Electrocardiographic criteria for the diagnosis of left ventricular hypertrophy based on data mining

Left ventricular hypertrophy (LVH) is the main manifestation of cardiovascular disease in patients with hypertension, and is an independent risk factor for multiple cardiovascular complications. So the medical researchers attach enough importance to it in aspect of clinical practice. Electrocardiograph (ECG) has the unique advantages of simple operation and low price, which support its widespread use for screening LVH. So far, the American Heart Association has approved 37 different ECG criteria, but it is widely believed that all of these criteria show problems of high specificity but low sensitivity. Therefore, finding a new ECG criterion which has better diagnostic value becomes a research hotspot. The purpose of this study is to propose a new method for the selection of ECG diagnostic criteria based on the method of data mining. The process of algorithm research including ECG signal preprocessing, feature extraction, feature selection, and diagnostic criteria establishment. The results show that the criteria proposed by our method significantly improve the sensitivity (up to 80%) and the overall accuracy.