A five-factor machine learning model accurately predicted periprocedural adverse events in patients with bicuspid aortic valve stenosis undergoing TAVR, achieving an area under the curve of 0.801.
Can a machine learning model predict periprocedural adverse events in patients with bicuspid aortic valve stenosis undergoing TAVR?
A machine learning model incorporating clinical and imaging factors can effectively predict the risk of periprocedural adverse events in patients with bicuspid aortic valve stenosis undergoing TAVR.
Absolute Event Rate: 0% vs 0%
Background Transcatheter aortic valve replacement (TAVR) has increasingly emerged as one of the primary treatments for patients with severe bicuspid aortic valve (BAV) stenosis. Nevertheless, these patients encounter multiple procedural challenges. Objective To develop a machine learning (ML) model for assessing the risk of periprocedural adverse events (PAEs) in TAVR population with BAV. Methods This multicentre study retrospectively enrolled 1266 patients with BAV stenosis. Clinical characteristics and imaging data of the patients were collected, and an ML prediction model was developed. PAE was collectively defined as all-cause death, disabling stroke, life-threatening haemorrhage, acute kidney injury (≥stage 3), major vascular complications, valve-related dysfunction necessitating reoperation and other major complications that occurred prior to discharge. Results The average age was 72.6±6.3 years, and 58.3% (n=738) of male. In the derivation dataset, five predictive factors were identified: Type 0 BAV, aortic root calcification volume, horizontal aorta, annular ellipticity and previous atrial fibrillation. A robust risk scoring model was thereby established (area under the curve=0.801 95% CI 0.768 to 0.832). A graded relationship was observed between the quartiles of the score and PAE (0.6%, 1.7%, 3.2% and 9.6%; overall p<0.001). A nomogram was constructed to enable calculation of individual scores and the corresponding PAE probabilities. Additionally, similar results were observed in the validation dataset. Conclusions The ML model developed in this study could predict the PAEs occurrence of TAVR in patients with BAV stenosis. This is conducive to individualised procedural planning and in-hospital management.
Mao et al. (Thu,) reported a other. A five-factor machine learning model accurately predicted periprocedural adverse events in patients with bicuspid aortic valve stenosis undergoing TAVR, achieving an area under the curve of 0.801.