arXiv:2511.14087v1 Announce Type: new 
Abstract: Medical image segmentation underpins computer-aided diagnosis and therapy by supporting clinical diagnosis, preoperative planning, and disease monitoring. While U-Net style convolutional neural networks perform well due to their encoder-decoder structures with skip connections, they struggle to capture long-range dependencies. Transformer-based variants address global context but often require heavy computation and large training datasets. This paper proposes GCA-ResUNet, an efficient segmentation network that integrates Grouped Coordinate Attention (GCA) into ResNet-50 residual blocks. GCA uses grouped coordinate modeling to jointly encode global dependencies across channels and spatial locations, strengthening feature representation and boundary delineation while adding minimal parameter and FLOP overhead compared with self-attention. On the Synapse dataset, GCA-ResUNet achieves a Dice score of 86.11%, and on the ACDC dataset, it reaches 92.64%, surpassing several state-of-the-art baselines while maintaining fast inference and favorable computational efficiency. These results indicate that GCA offers a practical way to enhance convolutional architectures with global modeling capability, enabling high-accuracy and resource-efficient medical image segmentation.

تقدم الورقة GCA-ResUNet، وهو شبكة جديدة لت segmenting الصور الطبية. يدمج الانتباه المنسق المجموع (GCA) في الكتل المتبقية من ResNet-50، مما يعزز قدرة النموذج على التقاط الاعتماديات طويلة المدى مع الحفاظ على الكفاءة. يحسن هذا النهج تمثيل الميزات وتحديد الحدود مع الحد الأدنى من التكلفة الحسابية الإضافية. حقق GCA-ResUNet درجة Dice تبلغ 86.11% على مجموعة بيانات Synapse، مما يدل على فعاليته في مهام تقسيم الصور الطبية.

El artículo presenta GCA-ResUNet, una nueva red de segmentación de imágenes diseñada para imágenes médicas. Integra la Atención por Coordenadas Agrupadas (GCA) en los bloques residuales de ResNet-50, mejorando la capacidad del modelo para capturar dependencias a largo plazo mientras mantiene la eficiencia. Este enfoque mejora la representación de características y la delimitación de bordes con un costo computacional adicional mínimo. GCA-ResUNet logró un puntaje de Dice del 86.11% en el conjunto de datos Synapse, demostrando su efectividad en tareas de segmentación de imágenes médicas.

Cet article présente GCA-ResUNet, un nouveau réseau de segmentation d'images conçu pour les images médicales. Il intègre l'attention par coordonnées groupées (GCA) dans les blocs résiduels de ResNet-50, améliorant la capacité du modèle à capturer les dépendances à long terme tout en maintenant l'efficacité. Cette approche améliore la représentation des caractéristiques et la délimitation des contours avec un coût computationnel supplémentaire minimal. GCA-ResUNet a atteint un score de Dice de 86,11 % sur le jeu de données Synapse, démontrant son efficacité dans les tâches de segmentation d'ima…

The paper presents GCA-ResUNet, a novel image segmentation network designed for medical images. It integrates Grouped Coordinate Attention (GCA) into ResNet-50 residual blocks, enhancing the model's ability to capture long-range dependencies while maintaining efficiency. This approach improves feature representation and boundary delineation with minimal additional computational cost. The GCA-ResUNet achieved a Dice score of 86.11% on the Synapse dataset, demonstrating its effectiveness in medical image segmentation tasks.

GCA-ResUNet:Image segmentation in medical images using grouped coordinate attention

arXiv:2512.04967v1 Announce Type: new 
Abstract: Automated retinal disease diagnosis is vital given the rising prevalence of conditions such as diabetic retinopathy and macular degeneration. Conventional deep learning approaches require large annotated datasets, which are costly and often imbalanced across disease categories, limiting their reliability in practice. Few-shot learning (FSL) addresses this challenge by enabling models to generalize from only a few labeled samples per class. In this study,we propose a balanced few-shot episodic learning framework tailored to the Retinal Fundus Multi-Disease Image Dataset (RFMiD). Focusing on the ten most represented classes, which still show substantial imbalance between majority diseases (e.g., Diabetic Retinopathy, Macular Hole) and minority ones (e.g., Optic Disc Edema, Branch Retinal Vein Occlusion), our method integrates three key components: (i) balanced episodic sampling, ensuring equal participation of all classes in each 5-way 5-shot episode; (ii) targeted augmentation, including Contrast Limited Adaptive Histogram Equalization (CLAHE) and color/geometry transformations, to improve minority-class di- versity; and (iii) a ResNet-50 encoder pretrained on ImageNet, selected for its superior ability to capture fine-grained retinal features. Prototypes are computed in the embedding space and classification is performed with cosine similarity for improved stability. Trained on 100 episodes and evaluated on 1,000 test episodes, our framework achieves substantial accuracy gains and reduces bias toward majority classes, with notable improvements for underrepresented diseases. These results demonstrate that dataset-aware few-shot pipelines, combined with balanced sampling and CLAHE-enhanced preprocessing, can deliver more robust and clinically fair retinal disease diagnosis under data-constrained conditions.

تقدم دراسة جديدة إطارًا للتعلم الإepisodic المتوازن القليل من اللقطات يهدف إلى تحسين دقة تشخيص الأمراض الشبكية الآلي، خاصةً لحالات مثل اعتلال الشبكية السكري والتنكس البقعي. تستخدم هذه الطريقة مجموعة بيانات صور قاع العين متعددة الأمراض (RFMiD) وتعالج تحدي البيانات غير المتوازنة في الأساليب التقليدية للتعلم العميق.

Un nuevo estudio presenta un marco de aprendizaje episódico equilibrado de pocos disparos que busca mejorar la precisión del diagnóstico automatizado de enfermedades retinianas, especialmente para condiciones como la retinopatía diabética y la degeneración macular. Este método utiliza el Conjunto de Datos de Imágenes de Fondo Retiniano Multi-Enfermedad (RFMiD) y aborda el desafío de los conjuntos de datos desequilibrados en los enfoques convencionales de aprendizaje profundo.

Une nouvelle étude présente un cadre d'apprentissage épisodique équilibré par apprentissage peu profond, visant à améliorer la précision du diagnostic automatisé des maladies rétiniennes, en particulier pour des conditions telles que la rétinopathie diabétique et la dégénérescence maculaire. Cette méthode utilise le jeu de données d'images rétiniennes multi-maladies (RFMiD) et aborde le défi des ensembles de données déséquilibrés dans les approches d'apprentissage profond conventionnelles.

A new study introduces a balanced few-shot episodic learning framework aimed at improving the accuracy of automated retinal disease diagnosis, particularly for conditions like diabetic retinopathy and macular degeneration. This method utilizes the Retinal Fundus Multi-Disease Image Dataset (RFMiD) and addresses the challenge of imbalanced datasets in conventional deep learning approaches.

Balanced Few-Shot Episodic Learning for Accurate Retinal Disease Diagnosis

arXiv:2512.00626v2 Announce Type: replace 
Abstract: Accurate and timely diagnosis of multi-class skin lesions is hampered by subjective methods, inherent data imbalance in datasets like HAM10000, and the "black box" nature of Deep Learning (DL) models. This study proposes a trustworthy and highly accurate Computer-Aided Diagnosis (CAD) system to overcome these limitations. The approach utilizes Deep Convolutional Generative Adversarial Networks (DCGANs) for per class data augmentation to resolve the critical class imbalance problem. A fine-tuned ResNet-50 classifier is then trained on the augmented dataset to classify seven skin disease categories. Crucially, LIME and SHAP Explainable AI (XAI) techniques are integrated to provide transparency by confirming that predictions are based on clinically relevant features like irregular morphology. The system achieved a high overall Accuracy of 92.50 % and a Macro-AUC of 98.82 %, successfully outperforming various prior benchmarked architectures. This work successfully validates a verifiable framework that combines high performance with the essential clinical interpretability required for safe diagnostic deployment. Future research should prioritize enhancing discrimination for critical categories, such as Melanoma NOS (F1-Score is 0.8602).

قدمت دراسة جديدة نظام تشخيص مدعوم بالكمبيوتر (CAD) يستخدم الشبكات التنافسية التوليدية العميقة (DCGAN) لزيادة البيانات اللازمة لتدريب مصنف ResNet-50 المعدل، محققًا دقة مثيرة للإعجاب تبلغ 92.50% في تصنيف سبع فئات من الأمراض الجلدية. كما أن دمج تقنيات الذكاء الاصطناعي القابلة للتفسير، LIME وSHAP، يعزز من شفافية التنبؤات بناءً على ميزات ذات صلة سريريًا.

Un nuevo estudio ha presentado un sistema de Diagnóstico Asistido por Computadora (CAD) que utiliza Redes Generativas Antagónicas Convolucionales Profundas (DCGAN) para aumentar los datos necesarios para entrenar un clasificador ResNet-50 ajustado, logrando una impresionante precisión del 92.50% en la clasificación de siete categorías de enfermedades de la piel. La integración de técnicas de IA Explicable, LIME y SHAP, mejora la transparencia de las predicciones basadas en características clínicamente relevantes.

Une nouvelle étude a présenté un système de diagnostic assisté par ordinateur (CAD) qui utilise des réseaux antagonistes génératifs convolutifs profonds (DCGAN) pour augmenter les données nécessaires à l'entraînement d'un classificateur ResNet-50 affiné, atteignant une précision impressionnante de 92,50 % dans la classification de sept catégories de maladies cutanées. L'intégration de techniques d'IA explicable, LIME et SHAP, améliore la transparence des prédictions basées sur des caractéristiques cliniquement pertinentes.

A new study has introduced a Computer-Aided Diagnosis (CAD) system that utilizes Deep Convolutional Generative Adversarial Networks (DCGANs) to augment data for training a fine-tuned ResNet-50 classifier, achieving an impressive accuracy of 92.50% in classifying seven skin disease categories. The integration of Explainable AI techniques, LIME and SHAP, enhances the transparency of predictions based on clinically relevant features.

XAI-Driven Skin Disease Classification: Leveraging GANs to Augment ResNet-50 Performance

arXiv:2512.02364v1 Announce Type: new 
Abstract: This study explores the application of machine learning models, specifically a pretrained ResNet-50 model and a general SqueezeNet model, in diagnosing tuberculosis (TB) using chest X-ray images. TB, a persistent infectious disease affecting humanity for millennia, poses challenges in diagnosis, especially in resource-limited settings. Traditional methods, such as sputum smear microscopy and culture, are inefficient, prompting the exploration of advanced technologies like deep learning and computer vision. The study utilized a dataset from Kaggle, consisting of 4,200 chest X-rays, to develop and compare the performance of the two machine learning models. Preprocessing involved data splitting, augmentation, and resizing to enhance training efficiency. Evaluation metrics, including accuracy, precision, recall, and confusion matrix, were employed to assess model performance. Results showcase that the SqueezeNet achieved a loss of 32%, accuracy of 89%, precision of 98%, recall of 80%, and an F1 score of 87%. In contrast, the ResNet-50 model exhibited a loss of 54%, accuracy of 73%, precision of 88%, recall of 52%, and an F1 score of 65%. This study emphasizes the potential of machine learning in TB detection and possible implications for early identification and treatment initiation. The possibility of integrating such models into mobile devices expands their utility in areas lacking TB detection resources. However, despite promising results, the need for continued development of faster, smaller, and more accurate TB detection models remains crucial in contributing to the global efforts in combating TB.

أظهرت دراسة حديثة استخدام نماذج التعلم الآلي، وخاصة ResNet-50 وSqueezeNet، في تشخيص مرض السل (TB) من خلال صور الأشعة السينية للصدر. استخدمت الدراسة مجموعة بيانات تتكون من 4200 صورة أشعة سينية من Kaggle، مما يبرز القيود المفروضة على طرق التشخيص التقليدية في البيئات ذات الموارد المحدودة. أظهرت النتائج أن SqueezeNet حققت أداءً ملحوظًا مع خسارة بنسبة 32% ومقاييس دقة تؤكد على إمكانية استخدام التعلم العميق في الكشف عن مرض السل.

Un estudio reciente ha investigado el uso de modelos de aprendizaje automático, específicamente ResNet-50 y SqueezeNet, para diagnosticar la tuberculosis (TB) a través de imágenes de radiografías de tórax. La investigación utilizó un conjunto de datos de 4,200 radiografías de Kaggle, destacando las limitaciones de los métodos de diagnóstico tradicionales en entornos con recursos limitados. Los resultados indicaron que SqueezeNet logró un rendimiento notable con una pérdida del 32% y métricas de precisión que subrayan el potencial del aprendizaje profundo en la detección de la TB.

Une étude récente a examiné l'utilisation de modèles d'apprentissage automatique, notamment ResNet-50 et SqueezeNet, pour diagnostiquer la tuberculose (TB) à partir d'images de radiographies thoraciques. La recherche a utilisé un ensemble de données de 4 200 radiographies provenant de Kaggle, mettant en évidence les limites des méthodes de diagnostic traditionnelles dans les contextes à ressources limitées. Les résultats ont indiqué que SqueezeNet a atteint une performance notable avec une perte de 32 % et des métriques de précision qui soulignent le potentiel de l'apprentissage profond dans l…

A recent study has investigated the use of machine learning models, specifically ResNet-50 and SqueezeNet, for diagnosing tuberculosis (TB) through chest X-ray images. The research utilized a dataset of 4,200 X-rays from Kaggle, highlighting the limitations of traditional diagnostic methods in resource-limited settings. Results indicated that SqueezeNet achieved a notable performance with a loss of 32% and accuracy metrics that underscore the potential of deep learning in TB detection.

Tackling Tuberculosis: A Comparative Dive into Machine Learning for Tuberculosis Detection

arXiv:2512.02088v1 Announce Type: cross 
Abstract: This study compares baseline (J0) and 24-hour (J1) diffusion magnetic resonance imaging (MRI) for predicting three-month functional outcomes after acute ischemic stroke (AIS). Seventy-four AIS patients with paired apparent diffusion coefficient (ADC) scans and clinical data were analyzed. Three-dimensional ResNet-50 embeddings were fused with structured clinical variables, reduced via principal component analysis (<=12 components), and classified using linear support vector machines with eight-fold stratified group cross-validation. J1 multimodal models achieved the highest predictive performance (AUC = 0.923 +/- 0.085), outperforming J0-based configurations (AUC <= 0.86). Incorporating lesion-volume features further improved model stability and interpretability. These findings demonstrate that early post-treatment diffusion MRI provides superior prognostic value to pre-treatment imaging and that combining MRI, clinical, and lesion-volume features produces a robust and interpretable framework for predicting three-month functional outcomes in AIS patients.

أجرت دراسة مقارنة بين تصوير الرنين المغناطيسي بالانتشار الأساسي وتصوير الرنين المغناطيسي بعد 24 ساعة للتنبؤ بالنتائج الوظيفية بعد ثلاثة أشهر من السكتة الدماغية الإقفارية الحادة (AIS) لدى 74 مريضًا. استخدمت الدراسة تمثيلات ResNet-50 ثلاثية الأبعاد مع بيانات سريرية، محققة أداءً تنبؤيًا قدره AUC = 0.923 لتصوير الرنين المغناطيسي بعد 24 ساعة، متفوقة على AUC البالغ 0.86 للتصوير الأساسي. كما حسنت إضافة ميزات حجم الآفة من استقرار النموذج وقابليته للتفسير.

Un estudio ha comparado la resonancia magnética de difusión de referencia y la de 24 horas para predecir los resultados a tres meses después de un accidente cerebrovascular isquémico agudo (AIS) en 74 pacientes. La investigación utilizó embeddings ResNet-50 en tres dimensiones combinados con datos clínicos, logrando un rendimiento predictivo de AUC = 0.923 para la resonancia magnética de 24 horas, superando el AUC de 0.86 de la referencia. La incorporación de características de volumen de lesión mejoró la estabilidad e interpretabilidad del modelo.

Une étude a comparé l'IRM de diffusion de base et celle de 24 heures pour prédire les résultats à trois mois après un AVC ischémique aigu (AIS) chez 74 patients. La recherche a utilisé des embeddings ResNet-50 en trois dimensions combinés à des données cliniques, atteignant une performance prédictive de AUC = 0,923 pour l'IRM de 24 heures, surpassant l'AUC de 0,86 de la base. L'incorporation de caractéristiques de volume de lésion a amélioré la stabilité et l'interprétabilité du modèle.

A study has compared baseline and 24-hour diffusion MRI to predict three-month outcomes after acute ischemic stroke (AIS) in 74 patients. The research utilized three-dimensional ResNet-50 embeddings combined with clinical data, achieving a predictive performance of AUC = 0.923 for the 24-hour MRI, surpassing the baseline's AUC of 0.86. Incorporating lesion-volume features enhanced model stability and interpretability.

GCA-ResUNet:Image segmentation in medical images using grouped coordinate attention

Was this article worth reading? Share it

Airparser

Attentive AI

Augmeta