Table 4.

Summary of publications describing artificial intelligence application in guiding mechanical circulatory support practice.

Author	Title	Data source used	Primary outcome	Best model performance	Study conclusion
Aras (36)	InDetector – Automatic detection of infected driveline regions	Local institutional data – University of Arizona, United states	Diagnosis of LVAD driveline infections	The convolution neural network with image augmentation had 93.75% accuracy.	InDetector, a smartphone-based application, allows at-home patients to send images of their driveline to a remote server which used an AI model to classify the driveline region as clean or infected.
Maw (37)	Development of a suction detection algorithm from patient pump data	Local Institutional Data, Medical University of Vienna	Detection of HVAD suction events	The supervised learning algorithm had 92.5% sensitivity and 100% specificity.	The proposed algorithm for suction detection may be used as diagnostic marker, or as a component of an automatic physiologic controller in patients with HVAD pumps.
Topkara (38)	Machine learning-based prediction of myocardial recovery in patients with left ventricular assist device support	INTERMACS, 2008–2017	LVAD explant specifically due to myocardial recovery	Bayesian logistic regression model achieved the highest AUC of 0.824.	Machine learning can be a valuable tool to identify subsets of LVAD patients who may be more likely to respond to myocardial recovery protocols.

HVAD, heartware ventricular assist device. LVAD, left ventricular assist device. INTERMACS, Interagency Registry for Mechanically Assisted Circulatory Support. strong recommendation, weak recommendation/promising but not ready to implement, as suggested by the respective paper.