Back

A radiomics approach to artificial intelligence in echocardiography: Predicting post-operative right ventricular failure

Shad, R.; Quach, N.; Fong, R.; Langlotz, C. P.; Kong, S.; Kasinpila, P.; Amsallem, M.; Haddad, F.; Shudo, Y.; Woo, Y. J.; Teuteberg, J.; Hiesinger, W.

2020-05-11 cardiovascular medicine
10.1101/2020.05.05.20092494 medRxiv
Show abstract

In this study, we describe a novel radiomics approach to an echocardiography artificial intelligence system that enables the extraction of hundreds of thousands of motion parameters per echocardiography video. We apply this AI system to the clinical problem of predicting post-operative right ventricular failure (RV failure) in heart failure patients receiving implantable circulatory life support systems. Post-operative RV failure is the single largest contributor to short-term mortality in patients with left ventricular assist devices (LVAD); yet predicting which patient is at risk of developing this complication in the pre-operative setting, has remained beyond the abilities of experts in the field. We report results on testing datasets using a standard 10-fold cross validation. The AUC for the AI system trained using the Stanford LVAD dataset was 0.860 (95% CI 0.815-0.905; n = 290 patients) using pre-operative echocardiograms alone. We further show that our system outperforms board certified clinicians equipped with both contemporary risk scores (AUC 0.502 - 0.584) and independently measured echocardiographic metrics (0.519 - 0.598).

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.