Diagnostics, Volume 13, Issue 19 (October-1 2023) – 139 articles

Skin lesion classification plays a crucial role in dermatology, aiding in the early detection, diagnosis, and management of life-threatening malignant lesions. However, standalone transfer learning (TL) models failed to deliver optimal performance. In this study, we present an attention-enabled ensemble-based deep learning technique, a powerful, novel, and generalized method for extracting features for the classification of skin lesions. This technique holds significant promise in enhancing diagnostic accuracy by using seven pre-trained TL models for classification. Six ensemble-based DL (EBDL) models were created using stacking, softmax voting, and weighted average techniques. Furthermore, we investigated the attention mechanism as an effective paradigm and created seven attention-enabled transfer learning (aeTL) models before branching out to construct three attention-enabled ensemble-based DL (aeEBDL) models to create a reliable, adaptive, and generalized paradigm. The mean accuracy of the TL models is 95.30%, and the use of an ensemble-based paradigm increased it by 4.22%, to 99.52%. The aeTL models’ performance was superior to the TL models in accuracy by 3.01%, and aeEBDL models outperformed aeTL models by 1.29%. Statistical tests show significant p-value and Kappa coefficient along with a 99.6% reliability index for the aeEBDL models. The approach is highly effective and generalized for the classification of skin lesions. Full article

Aim: The study aims to test the hypothesis that concentrations of adropin and afamin differ between patients in various stages of chronic kidney disease when compared with healthy controls. The study also investigates the association of the biomarkers (adropin and afamin) with CKD-MBD and traditional cardiovascular risk parameters in CKD patients. Methodology: The cross-sectional study includes the subjects divided into four groups comprising the control group (healthy volunteers = 50), CKD stages 1–2 patients (n = 50), CKD stages 3–4 patients (n = 50), CKD stage 5 patients (n = 50). Serum concentrations of adropin and afamin were determined using ELISA. Clinical variables (renal, lipid, and CKD-MBD parameters) were correlated to adropin and afamin concentrations. Results: Afamin concentration was found to be higher in group IV, followed by groups III and II when compared to the control group, i.e., (83.243 ± 1.46, 64.233 ± 0.99, and 28.948 ± 0.72 vs. 14.476 ± 0.5) mg/L (p < 0.001), and adropin concentration was found to be lower in group IV as compared to groups III, II, and I (200.342 ± 8.37 vs. 284.682 ± 9.89 vs. 413.208 ± 12.32 vs. 706.542 ± 11.32) pg/mL (p < 0.001), respectively. Pearson correlation analysis showed that afamin was positively correlated with traditional cardiovascular risk biomarkers, while adropin showed a negative correlation. Conclusions: Adropin and afamin may potentially serve as futuristic predictors for the deterioration of renal function and may be involved in the pathological mechanisms of CKD and its associated complications such as CKD-MBD and high lipid levels. Full article

(1) Introduction: Although historically, the lung has been considered a sterile organ, recent studies through 16S rRNA gene sequencing have identified a substantial number of microorganisms. The human microbiome has been considered an “essential organ,” carrying about 150 times more information (genes) than are found in the entire human genome. The purpose of the present study is to characterize and compare the microbiome in three different interstitial lung diseases: idiopathic pulmonary fibrosis (IPF), hypersensitivity pneumonitis, and nondifferential interstitial lung disease. (2) Material and methods: This was a prospective cohort study where the DNA of 28 patients with ILD was extracted from the lavage and then processed using the standard technique of 16S RNA gene sequencing. In a tertiary teaching hospital in the northern, western part of Romania, samples were collected through bronchoscopy and then processed. (3) Results: The same four species were found in all the patients but in different quantities and compositions: Firmicutes, Actinobacteria, Proteobacteria and Bacteroides. Streptococcus was the most prevalent genus, followed by Staphylococcus and Prevotella. Statistically significant differences in the OUT count for the ten most abundant taxa were found for the genus: Gemella, Actinobacteria, Prevotella, Neisseria, Haemophilus, and Bifidobacterium. The comparative analysis showed a richer microbiota in patients with IPF, as shown by the alpha diversity index. (4) Conclusions: In interstitial lung diseases, the microorganisms normally found in the lung are reduced to a restricted flora dominated by the Firmicutes family. These changes significantly disrupt the continuity of the observed bacterial pattern from the oropharynx to the bronchial tree and lung, possibly impacting the evolution and severity of interstitial lung diseases. Full article

Maxillary sinus fungal ball (MSFB) is the most common type of non-invasive fungal rhinosinusitis. Since MSFB requires a unique treatment strategy and is associated with potentially severe complications, timely and precise diagnosis is crucial. Computed tomography (CT) is the first-line imaging tool for evaluating chronic rhinosinusitis. Accordingly, we aimed to investigate the clinical and CT imaging characteristics of MSFB. We retrospectively enrolled 97 patients with unilateral MSFB and 158 with unilateral non-fungal maxillary rhinosinusitis. The clinical characteristics, laboratory data, and CT imaging features of participants were evaluated. Older age, female sex, lower white blood cell and neutrophil counts, and CT imaging features (including an irregular surface, erosion of the medial sinus wall, sclerosis of the lateral sinus wall, and intralesional hyperdensity) were significantly associated with MSFB. The presence of adjacent maxillary odontogenic pathology was associated with a decreased likelihood of the incidence of MSFB in unilateral maxillary rhinosinusitis. Separate nomograms were created for patients, without and with the use of CT scan, to predict the probabilities of MSFB in patients with unilateral maxillary rhinosinusitis. We proposed two nomograms based on the clinical and CT characteristics of patients with MSFB. These could serve as evaluation tools to assist clinicians in determining the need for undergoing CT and facilitate the accurate and timely diagnosis of MSFB. Full article

The study utilizes osteosarcoma hematoxylin and the Eosin-stained image dataset, which is unevenly dispersed, and it raises concerns about the potential impact on the overall performance and reliability of any analyses or models derived from the dataset. In this study, a deep-learning-based convolution neural network (CNN) and adapted heterogeneous ensemble-learning-based voting classifier have been proposed to classify osteosarcoma. The proposed methods can also resolve the issue and develop unbiased learning models by introducing an evenly distributed training dataset. Data augmentation is employed to boost the generalization abilities. Six different pre-trained CNN models, namely MobileNetV1, Mo-bileNetV2, ResNetV250, InceptionV2, EfficientNetV2B0, and NasNetMobile, are applied and evaluated in frozen and fine-tuned-based phases. In addition, a novel CNN model and adapted heterogeneous ensemble-learning-based voting classifier developed from the proposed CNN model, fine-tuned NasNetMobile model, and fine-tuned Efficient-NetV2B0 model are also introduced to classify osteosarcoma. The proposed CNN model outperforms other pre-trained models. The Kappa score obtained from the proposed CNN model is 93.09%. Notably, the proposed voting classifier attains the highest Kappa score of 96.50% and outperforms all other models. The findings of this study have practical implications in telemedicine, mobile healthcare systems, and as a supportive tool for medical professionals. Full article

Background: Colonic endoscopic submucosal dissection (ESD) at “challenging sites” such as the cecum, ascending colon, and colonic flexures could be difficult even for expert endoscopists due to poor endoscope stability/maneuverability, steep angles, and thinner wall thickness. A double-balloon endoluminal intervention platform (EIP) has been introduced in the market to fasten and facilitate ESD, particularly when located at difficult sites. Here, we report our initial experience with an EIP comparing the outcomes of an EIP versus standard ESD (S-ESD) at “challenging sites”. Materials and methods: We retrospectively collected data on consecutive patients with colonic lesions located in the right colon and at flexures who underwent ESD in our tertiary referral center between March 2019 and May 2023. Endoscopic and clinical outcomes (technical success, en bloc resection rate, R0 resection rate, procedure time, time to reach the lesion, and adverse events) and 6-month follow-up outcomes were analyzed. Results: Overall, 139 consecutive patients with lesions located at these challenging sites were enrolled (EIP: 31 and S-ESD: 108). Demographic characteristics did not differ between groups. En bloc resection was achieved in 92.3% and 93.5% of patients, respectively, in the EIP and S-ESD groups. Both groups showed a comparable R0 resection rate (EIP vs. S-ESD: 92.3% vs. 97.2%). In patients undergoing EIP-assisted ESD, the total procedure time was shorter (96.1 [30.6] vs. 113.6 [42.3] minutes, p = 0.01), and the mean size of the resected lesions was smaller (46.2 ± 12.7 vs. 55.7 ± 17.6 mm, p = 0.003). The time to reach the lesion was significantly shorter in the EIP group (1.9 ± 0.3 vs. 8.2 ± 2.7 min, p ≤ 0.01). Procedure speed was comparable between groups (14.9 vs. 16.6 mm²/min, p = 0.29). Lower adverse events were observed in the EIP patients (3.8 vs. 10.2%, p = 0.31). Conclusions: EIP allows results that do not differ from S-ESD in the resection of colorectal superficial neoplasms localized in “challenging sites” in terms of efficacy and safety. EIP reduces the time to reach the lesions and may more safely facilitate endoscopic resection. Full article

Background. Technological advancement may bridge gaps between long-practiced medical competencies and modern technologies. Such a domain is the application of digital stethoscopes used for physical examination in telemedicine. This study aimed to validate the level of consensus among physicians regarding the interpretation of remote, digital auscultation of heart and lung sounds. Methods. Seven specialist physicians considered both the technical quality and clinical interpretation of auscultation findings of pre-recorded heart and lung sounds of patients hospitalized in their homes. TytoCare^TM system was used as a remote, digital stethoscope. Results. In total, 140 sounds (70 heart and 70 lungs) were presented to seven specialists. The level of agreement was measured using Fleiss’ Kappa (FK) variable. Agreement relating to heart sounds reached low-to-moderate consensus: the overall technical quality (FK = 0.199), rhythm regularity (FK = 0.328), presence of murmurs (FK = 0.469), appreciation of sounds as remote (FK = 0.011), and an overall diagnosis as normal or pathologic (FK = 0.304). The interpretation of some of the lung sounds reached a higher consensus: the overall technical quality (FK = 0.169), crepitus (FK = 0.514), wheezing (FK = 0.704), bronchial sounds (FK = 0.034), and an overall diagnosis as normal or pathological (FK = 0.386). Most Fleiss’ Kappa values were in the range of “fare consensus”, while in the domains of diagnosing lung crepitus and wheezing, the values increased to the “substantial” level. Conclusions. Bio signals, as recorded auscultations of the heart and lung sounds serving the process of clinical assessment of remotely situated patients, do not achieve a high enough level of agreement between specialized physicians. These findings should serve as a catalyzer for improving the process of telemedicine-attained bio-signals and their clinical interpretation. Full article

Glomeruli are interconnected capillaries in the renal cortex that are responsible for blood filtration. Damage to these glomeruli often signifies the presence of kidney disorders like glomerulonephritis and glomerulosclerosis, which can ultimately lead to chronic kidney disease and kidney failure. The timely detection of such conditions is essential for effective treatment. This paper proposes a modified UNet model to accurately detect glomeruli in whole-slide images of kidney tissue. The UNet model was modified by changing the number of filters and feature map dimensions from the first to the last layer to enhance the model’s capacity for feature extraction. Moreover, the depth of the UNet model was also improved by adding one more convolution block to both the encoder and decoder sections. The dataset used in the study comprised 20 large whole-side images. Due to their large size, the images were cropped into 512 × 512-pixel patches, resulting in a dataset comprising 50,486 images. The proposed model performed well, with 95.7% accuracy, 97.2% precision, 96.4% recall, and 96.7% F1-score. These results demonstrate the proposed model’s superior performance compared to the original UNet model, the UNet model with EfficientNetb3, and the current state-of-the-art. Based on these experimental findings, it has been determined that the proposed model accurately identifies glomeruli in extracted kidney patches. Full article

Chronic kidney disease (CKD) is a significant global health challenge that requires timely detection and accurate prognosis for effective treatment and management. The application of machine learning (ML) algorithms for CKD detection and prediction holds promising potential for improving patient outcomes. By incorporating key features which contribute to CKD, these algorithms enhance our ability to identify high-risk individuals and initiate timely interventions. This research highlights the importance of leveraging machine learning techniques to augment existing medical knowledge and improve the identification and management of kidney disease. In this paper, we explore the utilization of diverse ML algorithms, including gradient boost (GB), decision tree (DT), K-nearest neighbor (KNN), random forest (RF), histogram boost (HB), and XGBoost (XGB) to detect and predict chronic kidney disease (CKD). The aim is to improve early detection and prognosis, enhancing patient outcomes and reducing the burden on healthcare systems. We evaluated the performance of the ML algorithms using key metrics like accuracy, precision, recall, and F1 score. Additionally, we conducted feature significance analysis to identify the most influential characteristics in the detection and prediction of kidney disease. The dataset used for training and evaluation contained various clinical and demographic attributes of patients, including serum creatinine level, blood pressure, and age, among others. The proficiency analysis of the ML algorithms revealed consistent predictors across all models, with serum creatinine level, blood pressure, and age emerging as particularly effective in identifying individuals at risk of kidney disease. These findings align with established medical knowledge and emphasize the pivotal role of these attributes in early detection and prognosis. In conclusion, our study demonstrates the effectiveness of diverse machine learning algorithms in detecting and predicting kidney disease. The identification of influential predictors, such as serum creatinine level, blood pressure, and age, underscores their significance in early detection and prognosis. By leveraging machine learning techniques, we can enhance the accuracy and efficiency of kidney disease diagnosis and treatment, ultimately improving patient outcomes and healthcare system effectiveness. Full article

Patients with type 1 diabetes must continually decide how much insulin to inject before each meal to maintain blood glucose levels within a healthy range. Recent research has worked on a solution for this burden, showing the potential of reinforcement learning as an emerging approach for the task of controlling blood glucose levels. In this paper, we test and evaluate several deep Q-learning algorithms for automated and personalized blood glucose regulation in an in silico type 1 diabetes patient with the goal of estimating and delivering proper insulin doses. The proposed algorithms are model-free approaches with no prior information about the patient. We used the Hovorka model with meal variation and carbohydrate counting errors to simulate the patient included in this work. Our experiments compare different deep Q-learning extensions showing promising results controlling blood glucose levels, with some of the proposed algorithms outperforming standard baseline treatment. Full article

Cryptococcosis is an opportunistic infection in immunocompromised patients, involving mainly the lungs and central nervous system; however, the skin, eyes and genitourinary tract could also be involved as secondary sites of infection. Primary cutaneous cryptococcosis (PCC) is a distinct clinical entity that can occur in both immunocompetent and -compromised patients, usually trough skin injury. In immunocompetent patients, it is a very rare infection, presenting with non-specific clinical pictures and being challenging to diagnose. Herein, we present the case of an immunocompetent man with PCC due to Cryptococcus neoformans on his right forearm. PCC was diagnosed by a histological and cultural examination. Causes of concomitant immunosuppression were ruled out. A secondary cutaneous cryptococcosis was excluded with careful investigations. Therapy with oral fluconazole for three months was successfully performed, without evidence of recurrence in the following six months. Complete clinical recovery was achieved after three months of oral antifungal therapy, suggesting that longer courses of treatment could be avoided when faced with PCC in immunocompetent patients. Full article

The study by Chen et al. of a 56-year-old man diagnosed with acute hemorrhagic encephalomyelitis (AHEM) had a significant impact on us. The authors provided a comprehensive account of their diagnostic journey and emphasized the need to differentiate myelin oligodendrocyte glycoprotein antibody-associated disease (MOGAD) from AHEM. However, recent research suggests that AHEM may not be an isolated entity, but rather a phenotype within MOGAD. The patient’s clinical presentation included MRI brain lesions characteristic of MOGAD in addition to hemorrhagic abnormalities. These findings raise the possibility that AHEM in this case represents a MOGAD phenotype. In conclusion, it is important to recognize the potential association between AHEM and MOGAD, especially when distinct MOGAD brain MRI patterns are present, as in this case. Full article

Skin lesions are essential for the early detection and management of a number of dermatological disorders. Learning-based methods for skin lesion analysis have drawn much attention lately because of improvements in computer vision and machine learning techniques. A review of the most-recent methods for skin lesion classification, segmentation, and detection is presented in this survey paper. The significance of skin lesion analysis in healthcare and the difficulties of physical inspection are discussed in this survey paper. The review of state-of-the-art papers targeting skin lesion classification is then covered in depth with the goal of correctly identifying the type of skin lesion from dermoscopic, macroscopic, and other lesion image formats. The contribution and limitations of various techniques used in the selected study papers, including deep learning architectures and conventional machine learning methods, are examined. The survey then looks into study papers focused on skin lesion segmentation and detection techniques that aimed to identify the precise borders of skin lesions and classify them accordingly. These techniques make it easier to conduct subsequent analyses and allow for precise measurements and quantitative evaluations. The survey paper discusses well-known segmentation algorithms, including deep-learning-based, graph-based, and region-based ones. The difficulties, datasets, and evaluation metrics particular to skin lesion segmentation are also discussed. Throughout the survey, notable datasets, benchmark challenges, and evaluation metrics relevant to skin lesion analysis are highlighted, providing a comprehensive overview of the field. The paper concludes with a summary of the major trends, challenges, and potential future directions in skin lesion classification, segmentation, and detection, aiming to inspire further advancements in this critical domain of dermatological research. Full article

Infective endocarditis (IE) is a serious and diagnostically challenging condition. [¹⁸F]FDG PET/CT is valuable for evaluating suspected IE, but it is susceptible to motion-related artefacts. This study investigated the potential benefits of cardiac motion correction for [¹⁸F]FDG PET/CT. In this prospective study, patients underwent [¹⁸F]FDG PET/CT for suspected IE, combined with a conventional cardiac gating sequence, a data-driven cardiac and respiratory gating sequence (CardioFreeze^TM), or both. Scans were performed in adherence to EANM guidelines and assessors were blinded to patients’ clinical contexts. Final diagnosis of IE was established based on multidisciplinary consensus after a minimum of 4 months follow-up and surgical findings, whenever performed. Seven patients participated in the study, undergoing both an ungated [¹⁸F] FDG-PET/CT and a scan with either conventional cardiac gating, CardioFreeze^TM, or both. Cardiac motion correction improved the interpretability of [¹⁸F]FDG PET/CT in four out of five patients with valvular IE lesions, regardless of the method of motion correction used, which was statistically significant by Wilcoxon’s signed rank test: p = 0.046. In one patient the motion-corrected sequence confirmed the diagnosis of endocarditis, which had been missed on non-gated PET. The performance of the two gating sequences was comparable. In conclusion, in this exploratory study, cardiac motion correction of [¹⁸F]FDG PET/CT improved the interpretability of [¹⁸F]FDG PET/CT. This may improve the sensitivity of PET/CT for suspected IE. Further larger comparative studies are necessary to confirm the additive value of these cardiac motion correction methods. Full article

Background and Objectives: Peripheral artery disease (PAD) affects both genders, but the knowledge of clinical and therapeutic aspects particular to each gender has a prognostic value, modulating the risk of amputation and helping to reduce the risk of death or the occurrence of an acute vascular event secondary to optimal management. Materials and Methods: We conducted a retrospective, descriptive study that included 652 patients with PAD who were evaluated at “St. Spiridon” Hospital’s Cardiology Department and divided into two groups according to gender: women (100 cases) and men (552 cases). We evaluated demographics, anthropometric data, as well as clinical and paraclinical parameters in the two groups. Results: Men had a lower mean age (p < 0.001), higher mean BMI (p = 0.049) and were more frequent smokers. (p = 0.008). Hypercholesterolemia (p = 0.026), obesity (p = 0.009), concomitant cerebrovascular (p = 0.005) and chronic kidney disease (p = 0.046) were more common in women, while coronary artery disease (p = 0.033) was more common in men. The number of angiographic stenotic lesions (p = 0.037) is a statistically significant parameter in our study, with both genders predominantly associated with stenotic lesions. In addition, directly proportional relationships were found between smoking, uric acid, inflammatory markers, and the number of stenotic lesions and thromboses or the ankle–brachial index (ABI). In the subgroup of men, the number of stenotic and thrombosed lesions positively correlated with the ABI value (p < 0.001). The presence of more than three cardiovascular risk factors (p = 0.001) and serum triglyceride levels (p = 0.019) significantly correlated with the number of angiographically detected lesions. We applied several risk scores (PREVENT III, Finnvasc Score, or GermanVasc risk score) in our study group for prognostic purposes, without showing statistically significant differences between genders. Men, rest pain, gangrene, smoking status, the presence of more than three cardiovascular risk factors, or a serum HDL-cholesterol level below 40 mg/dL (p < 0.001 for all parameters) are independent predictors associated with amputation in our study group. Conclusions: In our study, we demonstrated that several clinical–paraclinical particularities guide the diagnosis, providing the clinician with prognostic and therapeutic tools to choose the optimal management with maximum benefits. Full article

Hepatitis B e antigen (HBeAg) is a marker of wild-type hepatitis B virus replication. In resource-limited countries where access to enzyme-linked immunosorbent assay (ELISA) remains a challenge, rapid diagnostic tests (RDT) constitute a good alternative. The HBeAg status is employed to evaluate eligibility for antiviral therapy and to prevent the transmission of hepatitis B from mother to child (PMTCT). The objective of this study was to assess the diagnostic performance of the SD-Bioline^®HBeAg RDT commonly used for detecting HBeAg in laboratories in Burkina Faso. The sample panel used was collected from HBsAg-positive patients received in the laboratory for the detection of HBeAg with the rapid test. The samples were retested for HBeAg using the VIDAS HBe/Anti-HBe enzyme-linked fluorescent assay (ELFA) (Gold standard). Then, the viral load (VL) of HBV DNA was determined using the GENERIC HBV CHARGE VIRLAE kit (GHBV-CV). The diagnostic performances of the SD-Bioline^®HBeAg and its agreement with the gold standard were calculated with their 95% confidence intervals. Overall, 340 sera obtained from HBsAg-positive patients were included in this evaluation Compared to the VIDAS HBe/Anti-HBe ELFA test, the sensitivity (Se) and specificity (Sp) of the SD-Bioline^®HBeAg test were 33.3% and 97.9%, respectively. The concordance between the two tests was 0.42. Depending on the viral load, the Se and Sp varied from 8.8% and 98.3% for a VL < 2000 IU/mL to 35.5% and 98.4% for a VL > 2,000,000 IU/mL. The results showed a low sensibility of the SD-Bioline^®HBeAg RDT test, indicating that its use is inappropriate for the clinical management of HBV-infected patients. They also highlight the urgent need to develop HBeAg rapid tests with better sensitivities. Full article

Background: In STEMIs, the evaluation of the relationship between biomarkers of myocardial injury and patients’ prognoses has not been completely explored. Increased levels of CK-MB in patients with a STEMI undergoing primary angioplasty are known to be associated with higher mortality rates, yet the correlation of these values with short-term evolution remains unknown. Material and Methods: The research encompassed a sample of 80 patients diagnosed with STEMIs, and its methodology entailed a retrospective analysis of the data collected during their hospital stays. The study population was then categorized into three distinct analysis groups based on the occurrence or absence of acute complications and fatalities. Results: The findings indicated that there is a notable correlation between rising levels of CK-MB upon admission and peak CK-MB levels with a reduction in left ventricular ejection fraction. Moreover, the CK-MB variation established a point of reference for anticipating complications at 388 U/L, and a cut-off value for predicting death at 354 U/L. Conclusion: CK-MB values are reliable indicators of the progress of patients with STEMIs. Furthermore, the difference between the peak and admission CK-MB levels demonstrates a high accuracy of predicting complications and has a significant predictive power to estimate mortality risk. Full article

Among the most serious types of cancer is skin cancer. Despite the risk of death, when caught early, the rate of survival is greater than 95%. This inspires researchers to explore methods that allow for the early detection of skin cancer that could save millions of lives. The ability to detect the early signs of skin cancer has become more urgent in light of the rising number of illnesses, the high death rate, and costly healthcare treatments. Given the gravity of these issues, experts have created a number of existing approaches for detecting skin cancer. Identifying skin cancer and whether it is benign or malignant involves detecting features of the lesions such as size, form, symmetry, color, etc. The aim of this study is to determine the most successful skin cancer detection methods by comparing the outcomes and effectiveness of the various applications that categorize benign and malignant forms of skin cancer. Descriptors such as the Local Binary Pattern (LBP), the Local Directional Number Pattern (LDN), the Pyramid of Histogram of Oriented Gradients (PHOG), the Local Directional Pattern (LDiP), and Monogenic Binary Coding (MBC) are used to extract the necessary features. Support vector machines (SVM) and XGBoost are used in the classification process. In addition, this study uses colored histogram-based features to classify the various characteristics obtained from the color images. In the experimental results, the applications implemented with the proposed color histogram-based features were observed to be more successful. Under the proposed method (the colored LDN feature obtained using the YCbCr color space with the XGBoost classifier), a 90% accuracy rate was achieved on Dataset 1, which was obtained from the Kaggle website. For the HAM10000 data set, an accuracy rate of 96.50% was achieved under a similar proposed method (the colored MBC feature obtained using the HSV color space with the XGBoost classifier). Full article

Since there are no morphological clues capable of making a pathologist suspect a possible mammary origin of a metastatic lesion without adequate clinical information, the histologic diagnosis of brain metastasis from BC is still based on the immunohistochemical expression of mammary gland markers such as GATA-3, ERs, PgRs and HER-2. The present retrospective study aimed to select purely morphological features capable of suggesting the mammary origin of a metastatic carcinoma in the brain. The following histological features were collected from a series of 30 cases of brain metastases from breast cancer: (i) a solid growth pattern; (ii) the presence of comedonecrosis; and (iii) glandular differentiation. Our results showed that most cases histologically exhibited a solid growth pattern with at least focal comedonecrosis, producing an overall morphology closely reminiscent of mammary high-grade ductal carcinoma in situ. Although the above-mentioned morphological parameters are not strictly specific to a mammary origin, they may have an important diagnostic utility for leading pathologists to suspect a possible breast primary tumor and to include GATA-3, ERs, PgRs and HER-2 in the immunohistochemical panel. Full article

Since its introduction in 2016, researchers have applied the idea of Federated Learning (FL) to several domains ranging from edge computing to banking. The technique’s inherent security benefits, privacy-preserving capabilities, ease of scalability, and ability to transcend data biases have motivated researchers to use this tool on healthcare datasets. While several reviews exist detailing FL and its applications, this review focuses solely on the different applications of FL to medical imaging datasets, grouping applications by diseases, modality, and/or part of the body. This Systematic Literature review was conducted by querying and consolidating results from ArXiv, IEEE Xplorer, and PubMed. Furthermore, we provide a detailed description of FL architecture, models, descriptions of the performance achieved by FL models, and how results compare with traditional Machine Learning (ML) models. Additionally, we discuss the security benefits, highlighting two primary forms of privacy-preserving techniques, including homomorphic encryption and differential privacy. Finally, we provide some background information and context regarding where the contributions lie. The background information is organized into the following categories: architecture/setup type, data-related topics, security, and learning types. While progress has been made within the field of FL and medical imaging, much room for improvement and understanding remains, with an emphasis on security and data issues remaining the primary concerns for researchers. Therefore, improvements are constantly pushing the field forward. Finally, we highlighted the challenges in deploying FL in medical imaging applications and provided recommendations for future directions. Full article

Neoadjuvant chemotherapy plus radical surgery could be a safe alternative to chemo-radiation in cervical cancer patients who are not willing to receive radiotherapy. The response to neoadjuvant chemotherapy is the main factor influencing the need for adjunctive treatments and survival. In the present paper we aim to develop a machine learning model based on cervix magnetic resonance imaging (MRI) images to stratify the single-subject risk of cervical cancer. We collected MRI images from 72 subjects. Among these subjects, 28 patients (38.9%) belonged to the “Not completely responding” class and 44 patients (61.1%) belonged to the ’Completely responding‘ class according to their response to treatment. This image set was used for the training and cross-validation of different machine learning models. A robust radiomic approach was applied, under the hypothesis that the radiomic features could be able to capture the disease heterogeneity among the two groups. Three models consisting of three ensembles of machine learning classifiers (random forests, support vector machines, and k-nearest neighbor classifiers) were developed for the binary classification task of interest (“Not completely responding” vs. “Completely responding”), based on supervised learning, using response to treatment as the reference standard. The best model showed an ROC-AUC (%) of 83 (majority vote), 82.3 (mean) [79.9–84.6], an accuracy (%) of 74, 74.1 [72.1–76.1], a sensitivity (%) of 71, 73.8 [68.7–78.9], and a specificity (%) of 75, 74.2 [71–77.5]. In conclusion, our preliminary data support the adoption of a radiomic-based approach to predict the response to neoadjuvant chemotherapy. Full article

Ampullary neoplastic lesions (ANLs) represent a rare cancer, accounting for about 0.6–0.8% of all gastrointestinal malignancies, and about 6–17% of periampullary tumors. They can be sporadic or occur in the setting of a hereditary predisposition syndrome, mainly familial adenomatous polyposis (FAP). Usually, noninvasive ANLs are asymptomatic and detected accidentally during esophagogastroduodenoscopy (EGD). When symptomatic, ANLs can manifest differently with jaundice, pain, pancreatitis, cholangitis, and melaena. Endoscopy with a side-viewing duodenoscopy, endoscopic ultrasound (EUS), and magnetic resonance cholangiopancreatography (MRCP) play a crucial role in the ANL evaluation, providing an accurate assessment of the size, location, and characteristics of the lesions, including the staging of the depth of tumor invasion into the surrounding tissues and the involvement of local lymph nodes. Endoscopic papillectomy (EP) has been recognized as an effective treatment for ANLs in selected patients, providing an alternative to traditional surgical methods. Originally, EP was recommended for benign lesions and patients unfit for surgery. However, advancements in endoscopic techniques have broadened its indications to comprise early ampullary carcinoma, giant laterally spreading lesions, and ANLs with intraductal extension. In this paper, we review the existing evidence on endoscopic diagnosis and treatment of ampullary neoplastic lesions. Full article

Advancements in artificial intelligence (AI) have rapidly transformed various sectors, and the field of echocardiography is no exception. AI-driven technologies hold immense potential to revolutionize echo labs’ diagnostic capabilities and improve patient care. This paper explores the importance for echo labs to embrace AI and stay ahead of the curve in harnessing its power. Our manuscript provides an overview of the growing impact of AI on medical imaging, specifically echocardiography. It highlights how AI-driven algorithms can enhance image quality, automate measurements, and accurately diagnose cardiovascular diseases. Additionally, we emphasize the importance of training echo lab professionals in AI implementation to optimize its integration into routine clinical practice. By embracing AI, echo labs can overcome challenges such as workload burden and diagnostic accuracy variability, improving efficiency and patient outcomes. This paper highlights the need for collaboration between echocardiography laboratory experts, AI researchers, and industry stakeholders to drive innovation and establish standardized protocols for implementing AI in echocardiography. In conclusion, this article emphasizes the importance of AI adoption in echocardiography labs, urging practitioners to proactively integrate AI technologies into their workflow and take advantage of their present opportunities. Embracing AI is not just a choice but an imperative for echo labs to maintain their leadership and excel in delivering state-of-the-art cardiac care in the era of advanced medical technologies. Full article

Purpose: By incorporating the cost of multiple tumor-marker tests, this work aims to comprehensively evaluate the financial burden of patients and the accuracy of machine learning models in diagnosing malignant pleural effusion (MPE) using tumor-marker combinations. Methods: Carcinoembryonic antigen (CEA), carbohydrate antigen (CA)19-9, CA125, and CA15-3 were collected from pleural effusion (PE) and peripheral blood (PB) of 319 patients with pleural effusion. A stacked ensemble (stacking) model based on five machine learning models was utilized to evaluate the diagnostic accuracy of tumor markers. We evaluated the discriminatory accuracy of various tumor-marker combinations using the area under the curve (AUC), sensitivity, and specificity. To evaluate the cost-effectiveness of different tumor-marker combinations, a comprehensive score (C-score) with a tuning parameter w was proposed. Results: In most scenarios, the stacking model outperformed the five individual machine learning models in terms of AUC. Among the eight tumor markers, the CEA in PE (PE.CEA) showed the best AUC of 0.902. Among all tumor-marker combinations, the PE.CA19-9 + PE.CA15-3 + PE.CEA + PB.CEA combination (C9 combination) achieved the highest AUC of 0.946. When w puts more weight on the cost, the highest C-score was achieved with the single PE.CEA marker. As w puts over 0.8 weight on AUC, the C-score favored diagnostic models with more expensive tumor-marker combinations. Specifically, when w was set to 0.99, the C9 combination achieved the best C-score. Conclusion: The stacking diagnostic model using PE.CEA is a relatively accurate and affordable choice in diagnosing MPE for patients without medical insurance or in a low economic level. The stacking model using the combination PE.CA19-9 + PE.CA15-3 + PE.CEA + PB.CEA is the most accurate diagnostic model and the best choice for patients without an economic burden. From a cost-effectiveness perspective, the stacking diagnostic model with PE.CA19-9 + PE.CA15-3 + PE.CEA combination is particularly recommended, as it gains the best trade-off between the low cost and high effectiveness. Full article

Medical diagnosis is the basis for treatment and management decisions in healthcare. Conventional methods for medical diagnosis commonly use established clinical criteria and fixed numerical thresholds. The limitations of such an approach may result in a failure to capture the intricate relations between diagnostic tests and the varying prevalence of diseases. To explore this further, we have developed a freely available specialized computational tool that employs Bayesian inference to calculate the posterior probability of disease diagnosis. This novel software comprises of three distinct modules, each designed to allow users to define and compare parametric and nonparametric distributions effectively. The tool is equipped to analyze datasets generated from two separate diagnostic tests, each performed on both diseased and nondiseased populations. We demonstrate the utility of this software by analyzing fasting plasma glucose, and glycated hemoglobin A1c data from the National Health and Nutrition Examination Survey. Our results are validated using the oral glucose tolerance test as a reference standard, and we explore both parametric and nonparametric distribution models for the Bayesian diagnosis of diabetes mellitus. Full article

Introduction: This study aimed to evaluate whether radiomic features extracted solely from the edema of soft tissue sarcomas (STS) could predict the occurrence of lung metastasis in comparison with features extracted solely from the tumoral mass. Materials and Methods: We retrospectively analyzed magnetic resonance imaging (MRI) scans of 32 STSs, including 14 with lung metastasis and 18 without. A segmentation of the tumor mass and edema was assessed for each MRI examination. A total of 107 radiomic features were extracted for each mass segmentation and 107 radiomic features for each edema segmentation. A two-step feature selection process was applied. Two predictive features for the development of lung metastasis were selected from the mass-related features, as well as two predictive features from the edema-related features. Two Random Forest models were created based on these selected features; 100 random subsampling runs were performed. Key performance metrics, including accuracy and area under the ROC curve (AUC), were calculated, and the resulting accuracies were compared. Results: The model based on mass-related features achieved a median accuracy of 0.83 and a median AUC of 0.88, while the model based on edema-related features achieved a median accuracy of 0.75 and a median AUC of 0.79. A statistical analysis comparing the accuracies of the two models revealed no significant difference. Conclusion: Both models showed promise in predicting the occurrence of lung metastasis in soft tissue sarcomas. These findings suggest that radiomic analysis of edema features can provide valuable insights into the prediction of lung metastasis in soft tissue sarcomas. Full article

Although wireless capsule endoscopy (WCE) detects small bowel diseases effectively, it has some limitations. For example, the reading process can be time consuming due to the numerous images generated per case and the lesion detection accuracy may rely on the operators’ skills and experiences. Hence, many researchers have recently developed deep-learning-based methods to address these limitations. However, they tend to select only a portion of the images from a given WCE video and analyze each image individually. In this study, we note that more information can be extracted from the unused frames and the temporal relations of sequential frames. Specifically, to increase the accuracy of lesion detection without depending on experts’ frame selection skills, we suggest using whole video frames as the input to the deep learning system. Thus, we propose a new Transformer-architecture-based neural encoder that takes the entire video as the input, exploiting the power of the Transformer architecture to extract long-term global correlation within and between the input frames. Subsequently, we can capture the temporal context of the input frames and the attentional features within a frame. Tests on benchmark datasets of four WCE videos showed 95.1% sensitivity and 83.4% specificity. These results may significantly advance automated lesion detection techniques for WCE images. Full article

Sarcoidosis has been a well-recognised risk factor for pulmonary hypertension (PH) for a long time, but still, the knowledge about this concatenation is incomplete. Sarcoidosis-associated PH (SAPH) is an uncommon but serious complication associated with increased morbidity and mortality among sarcoidosis patients. The real epidemiology of SAPH remains unknown, and its pathomechanisms are not fully explained. Sarcoidosis is a heterogeneous and dynamic condition, and SAPH pathogenesis is believed to be multifactorial. The main roles in SAPH development play: parenchymal lung disease with the destruction of pulmonary vessels, the extrinsic compression of pulmonary vessels by conglomerate masses, lymphadenopathy or fibrosing mediastinitis, pulmonary vasculopathy, LV dysfunction, and portal hypertension. Recently, it has been recommended to individually tailor SAPH management according to the predominant pathomechanism, i.e., SAPH phenotype. Unfortunately, SAPH phenotyping is not a straightforward process. First, there are gaps in our understanding of undergoing processes. Second, the assessment of such a pivotal element as pulmonary vasculature on a microscopic level is non-feasible in SAPH patients antemortem. Finally, SAPH is a dynamic condition, multiple phenotypes usually coexist, and patients can switch between phenotypes during the course of sarcoidosis. In this article, we summarise the basic knowledge of SAPH, describe SAPH phenotypes, and highlight some practical problems related to SAPH phenotyping. Full article

Accurate and early detection of malignant pelvic mass is important for a suitable referral, triage, and for further care for the women diagnosed with a pelvic mass. Several deep learning (DL) methods have been proposed to detect pelvic masses but other methods cannot provide sufficient accuracy and increase the computational time while classifying the pelvic mass. To overcome these issues, in this manuscript, the evolutionary gravitational neocognitron neural network optimized with nomadic people optimizer for gynecological abdominal pelvic masses classification is proposed for classifying the pelvic masses (EGNNN-NPOA-PM-UI). The real time ultrasound pelvic mass images are augmented using random transformation. Then the augmented images are given to the 3D Tsallis entropy-based multilevel thresholding technique for extraction of the ROI region and its features are further extracted with the help of fast discrete curvelet transform with the wrapping (FDCT-WRP) method. Therefore, in this work, EGNNN optimized with nomadic people optimizer (NPOA) was utilized for classifying the gynecological abdominal pelvic masses. It was executed in PYTHON and the efficiency of the proposed method analyzed under several performance metrics. The proposed EGNNN-NPOA-PM-UI methods attained 99.8%. Ultrasound image analysis using the proposed EGNNN-NPOA-PM-UI methods can accurately predict pelvic masses analyzed with the existing methods. Full article

Background: Many patients with Type B aortic dissection (TBAD) may not show noticeable symptoms until they become intervention and help prevent critically ill, which can result in fatal outcomes. Thus, it is crucial to screen people at high risk of TBAD and initiate the necessary preventive and therapeutic measures before irreversible harm occurs. By developing a prediction model for aortic arch morphology, it is possible to accurately identify those at high risk and take prompt action to prevent the adverse consequences of TBAD. This approach can facilitate timely the development of serious illnesses. Method: The predictive model was established in a primary population consisting of 173 patients diagnosed with acute Stanford TBAD, with data collected from January 2017 and December 2018, as well as 534 patients with healthy aortas, with data collected from April 2018 and December 2018. Explicitly, the data were randomly separated into the derivation set and validation set in a 7:3 ratio. Geometric and anatomical features were extracted from a three-dimensional multiplanar reconstruction of the aortic arch. The LASSO regression model was utilized to minimize the data dimension and choose relevant features. Multivariable logistic regression analysis and backward stepwise selection were employed for predictive model generation, combining demographic and clinical features as well as geometric and anatomical features. The predictive model’s performance was evaluated by examining its calibration, discrimination, and clinical benefit. Finally, we also conducted internal verification. Results: After applying LASSO logistic regression and backward stepwise selection, 12 features were entered into the prediction model. Age, aortic arch angle, total thoracic aorta distance, ascending aorta tortuosity, aortic arch tortuosity, distal descending aorta tortuosity, and type III arch were protective factors, while male sex, hypertension, aortic arch height, and aortic arch distance were risk factors. The model exhibited satisfactory discrimination (AUC, 0.917 [95% CI, 0.890–0.945]) and good calibration in the derivation set. Applying the predictive model to the validation set also provided satisfactory discrimination (AUC, 0.909 [95% CI, 0.864–0.953]) and good calibration. The TBAD nomogram for clinical use was established. Conclusions: This study demonstrates that a multivariable logistic regression model can be used to predict TBAD patients. Full article