Preprocedural CT angiography and machine learning for mortality prediction after transcatheter aortic valve...

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Predicting outcomes after transcatheter aortic valve replacement (TAVR) for aortic stenosis is complex, influenced by valve degeneration and myocardial remodeling.
• Current risk stratification methods may not fully capture patient prognosis.

Purpose of the Study:

• To evaluate machine learning models using pre-procedural computed tomography (CT) for predicting 1-year mortality following TAVR.
• To compare the performance of CT-integrated machine learning against existing clinical risk scores.

Main Methods:

• Machine learning (extreme gradient boosting) was trained on clinical variables and CT-derived imaging data (myocardial mass, valve characteristics).
• The model was validated externally on data from three international sites.
• Performance was compared against EuroScore II and a clinical data-only machine learning score.

Main Results:

• The machine learning model integrating CT data achieved an area under the receiver operator curve of 0.79 in external validation.
• This significantly outperformed EuroScore II (0.59) and a clinical data-only model (0.64) (p < 0.001).

Conclusions:

• Machine learning incorporating clinical data and CT imaging characteristics effectively predicts 1-year all-cause mortality post-TAVR.
• This approach offers superior prognostic risk identification compared to clinical variables or scores alone, aiding pre-procedural patient selection.