Application of a BP Neural Network Based on Principal Component Analysis in ECG Diagnosis of the Right Ventricular Hypertrophy

摘要

Objective: To find the prediction model of the Right Ventricular Hypertrophy (RVH) diagnosed by the ECG, and to study the techniques to improve the rate of the clinical diagnosis. Method : 60 cases of RVH and 25 cases of normal were selected. The following 5 indexes are collected from the 85 records respectively: ages, heart rates, the sums of the amplitude of the R wave in lead V1 and the depth of the S wave in lead V5, the amplitudes of the inverted T wave in lead V1, and the deviation degrees of the right axis etc. Firstly, we used the principal component analysis (PCA) to pre-analyze the original multi-objective variables; Then 55 cases (including RVH and normal) of the total 85 were used as the training sample to input the BP Neural network, and the residue cases as the testing sample. Result: 3 principal components were extracted and their total explained variance was 85.26%; Using the principal components as the input of the network, we got the prediction sensitivity was 99.5%, the specificity 100%. While using the original variables, the sensitivity is 98.5%, the specificity 100%. Conclusion: The BP neural network model based on the PCA can be used to predicate the RVH and to improve the accuracy.