Abstract
OBJECTIVES: Transcatheter aortic valve implantation (TAVI) is crucial for treating calcified aortic stenosis (CAS), yet post-procedural outcomes remain variable. The CAPRI score previously showed promising results in predicting 1-year all-cause mortality, by combining manually measured thoracic aortic calcium (TAC) volume with clinical risk factors. This study introduces an enhanced CAPRI score (eCAPRI), which automates TAC volume measurement and incorporates additional automatically extracted biomarkers from pre-operative CT scans. MATERIALS AND METHODS: TAC volume extraction was automated using a deep learning model trained on 66 patients and evaluated on 1111 CT scans. Additional automatically extracted imaging biomarkers were incorporated into the eCAPRI score, following the original methodology for biomarker selection. The eCAPRI score was trained on 765 pre-TAVI CT scans for one-year mortality prediction and then compared to CAPRI and EuroSCORE Logistic using AUC, bootstrap tests, and calibration curves on 192 CT scans. RESULTS: Automated TAC segmentation achieved a mean Dice score of 0.777 ± 0.108. The eCAPRI score included body surface area (BSA)-indexed right ventricle volume, BSA-indexed pulmonary arteries max diameter, and abdominal muscles surface at L3 level in addition to automatically computed TAC volume and clinical biomarkers previously identified in the CAPRI score. On the evaluation dataset, eCAPRI showed an AUC of 0.731, outperforming CAPRI (AUC = 0.669) and EuroSCORE Logistic (AUC = 0.588) significantly (p = 0.034), with better calibration. CONCLUSION: The eCAPRI score, combining fully automated TAC volume extraction and additional imaging biomarkers, improved one-year mortality prediction over CAPRI and EuroSCORE Logistic. It may support standardized risk stratification in TAVI patients. KEY POINTS: Question Can an eCAPRI using imaging biomarkers from pre-operative CT scans improve the prediction of one-year mortality in patients undergoing transcatheter aortic valve replacement? Findings The eCAPRI score, integrating automated TAC volume and additional biomarkers, outperformed CAPRI and EuroSCORE Logistic in predicting one-year mortality (AUC = 0.731, p = 0.034). Clinical relevance The eCAPRI score provides a standardized approach to mortality risk assessment in transcatheter valve procedures. By improving prediction accuracy, it supports more informed clinical decisions and personalized care planning, ultimately contributing to better outcomes for patients undergoing TAVI.