arXiv:2511.17043v1 Announce Type: cross 
Abstract: Chest radiography remains one of the most widely used imaging modalities for thoracic diagnosis, yet increasing imaging volumes and radiologist workload continue to challenge timely interpretation. In this work, we investigate the use of MedImageInsight, a medical imaging foundational model, for automated binary classification of chest X-rays into Normal and Abnormal categories. Two approaches were evaluated: (1) fine-tuning MedImageInsight for end-to-end classification, and (2) employing the model as a feature extractor for a transfer learning pipeline using traditional machine learning classifiers. Experiments were conducted using a combination of the ChestX-ray14 dataset and real-world clinical data sourced from partner hospitals. The fine-tuned classifier achieved the highest performance, with an ROC-AUC of 0.888 and superior calibration compared to the transfer learning models, demonstrating performance comparable to established architectures such as CheXNet. These results highlight the effectiveness of foundational medical imaging models in reducing task-specific training requirements while maintaining diagnostic reliability. The system is designed for integration into web-based and hospital PACS workflows to support triage and reduce radiologist burden. Future work will extend the model to multi-label pathology classification to provide preliminary diagnostic interpretation in clinical environments.

تم استخدام MedImageInsight للتصنيف الثنائي الآلي للأشعة السينية للصدر، حيث يتم تصنيفها على أنها طبيعية أو غير طبيعية. يتناول هذا النهج التحديات التي تطرحها زيادة أحجام التصوير وعبء العمل على أطباء الأشعة، مع إجراء تجارب باستخدام مجموعة بيانات ChestX-ray14 وبيانات سريرية حقيقية من مستشفيات شريكة.

MedImageInsight se ha utilizado para la clasificación binaria automatizada de radiografías de tórax, categorizándolas como Normales o Anormales. Este enfoque aborda los desafíos que presentan los volúmenes crecientes de imágenes y la carga de trabajo de los radiólogos, con experimentos realizados utilizando el conjunto de datos ChestX-ray14 y datos clínicos del mundo real de hospitales asociados.

MedImageInsight a été utilisé pour la classification binaire automatisée des radiographies thoraciques, les catégorisant comme Normales ou Anormales. Cette approche répond aux défis posés par l'augmentation des volumes d'imagerie et la charge de travail des radiologues, avec des expériences menées à l'aide du jeu de données ChestX-ray14 et de données cliniques réelles provenant d'hôpitaux partenaires.

MedImageInsight has been utilized for the automated binary classification of chest X-rays, categorizing them as Normal or Abnormal. This approach addresses the challenges posed by increasing imaging volumes and radiologist workloads, with experiments conducted using the ChestX-ray14 dataset and real-world clinical data from partner hospitals.

MedImageInsight for Thoracic Cavity Health Classification from Chest X-rays

arXiv:2511.17803v1 Announce Type: new 
Abstract: Radiology plays an integral role in modern medicine, yet rising imaging volumes have far outpaced workforce growth. Foundation models offer a path toward assisting with the full spectrum of radiology tasks, but existing medical models remain limited: they process volumetric CT and MRI as low-fidelity 2D slices, discard critical grayscale contrast information, and lack evaluation frameworks that reflect real clinical practice. We introduce Pillar-0, a radiology foundation model pretrained on 42,990 abdomen-pelvis CTs, 86,411 chest CTs, 14,348 head CTs, and 11,543 breast MRIs from a large academic center, together with RATE, a scalable framework that extracts structured labels for 366 radiologic findings with near-perfect accuracy using LLMs. Across internal test sets of 14,230 abdomen-pelvis CTs, 10,646 chest CTs, 4,906 head CTs, and 1,585 breast MRIs, Pillar-0 establishes a new performance frontier, achieving mean AUROCs of 86.4, 88.0, 90.1, and 82.9, outperforming MedGemma (Google), MedImageInsight (Microsoft), Lingshu (Alibaba), and Merlin (Stanford) by 7.8-15.8 AUROC points and ranking best in 87.2\% (319/366) tasks. Pillar-0 similarly outperforms all baselines in an external validation on the Stanford Abdominal CT dataset, including Merlin (82.2 vs 80.6 AUROC). Pillar-0 extends to tasks beyond its pretraining, such as long-horizon lung cancer risk prediction, where it improves upon the state-of-the-art Sybil by 3.0 C-index points on NLST, and generalizes with gains of 5.9 (MGH) and 1.9 (CGMH). In brain hemorrhage detection, Pillar-0 obtained a >95 AUROC when using only 1/20th of the data of the next most sample efficient baseline. Pillar-0 and RATE together provide an open, clinically rigorous foundation for building high-performance radiology systems, enabling applications that were previously infeasible due to computational, data, and evaluation constraints.

تم تقديم Pillar-0 كنموذج جديد في مجال الأشعة، تم تدريبه مسبقًا على مجموعة بيانات كبيرة من الأشعة المقطعية والتصوير بالرنين المغناطيسي، بهدف تحسين الكفاءة والدقة في التقييمات الإشعاعية. يعالج هذا النموذج القيود الموجودة في النماذج الطبية الحالية، التي غالبًا ما تعالج بيانات التصوير بطريقة تتجاهل المعلومات الحيوية وتفتقر إلى أطر تقييم قوية.

Pillar-0 se ha presentado como un nuevo modelo de fundación en radiología, preentrenado con un conjunto de datos sustancial de escaneos de CT y MRI, con el objetivo de mejorar la eficiencia y precisión de las evaluaciones radiológicas. Este modelo aborda las limitaciones de los modelos médicos existentes, que a menudo procesan los datos de imagen de una manera que descarta información crítica y carecen de marcos de evaluación robustos.

Pillar-0 a été présenté comme un nouveau modèle de fondation en radiologie, préentraîné sur un ensemble de données substantiel de scans CT et IRM, visant à améliorer l'efficacité et la précision des évaluations radiologiques. Ce modèle répond aux limites des modèles médicaux existants, qui traitent souvent les données d'imagerie de manière à écarter des informations critiques et manquent de cadres d'évaluation robustes.

Pillar-0 has been introduced as a new radiology foundation model, pretrained on a substantial dataset of CT and MRI scans, aiming to enhance the efficiency and accuracy of radiological assessments. This model addresses the limitations of existing medical models, which often process imaging data in a way that discards critical information and lacks robust evaluation frameworks.

Pillar-0: A New Frontier for Radiology Foundation Models

arXiv:2511.18839v1 Announce Type: new 
Abstract: The utility of deep learning models, such as CheXNet, in high stakes clinical settings is fundamentally constrained by their purely deterministic nature, failing to provide reliable measures of predictive confidence. This project addresses this critical gap by integrating robust Uncertainty Quantification (UQ) into a high performance diagnostic platform for 14 common thoracic diseases on the NIH ChestX-ray14 dataset. Initial architectural development failed to stabilize performance and calibration using Monte Carlo Dropout (MCD), yielding an unacceptable Expected Calibration Error (ECE) of 0.7588. This technical failure necessitated a rigorous architectural pivot to a high diversity, 9-member Deep Ensemble (DE). This resulting DE successfully stabilized performance and delivered superior reliability, achieving a State-of-the-Art (SOTA) average Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.8559 and an average F1 Score of 0.3857. Crucially, the DE demonstrated superior calibration (Mean ECE of 0.0728 and Negative Log-Likelihood (NLL) of 0.1916) and enabled the reliable decomposition of total uncertainty into its Aleatoric (irreducible data noise) and Epistemic (reducible model knowledge) components, with a mean Epistemic Uncertainty (EU) of 0.0240. These results establish the Deep Ensemble as a trustworthy and explainable platform, transforming the model from a probabilistic tool into a reliable clinical decision support system.

قدمت دراسة حديثة منصة تشخيص عالية الأداء تستخدم قياس عدم اليقين (UQ) المعتمد على مجموعة عميقة لتعزيز تشخيص 14 مرضًا صدريًا شائعًا باستخدام مجموعة بيانات NIH ChestX-ray14. تتناول هذه الطريقة القيود المفروضة على نماذج التعلم العميق الحالية، مثل CheXNet، التي تفتقر إلى مقاييس موثوقة لثقة التنبؤ. حقق النموذج الجديد متوسط منطقة تحت منحنى التشغيل (AUROC) يبلغ 0.8559.

Un estudio reciente ha presentado una plataforma de diagnóstico de alto rendimiento que utiliza la cuantificación de incertidumbre (UQ) basada en un conjunto profundo para mejorar el diagnóstico de 14 enfermedades torácicas comunes utilizando el conjunto de datos NIH ChestX-ray14. Este enfoque aborda las limitaciones de los modelos de aprendizaje profundo existentes, como CheXNet, que carecen de medidas fiables de confianza predictiva. El nuevo modelo logró un área bajo la curva característica (AUROC) promedio de 0.8559.

Une étude récente a introduit une plateforme de diagnostic haute performance utilisant la quantification d'incertitude (UQ) basée sur un ensemble profond pour améliorer le diagnostic de 14 maladies thoraciques courantes à l'aide du jeu de données NIH ChestX-ray14. Cette approche répond aux limites des modèles d'apprentissage profond existants, tels que CheXNet, qui manquent de mesures fiables de confiance prédictive. Le nouveau modèle a atteint une courbe caractéristique moyenne de l'aire sous la courbe (AUROC) de 0,8559.

A recent study has introduced a high-performance diagnostic platform utilizing Deep Ensemble-based Uncertainty Quantification (UQ) to enhance the diagnosis of 14 common thoracic diseases using the NIH ChestX-ray14 dataset. This approach addresses the limitations of existing deep learning models, such as CheXNet, which lack reliable measures of predictive confidence. The new model achieved a State-of-the-Art average Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.8559.

MedImageInsight for Thoracic Cavity Health Classification from Chest X-rays

Was this article worth reading? Share it