An intelligent diagnosis system based on principle component analysis and ANFIS for the heart valve diseases Engin Avci * , Ibrahim Turkoglu Firat University, Department of Electronic and Computer Science, 23119 Elazig, Turkey Abstract In this paper, an intelligent diagnosis system based on principle component analysis (PCA) and adaptive network based on fuzzy inference system (ANFIS) for the heart valve disease is introduced. This intelligent system deals with combination of the feature extrac- tion and classification from measured Doppler signal waveforms at the heart valve using the Doppler ultrasound (DHS). Here, the wave- let entropy is used as features. This intelligent system has three phases. In pre-processing phase, the data acquisition and pre-processing for DHS signals are performed. In feature extraction phase, the feature vector is extracted by calculating the 12 wavelet entropy values for per DHS signal and dimension of Doppler signal dataset, which are 12 features, is reduced to 6 features using PCA. In classification phase, these reduced wavelet entropy features are given to inputs ANFIS classifier. The correct diagnosis performance of the PCA– ANFIS intelligent system is calculated in 215 samples. The classification accuracy of this PCA–ANFIS intelligent system was 96% for normal subjects and 93.1% for abnormal subjects. Ó 2008 Elsevier Ltd. All rights reserved. Keywords: Principle component analysis; ANFIS; Doppler heart sounds; Discrete wavelet decomposition; Wavelet entropy 1. Introduction The heart valve disorders are significant heart diseases, which are the aortic inadequacy and mitral inadequacy etc (Turkoglu, Arslan, & ve Ilkay, 2003). In recently, the techniques used for diagnosis of heart valve disorders sepa- rate two groups: non-invasive and invasive techniques. The first techniques are heart sounds, electrocardiograms, ultra- sound imaging, murmur from stethoscope, and Doppler techniques. The second techniques are transozefagial echo- cardiograph angiography methods (Nanda et al., 1993). These non-invasive techniques and invasive techniques have many advantages and disadvantages (Plett et al., 2000). The diagnosis of heart valve disorders studies are related on improving human-machine interfaces in existing methods. The main aim of these studies is providing to be understood the output of the examination systems more easily and diagnose the problem more accurately by the cardiologists (Philpot, Yoganathan, & Nanda, 1993). In nowadays, the Doppler techniques are the most preferred due to their completely non-invasive and without risk in the serial studies for early diagnosis of the heart diseases. DHS are composed by valves motion, blood flow, and vibration of the other cardiovascular components (C ¸ omak, Arslan, & Turkoglu, 2006; Jing, Xuemin, Mingshi, & Wie, 1997). In Doppler literature, there are many Doppler tech- niques. These Doppler techniques are the pulsatility index (PI), Pourcelot or resistance index (RI) and A/B Systolic Diastolic ratio, which are highly correlated and led to highly erroneous diagnostic results (Chan, Chan, Lam, Lui, & Poon, 1997). Here, the peak systolic and end-diastolic velocities are used for diagnosis in these methods. In today’s, very vari- ous methods, which are the PCA, Generalized Discrimi- nate Analysis, Linear Discriminant Analysis etc., are used for feature extraction on pattern recognition. 0957-4174/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.eswa.2008.01.030 * Corresponding author. Tel.: +90 4242370000x4257; fax: +90 4242367064. E-mail address: enginavci23@hotmail.com (E. Avci). www.elsevier.com/locate/eswa Available online at www.sciencedirect.com Expert Systems with Applications 36 (2009) 2873–2878 Expert Systems with Applications