BLOOD-BIOMARKER-BASED ACOUSTICAL FEATURE SELECTION IMPROVES OVERALL CLASSIFICATION ACCURACY FOR PUMP THROMBOSIS IN LEFT VENTRICULAR ASSIST DEVICES

The performance of biosignal classification hinges on feature selection, yet in many cases such features are selected ad hoc with an empirical, black box approach. Combining heterogeneous data streams -specifically physiologically meaningful blood biomarkers and biosignal features – can