Classification of Syncope in Front-Loaded Head-Up Tilt Test with Support Vector Machine

Syncope describes a condition that causes a brief period of unconsciousness. It leads to a range of undesirable outcomes, including people suffering from fractures or having their bones broken. Clinically, a standard assessment protocol in syncope, head-up tilt (HUT) table test, was introduced. However, this assessment is limited with its unstable sensitivity and time consuming (40-45 minutes of tilting). Therefore, the aim of the study is to design an algorithm to classify subjects with and without syncope based on physiological signals (electrocardiography and blood pressure) acquired from front-loaded HUT test. The study selected a total of 52 people, 25 of whom were syncope negative and 27 of whom had syncope. Subject was rested in supine position for 10 minutes, followed with tilting at 70-degree on a tilt table for 20 minutes. Once the subject is tilted, an 800 micrograms of glyceryl trinitrate (GTN) was administered. A series of physiological signal processing and relevant hemodynamic parameters were computed. Then, feature selection was carried out with recursive feature elimination (RFE), to eliminate and determine the optimum number of features. A support vector machine (SVM) classifier was then used to classify the selected features via 5-fold cross-validation. Our model achieved an accuracy of 86.5%, precision of 85.7%, and recall of 88.9%. For patient evaluation, the proposed methodology is a viable strategy for determining if a patient is syncope positive or not.