta, G.; Abbate, V. Peptide-based positron emission tomography probes: Current strategies for synthesis and radiolabelling. RSC Med. Chem. 2023, 14, 592–623. [Google Scholar] [CrossRef] [Scilit]
  51. Burkett, B.J.; Johnson, D.R.; Lowe, V.J. Evaluation of Neurodegenerative Disorders with Amyloid-β, Tau, and Dopaminergic PET Imaging: Interpretation Pitfalls. J. Nucl. Med. 2024, 65, 829–837. [Google Scholar] [CrossRef] [Scilit]
  52. Cherry, S.R.; Diekmann, J.; Bengel, F.M. Total-body positron emission tomography: Adding new perspectives to cardiovascular research. Cardiovasc. Imaging 2023, 16, 1335–1347. [Google Scholar]
  53. Beňačka, R.; Szabóová, D.; Gul’ašová, Z.; Hertelyová, Z.; Radoňák, J. Classic and new markers in diagnostics and classification of breast cancer. Cancers 2022, 14, 5444. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  54. Jones, L.I.; Marshall, A.; Geach, R.; Elangovan, P.; O’Flynn, E.; Timlin, T.; McKeown-Keegan, S.; Rose, J.; Vinnicombe, S.; Taylor-Phillips, S.; et al. Optimising the diagnostic accuracy of First post-contrAst SubtracTed breast MRI (FAST MRI) through interpretation-training: A multicentre e-learning study, mapping the learning curve of NHS Breast Screening Programme (NHSBSP) mammogram readers using an enriched dataset. Breast Cancer Res. 2024, 26, 85. [Google Scholar] [PubMed]
  55. Hashimoto, K.; Nishimura, S.; Ito, T.; Oka, N.; Akagi, M. Limitations and usefulness of biopsy techniques for the diagnosis of metastatic bone and soft tissue tumors. Ann. Med. Surg. 2021, 68, 102581. [Google Scholar] [CrossRef] [Scilit]
  56. Borowsky, A.D.; Glassy, E.F.; Wallace, W.D.; Kallichanda, N.S.; Behling, C.A.; Miller, D.V.; Oswal, H.N.; Feddersen, R.M.; Bakhtar, O.R.; Mendoza, A.E.; et al. Digital whole slide imaging compared with light microscopy for primary diagnosis in surgical pathology: A multicenter, double-blinded, randomized study of 2045 cases. Arch. Pathol. Lab. Med. 2020, 144, 1245–1253. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  57. Yin, L.; Duan, J.J.; Bian, X.W.; Yu, S.c. Triple-negative breast cancer molecular subtyping and treatment progress. Breast Cancer Res. 2020, 22, 61. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  58. Bruzas, S.; Kuemmel, S.; Harrach, H.; Breit, E.; Ataseven, B.; Traut, A.; Rüland, A.; Kostara, A.; Chiari, O.; Dittmer-Grabowski, C.; et al. Next-generation sequencing-directed therapy in patients with metastatic breast cancer in routine clinical practice. Cancers 2021, 13, 4564. [Google Scholar] [CrossRef] [Scilit]
  59. Abd El-Sadek, I.; Miyazawa, A.; Shen, L.T.W.; Makita, S.; Mukherjee, P.; Lichtenegger, A.; Matsusaka, S.; Yasuno, Y. Three-dimensional dynamics optical coherence tomography for tumor spheroid evaluation. Biomed. Opt. Express 2021, 12, 6844–6863. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  60. Morato, N.M.; Brown, H.M.; Garcia, D.; Middlebrooks, E.H.; Jentoft, M.; Chaichana, K.; Quiñones-Hinojosa, A.; Cooks, R.G. High-throughput analysis of tissue microarrays using automated desorption electrospray ionization mass spectrometry. Sci. Rep. 2022, 12, 18851. [Google Scholar] [CrossRef] [Scilit]
  61. Bitencourt, A.G.; Lima, E.N.; Macedo, B.R.; Conrado, J.L.; Marques, E.F.; Chojniak, R. Can positron emission mammography help to identify clinically significant breast cancer in women with suspicious calcifications on mammography? Eur. Radiol. 2017, 27, 1893–1900. [Google Scholar] [CrossRef] [Scilit]
  62. Kim, Y.J.; Seo, J.Y.; Kim, K.W.; Hwang, C.M.; Oh, D.H. The usefulness of addition of breast-specific gamma imaging to mammography in women with dense breast. Egypt. J. Radiol. Nucl. Med. 2023, 54, 94. [Google Scholar] [CrossRef] [Scilit]
  63. Kaur, A.; Kaushal, C.; Sandhu, J.K.; Damaševičius, R.; Thakur, N. Histopathological Image Diagnosis for Breast Cancer Diagnosis Based on Deep Mutual Learning. Diagnostics 2023, 14, 95. [Google Scholar] [CrossRef] [Scilit]
  64. Adam, R.; Dell’Aquila, K.; Hodges, L.; Maldjian, T.; Duong, T.Q. Deep learning applications to breast cancer detection by magnetic resonance imaging: A literature review. Breast Cancer Res. 2023, 25, 87. [Google Scholar] [CrossRef] [Scilit]
  65. Gautam, A. Recent advancements of deep learning in detecting breast cancer: A survey. Multimed. Syst. 2023, 29, 917–943. [Google Scholar] [CrossRef] [Scilit]
  66. Hamed, G.; Marey, M.A.E.R.; Amin, S.E.S.; Tolba, M.F. Deep learning in breast cancer detection and classification. In Proceedings of the Proceedings of the International Conference on Artificial Intelligence and Computer Vision (AICV2020); Springer: Berlin/Heidelberg, Germany, 2020; pp. 322–333. [Google Scholar]
  67. Nemade, V.; Pathak, S.; Dubey, A.K. A systematic literature review of breast cancer diagnosis using machine intelligence techniques. Arch. Comput. Methods Eng. 2022, 29, 4401–4430. [Google Scholar] [CrossRef] [Scilit]
  68. Gastounioti, A.; Desai, S.; Ahluwalia, V.S.; Conant, E.F.; Kontos, D. Artificial intelligence in mammographic phenotyping of breast cancer risk: A narrative review. Breast Cancer Res. 2022, 24, 14. [Google Scholar] [CrossRef] [Scilit]
  69. Kaba Gurmessa, D.; Jimma, W. Explainable machine learning for breast cancer diagnosis from Mammography and ultrasound images: A systematic review. BMJ Health Care Inform. 2024, 31, e100954. [Google Scholar] [CrossRef] [Scilit]
  70. Nasser, M.; Yusof, U.K. Deep learning based methods for breast cancer diagnosis: A systematic review and future direction. Diagnostics 2023, 13, 161. [Google Scholar] [CrossRef] [Scilit]
  71. Masud, M.; Eldin Rashed, A.E.; Hossain, M.S. Convolutional neural network-based models for diagnosis of breast cancer. Neural Comput. Appl. 2022, 34, 11383–11394. [Google Scholar] [CrossRef] [Scilit]
  72. Alam, T.; Shia, W.C.; Hsu, F.R.; Hassan, T. Improving Breast Cancer Detection and Diagnosis through Semantic Segmentation Using the Unet3+ Deep Learning Framework. Biomedicines 2023, 11, 1536. [Google Scholar] [CrossRef] [Scilit]
  73. Honnahalli, S.S.; Tiwari, H.; Chitragar, D.V. Future Fusion+ UNet (R2U-Net) Deep Learning Architecture for Breast Mass Segmentation. Eng. Proc. 2023, 59, 44. [Google Scholar]
  74. Gad, E.; Abou Khatwa, M.; A. Elattar, M.; Selim, S. A Novel Approach to Breast Cancer Segmentation Using U-Net Model with Attention Mechanisms and FedProx. In Proceedings of the Annual Conference on Medical Image Understanding and Analysis; Springer: Berlin/Heidelberg, Germany, 2023; pp. 310–324. [Google Scholar]
  75. Michael, E.; Ma, H.; Qi, S. Breast tumor segmentation in ultrasound images based on u-NET model. In Proceedings of the The International Conference on Innovations in Computing Research; Springer: Berlin/Heidelberg, Germany, 2022; pp. 22–31. [Google Scholar]
  76. Zhu, C.; Chai, X.; Xiao, Y.; Liu, X.; Zhang, R.; Yang, Z.; Wang, Z. Swin-Net: A Swin-Transformer-Based Network Combing with Multi-Scale Features for Segmentation of Breast Tumor Ultrasound Images. Diagnostics 2024, 14, 269. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  77. Qin, C.; Wu, Y.; Zeng, J.; Tian, L.; Zhai, Y.; Li, F.; Zhang, X. Joint transformer and multi-scale CNN for DCE-MRI breast cancer segmentation. Soft Comput. 2022, 26, 8317–8334. [Google Scholar] [CrossRef] [Scilit]
  78. Li, W.; Long, H.; Zhan, X.; Wu, Y. MDAA: Multi-scale and dual-adaptive attention network for breast cancer classification. Signal Image Video Process. 2024, 18, 3133–3143. [Google Scholar] [CrossRef] [Scilit]
  79. Lee, J.; Lee, G.; Kwak, T.Y.; Kim, S.W.; Jin, M.S.; Kim, C.; Chang, H. MurSS: A Multi-Resolution Selective Segmentation Model for Breast Cancer. Bioengineering 2024, 11, 463. [Google Scholar] [CrossRef] [Scilit]
  80. Praveena, P.; Kumar, N.S. Breast Cancer Detection Through Semantic Segmentation of MRI Images with DMFTNet. SN Comput. Sci. 2023, 4, 448. [Google Scholar] [CrossRef] [Scilit]
  81. Thakur, A.; Gupta, M.; Sinha, D.K.; Mishra, K.K.; Venkatesan, V.K.; Guluwadi, S. Transformative Breast Cancer Diagnosis using CNNs with Optimized ReduceLROnPlateau and Early Stopping Enhancements. Int. J. Comput. Intell. Syst. 2024, 17, 14. [Google Scholar]
  82. Arooj, S.; Zubair, M.; Khan, M.F.; Alissa, K.; Khan, M.A.; Mosavi, A. Breast cancer detection and classification empowered with transfer learning. Front. Public Health 2022, 10, 924432. [Google Scholar] [CrossRef] [Scilit]
  83. Azevedo, V.; Silva, C.; Dutra, I. Quantum transfer learning for breast cancer detection. Quantum Mach. Intell. 2022, 4, 5. [Google Scholar] [CrossRef] [Scilit]
  84. Ayana, G.; Choe, S.W. BUVITNET: Breast ultrasound detection via vision transformers. Diagnostics 2022, 12, 2654. [Google Scholar] [CrossRef] [Scilit]
  85. Chen, X.; Zhang, K.; Abdoli, N.; Gilley, P.W.; Wang, X.; Liu, H.; Zheng, B.; Qiu, Y. Transformers improve breast cancer diagnosis from unregistered multi-view mammograms. Diagnostics 2022, 12, 1549. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  86. Zhang, M.; Dong, B.; Li, Q. MS-GWNN: Multi-scale graph wavelet neural network for breast cancer diagnosis. In Proceedings of the 2022 IEEE 19th International Symposium on Biomedical Imaging (ISBI); IEEE: New York, NY, USA, 2022; pp. 1–5. [Google Scholar]
  87. Nóbrega, C.; Marinho, L. Towards explaining recommendations through local surrogate models. In Proceedings of the 34th ACM/SIGAPP Symposium on Applied Computing, Limassol, Cyprus, 8–12 April 2019; pp. 1671–1678. [Google Scholar]
  88. Prananda, A.R.; Frannita, E.L. Toward Better Analysis of Breast Cancer Diagnosis: Interpretable AI for Breast Cancer Classification. IT J. Res. Dev. 2023, 7, 220–227. [Google Scholar] [CrossRef] [Scilit]
  89. Ahmed, M.; Bibi, T.; Khan, R.A.; Nasir, S. Enhancing Breast Cancer Diagnosis in Mammography: Evaluation and Integration of Convolutional Neural Networks and Explainable AI. arXiv 2024, arXiv:2404.03892. [Google Scholar] [CrossRef] [Scilit]
  90. Hakkoum, H.; Idri, A.; Abnane, I. Artificial neural networks interpretation using LIME for breast cancer diagnosis. In Proceedings of the Trends and Innovations in Information Systems and Technologies; Springer: Berlin/Heidelberg, Germany, 2020; Volume 38, pp. 15–24. [Google Scholar]
  91. Linardatos, P.; Papastefanopoulos, V.; Kotsiantis, S. Explainable AI: A review of machine learning interpretability methods. Entropy 2020, 23, 18. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  92. Jansen, T.; Geleijnse, G.; Van Maaren, M.; Hendriks, M.P.; Ten Teije, A.; Moncada-Torres, A. Machine learning explainability in breast cancer survival. Stud. Health Technol. Inform. 2020, 270, 307–311. [Google Scholar]
  93. Zhang, G.; Shi, Y.; Yin, P.; Liu, F.; Fang, Y.; Li, X.; Zhang, Q.; Zhang, Z. A machine learning model based on ultrasound image features to assess the risk of sentinel lymph node metastasis in breast cancer patients: Applications of scikit-learn and SHAP. Front. Oncol. 2022, 12, 944569. [Google Scholar] [CrossRef] [Scilit]
  94. Xiao, M.; Zhang, L.; Shi, W.; Liu, J.; He, W.; Jiang, Z. A visualization method based on the Grad-CAM for medical image segmentation model. In Proceedings of the 2021 International Conference on Electronic Information Engineering and Computer Science (EIECS); IEEE: New York, NY, USA, 2021; pp. 242–247. [Google Scholar]
  95. Vani Kumari, S.; Usha Rani, K. Predicting of Breast Cancer using Grad-CAM and Bounding Box. J. Chem. Health Risks 2023, 13, 411–420. [Google Scholar]
  96. Zhang, Z.; Wang, W.; Wang, X.; Yu, X.; Zhu, Y.; Zhan, H.; Chen, Z.; Li, B.; Huang, J. Breast-specific gamma imaging or ultrasonography as adjunct imaging diagnostics in women with mammographically dense breasts. Eur. Radiol. 2020, 30, 6062–6071. [Google Scholar] [CrossRef] [Scilit]
  97. Yan, R.; Ren, F.; Wang, Z.; Wang, L.; Zhang, T.; Liu, Y.; Rao, X.; Zheng, C.; Zhang, F. Breast cancer histopathological image classification using a hybrid deep neural network. Methods 2020, 173, 52–60. [Google Scholar] [CrossRef] [Scilit]
  98. Drioua, W.R.; Benamrane, N.; Sais, L. Breast Cancer Histopathological Images Segmentation Using Deep Learning. Sensors 2023, 23, 7318. [Google Scholar] [CrossRef] [Scilit]
  99. McKinney, S.M.; Sieniek, M.; Godbole, V.; Godwin, J.; Antropova, N.; Ashrafian, H.; Back, T.; Chesus, M.; Corrado, G.S.; Darzi, A.; et al. International evaluation of an AI system for breast cancer screening. Nature 2020, 577, 89–94. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  100. Sabry, M.; Balaha, H.M.; Ali, K.M.; Soliman, T.H.A.; Gondim, D.; Ghazal, M.; Alghamdi, N.S.; El-Baz, A. Enhancing Breast Cancer Diagnosis with Multi-Resolution Vision Transformers and Robust Decision-Making. IEEE Access 2025, 13, 89704–89722. [Google Scholar] [CrossRef] [Scilit]
  101. Page, M.J.; McKenzie, J.E.; Bossuyt, P.M.; Boutron, I.; Hoffmann, T.C.; Mulrow, C.D.; Shamseer, L.; Tetzlaff, J.M.; Moher, D. Updating guidance for reporting systematic reviews: Development of the PRISMA 2020 statement. J. Clin. Epidemiol. 2021, 134, 103–112. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  102. Haddaway, N.R.; Page, M.J.; Pritchard, C.C.; McGuinness, L.A. PRISMA2020: An R package and Shiny app for producing PRISMA 2020-compliant flow diagrams, with interactivity for optimised digital transparency and Open Synthesis. Campbell Syst. Rev. 2022, 18, e1230. [Google Scholar] [CrossRef] [Scilit]
  103. Prodan, M.; Paraschiv, E.; Stanciu, A. Applying deep learning methods for mammography analysis and breast cancer detection. Appl. Sci. 2023, 13, 4272. [Google Scholar] [CrossRef] [Scilit]
  104. Pesapane, F.; Trentin, C.; Ferrari, F.; Signorelli, G.; Tantrige, P.; Montesano, M.; Cicala, C.; Virgoli, R.; D’Acquisto, S.; Nicosia, L.; et al. Deep learning performance for detection and classification of microcalcifications on mammography. Eur. Radiol. Exp. 2023, 7, 69. [Google Scholar] [CrossRef] [Scilit]
  105. Siddique, M.; Liu, M.; Duong, P.; Jambawalikar, S.; Ha, R. Deep Learning Approaches with Digital Mammography for Evaluating Breast Cancer Risk, a Narrative Review. Tomography 2023, 9, 1110–1119. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  106. Pedemonte, S.; Tsue, T.; Mombourquette, B.; Vu, Y.N.T.; Matthews, T.; Hoil, R.M.; Shah, M.; Ghare, N.; Zingman-Daniels, N.; Holley, S.; et al. A deep learning algorithm for reducing false positives in screening mammography. arXiv 2022, arXiv:2204.06671. [Google Scholar] [CrossRef] [Scilit]
  107. Khamparia, A.; Bharati, S.; Podder, P.; Gupta, D.; Khanna, A.; Phung, T.K.; Thanh, D.N. Diagnosis of breast cancer based on modern mammography using hybrid transfer learning. Multidimens. Syst. Signal Process. 2021, 32, 747–765. [Google Scholar] [CrossRef] [Scilit]
  108. Sechopoulos, I.; Teuwen, J.; Mann, R. Artificial intelligence for breast cancer detection in mammography and digital breast tomosynthesis: State of the art. In Proceedings of the Seminars in Cancer Biology; Elsevier: Amsterdam, The Netherlands, 2021; Volume 72, pp. 214–225. [Google Scholar]
  109. Dontchos, B.N.; Achibiri, J.; Mercaldo, S.F.; Wang, G.X.; Lamb, L.R.; Miles, R.C.; Narayan, A.K.; Lehman, C.D. Disparities in same-day diagnostic imaging in breast cancer screening: Impact of an immediate-read screening mammography program implemented during the COVID-19 pandemic. Am. J. Roentgenol. 2022, 218, 270–278. [Google Scholar] [CrossRef] [Scilit]
  110. Alshammari, M.M.; Almuhanna, A.; Alhiyafi, J. Mammography image-based diagnosis of breast cancer using machine learning: A pilot study. Sensors 2021, 22, 203. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  111. Kavitha, T.; Mathai, P.; Karthikeyan, C.; Ashok, M.; Kohar, R.; Avanija, J.; Neelakandan, S. Deep learning based capsule neural network model for breast cancer diagnosis using mammogram images. Interdiscip. Sci. Comput. Life Sci. 2022, 14, 113–129. [Google Scholar] [CrossRef] [Scilit]
  112. Vocaturo, E.; Zumpano, E. Artificial intelligence approaches on ultrasound for breast cancer diagnosis. In Proceedings of the 2021 IEEE International Conference on Bioinformatics and Biomedicine (BIBM); IEEE: New York, NY, USA, 2021; pp. 3116–3121. [Google Scholar]
  113. Yadav, A.C.; Kolekar, M.H.; Zope, M.K. ResNet-101 Empowered Deep Learning for Breast Cancer Ultrasound Image Classification. In Proceedings of the 17th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2024); SCITEPRESS: Setúbal, Portugal, 2024; Volume 1, pp. 763–769. [Google Scholar]
  114. Chowa, S.S.; Azam, S.; Montaha, S.; Payel, I.J.; Bhuiyan, M.R.I.; Hasan, M.Z.; Jonkman, M. Graph neural network-based breast cancer diagnosis using ultrasound images with optimized graph construction integrating the medically significant features. J. Cancer Res. Clin. Oncol. 2023, 149, 18039–18064. [Google Scholar] [CrossRef] [Scilit]
  115. Rezazadeh, A.; Jafarian, Y.; Kord, A. Explainable ensemble machine learning for breast cancer diagnosis based on ultrasound image texture features. Forecasting 2022, 4, 262–274. [Google Scholar] [CrossRef] [Scilit]
  116. Zakareya, S.; Izadkhah, H.; Karimpour, J. A new deep-learning-based model for breast cancer diagnosis from medical images. Diagnostics 2023, 13, 1944. [Google Scholar] [CrossRef] [Scilit]
  117. Brunetti, N.; Calabrese, M.; Martinoli, C.; Tagliafico, A.S. Artificial intelligence in breast ultrasound: From diagnosis to prognosis—A rapid review. Diagnostics 2022, 13, 58. [Google Scholar] [CrossRef] [Scilit]
  118. Gu, Y.; Xu, W.; Lin, B.; An, X.; Tian, J.; Ran, H.; Ren, W.; Chang, C.; Yuan, J.; Kang, C.; et al. Deep learning based on ultrasound images assists breast lesion diagnosis in China: A multicenter diagnostic study. Insights Into Imaging 2022, 13, 124. [Google Scholar] [CrossRef] [Scilit]
  119. Iacob, R.; Iacob, E.R.; Stoicescu, E.R.; Ghenciu, D.M.; Cocolea, D.M.; Constantinescu, A.; Ghenciu, L.A.; Manolescu, D.L. Evaluating the Role of Breast Ultrasound in Early Detection of Breast Cancer in Low-and Middle-Income Countries: A Comprehensive Narrative Review. Bioengineering 2024, 11, 262. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  120. Soni, M.; Khan, I.R.; Babu, K.S.; Nasrullah, S.; Madduri, A.; Rahin, S.A. Light weighted healthcare CNN model to detect prostate cancer on multiparametric MRI. Comput. Intell. Neurosci. 2022, 2022, 5497120. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  121. Khaled, R.; Vidal, J.; Vilanova, J.C.; Martí, R. A U-Net Ensemble for breast lesion segmentation in DCE MRI. Comput. Biol. Med. 2022, 140, 105093. [Google Scholar] [CrossRef] [Scilit]
  122. Kazama, T.; Takahara, T.; Hashimoto, J. Breast cancer subtypes and quantitative magnetic resonance imaging: A systemic review. Life 2022, 12, 490. [Google Scholar] [CrossRef] [Scilit]
  123. Onishi, N.; Sadinski, M.; Hughes, M.C.; Ko, E.S.; Gibbs, P.; Gallagher, K.M.; Fung, M.M.; Hunt, T.J.; Martinez, D.F.; Shukla-Dave, A.; et al. Ultrafast dynamic contrast-enhanced breast MRI may generate prognostic imaging markers of breast cancer. Breast Cancer Res. 2020, 22, 58. [Google Scholar] [CrossRef] [Scilit]
  124. Yu, Y.; Ren, W.; He, Z.; Chen, Y.; Tan, Y.; Mao, L.; Ouyang, W.; Lu, N.; Ouyang, J.; Chen, K.; et al. Machine learning radiomics of magnetic resonance imaging predicts recurrence-free survival after surgery and correlation of LncRNAs in patients with breast cancer: A multicenter cohort study. Breast Cancer Res. 2023, 25, 132. [Google Scholar] [CrossRef] [Scilit]
  125. Zhang, X.; Teng, X.; Zhang, J.; Lai, Q.; Cai, J. Enhancing pathological complete response prediction in breast cancer: The role of dynamic characterization of DCE-MRI and its association with tumor heterogeneity. Breast Cancer Res. 2024, 26, 77. [Google Scholar] [CrossRef] [Scilit]
  126. Xiao, J.; Rahbar, H.; Hippe, D.S.; Rendi, M.H.; Parker, E.U.; Shekar, N.; Hirano, M.; Cheung, K.J.; Partridge, S.C. Dynamic contrast-enhanced breast MRI features correlate with invasive breast cancer angiogenesis. npj Breast Cancer 2021, 7, 42. [Google Scholar] [CrossRef] [Scilit]
  127. He, M.; Ruan, H.; Ma, M.; Zhang, Z. Application of diffusion weighted imaging techniques for differentiating benign and malignant breast lesions. Front. Oncol. 2021, 11, 694634. [Google Scholar] [CrossRef] [Scilit]
  128. Hu, Q.; Whitney, H.M.; Li, H.; Ji, Y.; Liu, P.; Giger, M.L. Improved classification of benign and malignant breast lesions using deep feature maximum intensity projection MRI in breast cancer diagnosis using dynamic contrast-enhanced MRI. Radiol. Artif. Intell. 2021, 3, e200159. [Google Scholar] [CrossRef] [Scilit]
  129. Ayatollahi, F.; Shokouhi, S.B.; Mann, R.M.; Teuwen, J. Automatic breast lesion detection in ultrafast DCE-MRI using deep learning. Med. Phys. 2021, 48, 5897–5907. [Google Scholar] [CrossRef] [Scilit]
  130. Hruska, C.B.; Geske, J.R.; Conners, A.L.; Whaley, D.H.; Rhodes, D.J.; O’Connor, M.K.; Carter, R.E.; Scott, C.G.; Vachon, C.M. Background parenchymal uptake on molecular breast imaging and breast cancer risk: A cohort study. Am. J. Roentgenol. 2021, 216, 1193–1204. [Google Scholar] [CrossRef] [Scilit]
  131. Hunt, K.N.; Conners, A.L.; Gray, L.; Hruska, C.B.; O’connor, M.K. Molecular breast imaging biopsy with a dual-detector system. Radiol. Imaging Cancer 2024, 6, e230186. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  132. Carter, R.E.; Attia, Z.I.; Geske, J.R.; Conners, A.L.; Whaley, D.H.; Hunt, K.N.; O’Connor, M.K.; Rhodes, D.J.; Hruska, C.B. Classification of background parenchymal uptake on molecular breast imaging using a convolutional neural network. JCO Clin. Cancer Inform. 2019, 3, 1–11. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  133. Mann, R.M.; Hooley, R.; Barr, R.G.; Moy, L. Novel approaches to screening for breast cancer. Radiology 2020, 297, 266–285. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  134. Dibble, E.H.; Hunt, K.N.; Ehman, E.C.; O’Connor, M.K. Molecular breast imaging in clinical practice. Am. J. Roentgenol. 2020, 215, 277–284. [Google Scholar] [CrossRef] [Scilit]
  135. Tao, A.T.; Hruska, C.B.; Conners, A.L.; Hunt, K.N.; Swanson, T.N.; Tran, T.D.; Manduca, A.; Borges, L.; Maidment, A.D.; Lake, D.; et al. Dose reduction in molecular breast imaging with a new image-processing algorithm. Am. J. Roentgenol. 2020, 214, 185–193. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  136. Jimenez-Mesa, C.; Arco, J.E.; Martinez-Murcia, F.J.; Suckling, J.; Ramirez, J.; Gorriz, J.M. Applications of machine learning and deep learning in SPECT and PET imaging: General overview, challenges and future prospects. Pharmacol. Res. 2023, 197, 106984. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  137. Hellwig, D.; Hellwig, N.C.; Boehner, S.; Fuchs, T.; Fischer, R.; Schmidt, D. Artificial intelligence and deep learning for advancing PET image reconstruction: State-of-the-art and future directions. Nuklearmedizin-NuclearMedicine 2023, 62, 334–342. [Google Scholar] [CrossRef] [Scilit]
  138. Sanaat, A.; Zaidi, H. Depth of interaction estimation in a preclinical PET scanner equipped with monolithic crystals coupled to SiPMs using a deep neural network. Appl. Sci. 2020, 10, 4753. [Google Scholar] [CrossRef] [Scilit]
  139. Herraiz, J.L.; Bembibre, A.; Lopez-Montes, A. Deep-learning based positron range correction of PET images. Appl. Sci. 2020, 11, 266. [Google Scholar] [CrossRef] [Scilit]
  140. Hashimoto, F.; Onishi, Y.; Ote, K.; Tashima, H.; Reader, A.J.; Yamaya, T. Correction to: Deep learning-based PET image denoising and reconstruction: A review. Radiol. Phys. Technol. 2024, 17, 580. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  141. Chaudhari, A.S.; Mittra, E.; Davidzon, G.A.; Gulaka, P.; Gandhi, H.; Brown, A.; Zhang, T.; Srinivas, S.; Gong, E.; Zaharchuk, G.; et al. Author Correction: Low-count whole-body PET with deep learning in a multicenter and externally validated study. npj Digit. Med. 2021, 4, 139. [Google Scholar] [CrossRef] [Scilit]
  142. Artesani, A.; Bruno, A.; Gelardi, F.; Chiti, A. Empowering PET: Harnessing deep learning for improved clinical insight. Eur. Radiol. Exp. 2024, 8, 17. [Google Scholar] [CrossRef] [Scilit]
  143. Spadea, M.F.; Maspero, M.; Zaffino, P.; Seco, J. Deep learning based synthetic-CT generation in radiotherapy and PET: A review. Med. Phys. 2021, 48, 6537–6566. [Google Scholar] [CrossRef] [Scilit]
  144. Lim, H.J.; Ong, C.A.J.; Tan, J.W.S.; Teo, M.C.C. Utility of positron emission tomography/computed tomography (PET/CT) imaging in the evaluation of sarcomas: A systematic review. Crit. Rev. Oncol. 2019, 143, 1–13. [Google Scholar] [CrossRef] [Scilit]
  145. Mehranian, A.; Reader, A.J. Model-based deep learning PET image reconstruction using forward–backward splitting expectation–maximization. IEEE Trans. Radiat. Plasma Med. Sci. 2020, 5, 54–64. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  146. Thomas, A.M.; Adithya, G.; Arunselvan, A.; Karthik, R. Detection of breast cancer from histopathological images using image processing and deep-learning. In Proceedings of the 2022 Third International Conference on Intelligent Computing Instrumentation and Control Technologies (ICICICT); IEEE: New York, NY, USA, 2022; pp. 1008–1015. [Google Scholar]
  147. Sui, D.; Liu, W.; Chen, J.; Zhao, C.; Ma, X.; Guo, M.; Tian, Z. A pyramid architecture-based deep learning framework for breast cancer detection. Biomed Res. Int. 2021, 2021, 2567202. [Google Scholar] [CrossRef] [Scilit]
  148. Zou, Y.; Chen, S.; Sun, Q.; Liu, B.; Zhang, J. DCET-Net: Dual-stream convolution expanded transformer for breast cancer histopathological image classification. In Proceedings of the 2021 IEEE International Conference on Bioinformatics and Biomedicine (BIBM); IEEE: New York, NY, USA, 2021; pp. 1235–1240. [Google Scholar]
  149. He, Z.; Lin, M.; Xu, Z.; Yao, Z.; Chen, H.; Alhudhaif, A.; Alenezi, F. Deconv-transformer (DecT): A histopathological image classification model for breast cancer based on color deconvolution and transformer architecture. Inf. Sci. 2022, 608, 1093–1112. [Google Scholar] [CrossRef] [Scilit]
  150. Alirezazadeh, P.; Hejrati, B.; Monsef-Esfahani, A.; Fathi, A. Representation learning-based unsupervised domain adaptation for classification of breast cancer histopathology images. Biocybern. Biomed. Eng. 2018, 38, 671–683. [Google Scholar] [CrossRef] [Scilit]
  151. Hong, R.; Liu, W.; DeLair, D.; Razavian, N.; Fenyö, D. Predicting endometrial cancer subtypes and molecular features from histopathology images using multi-resolution deep learning models. Cell Rep. Med. 2021, 2, 100400. [Google Scholar] [CrossRef] [Scilit]
  152. Menezes, G.L.; Mann, R.M.; Meeuwis, C.; Bisschops, B.; Veltman, J.; Lavin, P.T.; van de Vijver, M.J.; Pijnappel, R.M. Optoacoustic imaging of the breast: Correlation with histopathology and histopathologic biomarkers. Eur. Radiol. 2019, 29, 6728–6740. [Google Scholar] [CrossRef] [Scilit]
  153. Sharma, S.; Mehra, R. Conventional machine learning and deep learning approach for multi-classification of breast cancer histopathology images—A comparative insight. J. Digit. Imaging 2020, 33, 632–654. [Google Scholar] [CrossRef] [Scilit]
  154. Singh, S.; Kumar, R. Histopathological image analysis for breast cancer detection using cubic SVM. In Proceedings of the 2020 7th International Conference on Signal Processing and Integrated Networks (SPIN); IEEE: New York, NY, USA, 2020; pp. 498–503. [Google Scholar]
  155. Mahmood, T.; Li, J.; Pei, Y.; Akhtar, F.; Imran, A.; Rehman, K.U. A brief survey on breast cancer diagnostic with deep learning schemes using multi-image modalities. IEEE Access 2020, 8, 165779–165809. [Google Scholar] [CrossRef] [Scilit]
  156. Lokaj, B.; Pugliese, M.T.; Kinkel, K.; Lovis, C.; Schmid, J. Barriers and facilitators of artificial intelligence conception and implementation for breast imaging diagnosis in clinical practice: A scoping review. Eur. Radiol. 2024, 34, 2096–2109. [Google Scholar] [CrossRef] [Scilit]
  157. Demetriou, D.; Lockhat, Z.; Brzozowski, L.; Saini, K.S.; Dlamini, Z.; Hull, R. The Convergence of Radiology and Genomics: Advancing Breast Cancer Diagnosis with Radiogenomics. Cancers 2024, 16, 1076. [Google Scholar] [CrossRef] [Scilit]
  158. Page, M.J.; McKenzie, J.E.; Bossuyt, P.M.; Boutron, I.; Hoffmann, T.C.; Mulrow, C.D.; Shamseer, L.; Tetzlaff, J.M.; Akl, E.A.; Brennan, S.E.; et al. The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. BMJ 2021, 372, n71. [Google Scholar] [CrossRef] [Scilit] [PubMed]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Article Metrics

Citations

Article Access Statistics

Multiple requests from the same IP address are counted as one view.