Causal Explainable Foundation Model for Medical Image Intelligence with Disease Progression Reasoning and Clinical Decision Support
Year: 2026
Type: article
Abstract: Medical image intelligence has seen tremendous advancements due to the development of foundation models; however, existing techniques make use of correlation-based machine learning with limited interpretability, which makes the approach less reliable in disease progression analysis and clinical decision support. In this paper, a new framework known as Causal Explainable Foundation Model (CEFM) is presented that incorporates foundation models, causal reasoning, explainable artificial intelligence (XAI), disease progression modeling, and clinical decision support in a single platform. The proposed design consists of a multimodal foundation encoder that can learn generalized representations from the medical images, clinical notes, and reports. A causal reasoning component develops disease-specific cause-and-effect relationships between imaging biomarkers, whereas the explainability component provides visual attributions, causal pathway evidence, and confidence scores. Also, the reasoning engine of disease progression reasons about the longitudinal progression of pathology to forecast disease stages and severity in the future, which helps in making personalized decisions for treatment. The proposed model was tested using a total of 48,620 multimodal medical images annotated using publicly available datasets involving diseases of thoracic, neurological, and oncological areas. Experimental results showed overall accuracy, precision, recall, F1 score, area under the curve (AUC), specificity, and Matthews Correlation Coefficient (MCC) values of 98.14%, 97.82%, 97.56%, 97.69%, 0.992, 98.31%, and 0.961, respectively. Also, the inclusion of causal reasoning increased disease progression prediction accuracy by 6.8% and explanation consistency by 14.5%. These results show that the use of foundational knowledge representations in conjunction with causal reasoning and temporal reasoning can significantly enhance diagnostic accuracy, interpretation, and temporal progression awareness. The presented CEFM framework can be considered an efficient, interpretable, and clinically relevant approach to intelligent medical image analysis and has great potential for AI-driven healthcare solutions.
Institution: Sri Ramakrishna Engineering College
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W7201865667