Imaging of fibrotic and calcified aortic valve changes and analysis using multi-task models for the prediction of morphological and clinical long-term changes

• Applicant: Professor Dr. Marc Dewey
• Subject Area: Radiology ; Medical Physics, Biomedical Technology
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 567210027

Background and Objective: Aortic stenosis (AS) is the most common heart valve disease in the elderly, characterized by progressive fibrotic and calcified thickening of the native aortic valve leaflets, leading to severe left ventricular outflow obstruction. With an aging population, AS prevalence is projected to double by 2050. Despite advances in valve replacement therapies, optimal timing of replacement therapies and prediction of long-term outcomes of native aortic valves with early changes remain challenging, as current imaging approaches lack sensitivity for early fibrotic changes. This project aims to harness computed tomography (CT) imaging and artificial intelligence (AI) to quantify fibrotic and calcified changes in the native aortic valve, correlate these findings with functional hemodynamics, and develop predictive models for clinical outcomes over 10 years. Methods and Work Programme: The project will leverage data from over 5,000 patients across three large clinical trials (CAD-Man, DISCHARGE, and IMPRO) with CT imaging and long-term follow-up. It encompasses three main aims. Aim 1: Automated Quantification. AI-supported tools will be developed to automate the segmentation and quantification of fibrotic and calcified tissue, validated against histopathology. Multi-task AI segmentation, will allow automated segmentation and characterization of aortic valve tissue types and enable rapid, reproducible assessments. Periaortic inflammation will also be evaluated as density the adipose tissue around the aortic valve in all 5,000 patients. Aim 2: Hemodynamic Correlation: CT-based anatomical measurements will be correlated with echocardiographic functional parameters at baseline and long-term AS progression until up to ten years of CT-imaging follow-up in the same patients as part of three of our large trials. Gender-specific analyses will explore differences in disease pathophysiology and progression rates. Aim 3: Outcome Prediction. An integrated AI-driven risk score combining imaging biomarkers, clinical variables, and functional data will be developed to predict 10-year clinical outcomes, including mortality, stroke, and procedural complications. External validation will ensure generalizability across diverse patient populations. Anticipated Gain of Knowledge: This project will address critical gaps in understanding AS development and progression by providing robust, noninvasive imaging biomarkers for early disease detection and monitoring. It will offer novel insights into the interplay between valve fibrosis, calcification, inflammation, and hemodynamic dysfunction, enabling personalized risk stratification. By integrating these findings into clinical workflows, the project aims to refine the timing of valve replacement and support the design of trials evaluating novel therapies. Ultimately, the resulting tools and risk models will enhance patient care by improving prognostic accuracy and guiding individualized treatment strategies.

• DFG Programme: Research Grants