An application of discriminant analysis on electrocardiogram

• Main Authors: Chen Yen-hung, 陳彥宏
• Other Authors: 白志東
• Format: Others
• Language: zh-TW
• Published: 1998
• Online Access: http://ndltd.ncl.edu.tw/handle/08608438031191275764

碩士 === 國立中山大學 === 應用數學研究所 === 86 === In this paper, we extract three statistics from ECG data: sample mean,sample standard deviation and sample median, from R-R intervals, P-R intervals and R-T intervals respectively. A quadratic discriminantfunction is applied to classify among: normal, cyanotic congenital heart disease (C. CHD), cyanotic congenital heart disease with congestive heart failure (C. CHD+CHF), noncyanotic congenital heartdisease (N. CHD), noncyanotic congenital heart disease with congestiveheart failure (N. CHD+CHF), ventricular septal defect (VSD), ventricularseptal defect with congestive heart failure (VSD+CHF) according to thesenine statistics. We compute the sensitivity and specificity of eachcombination of statistics, use odds ratio to determine the best discriminantvariable set. The future goal is to apply this study to build up an auto-diagnostic system of children''s ECG. We found that the sample standarddeviation of R-R interval is an important variable in classifying congenitalheart disease and congestive heart failure with other diseases. And thestatistics calculated from R-T interval are important variables of classifyingventricular septal defect patients with normal patients.