1 Automated Spectral Analysis for Pediatric Cardiac Auscultation Azra Rasouli Kenari*, Dr. M. Hassan Ghassemian** Faculty of Electrical and Computer Engineering, Tarbiat Modares University, Tehran, Iran. *azra.rasouli@modares.ac.ir ** ghassemi@modares.ac.ir Abstract: Early recognition of heart disease is an important goal in pediatrics. Developing countries have a large population of children living with undiagnosed heart murmurs. As a result of an accompanying skills shortage, most of these children will not get the necessary treatment. We designed a system for automatically detecting systolic murmurs due to a variety of conditions. This could enable health care providers in developing countries with tools to screen large amounts of children without the need for expensive equipment or specialist skills. For this purpose an algorithm was designed and tested to detect heart murmurs in digitally recorded signals. A specificity of 100% and a sensitivity of 90.57% were achieved using signal processing techniques and a k-nn as classifier. Keywords: Auscultation, cardiac, k-nn classifier, pediatric, wavelet. 1. Introduction A heart murmur is an extra heart sound in addition to the two heart sounds normally heard with each heartbeat. A heart murmur in pediatrics can be an indicator of congenital heart disease. Innocent heart murmurs are of no clinical consequence while pathologic heart murmurs indicate congenital heart disease is present [1]. The incidence of heart murmurs in the pediatric population is reportedly as high as 77% to 95%. However, less than 1% of this population has heart disease. Early recognition is an important goal, and equally important is avoiding misdiagnosing a pathological heart murmur in a healthy child without heart disease. To acquire high-quality auscultation skills, requires the guidance of an experienced instructor using a sizable number of patients along with frequent practice. Unfortunately, the interpretation of auscultation findings overall remains prone to error. Imaging technologies can provide more direct evidence of heart disease; however, they are generally more costly. There is an acute shortage of physicians in developing countries and many rural clinics are run by nurses. Given the high incidence of heart murmurs, automated screening based on electronic auscultation at clinic level would be of great benefit. Acceptance will obviously depend on the sensitivity and specificity of the system. Selection of representative data for diagnosis must be relatively simple for the system to be of practical use in rural clinics [2]. Recent advances in digital signal processing have led to a reexamination of the potential role of spectral analysis of heart sounds in cardiac diagnosis [3-7]. In this study, we investigated a new technique for evaluating heart murmurs in children and young adults using automated analysis of the systolic energy content found in digital recordings of cardiac auscultations. In this work we build on the “prototypical systole” and then extract two different feature sets that characterize acoustic activity in the systolic phase. One feature set is related to physiological activity of heart sounds. The other is the first three principal components derived from principal component analysis. The objective of our study was to assist the clinician in the detection and evaluation of heart murmurs. 2. Methods 2.1 Data Acquisition Cardiac auscultatory examinations of 93 children and young adults were recorded, digitized, and stored along with corresponding echocardiographic diagnoses. 40 subjects were diagnosed as normal and the other 53 subjects were found to have one of the pathological cardiac conditions namely VSD, AS and PS. For each patient a 10 second, ECG and heart sound (HS) recording was made at the Apex in the supine position. The recordings were made by experienced pediatric cardiologists, and in most cases, in a noisy clinical environment. Echocardiography was done on all patients to confirm patient diagnosis. The WelchAllyn Meditron Analyzer, ECG and electronic stethoscope were used during the auscultation. The digital recordings were made at a sampling frequency of 44.1 KHZ, 16 bit precision and saved via USB onto a laptop in the uncompressed WAV file format. 2.2 Pre-Processing Pre-processing of heart sounds is necessary to obtain consistent useful data for analysis and improve robustness in the presence of noise. Heart sounds were