Classification of internal carotid artery Doppler signals using fuzzy discrete hidden Markov model Harun Ug ˘uz ⇑ , Halife Kodaz Department of Computer Engineering, Selçuk University, Konya, Turkey article info Keywords: Fuzzy discrete hidden Markov model Doppler signal Carotid artery Power spectral density Autoregressive method abstract We developed a biomedical system based on Discrete Hidden Markov Model (DHMM). The aim of our system is to classify the internal carotid artery Doppler signals. We applied a fuzzy approach to DHMM. Thus we decreased information loss and increased the classification performance. Our system reached 97.38% of classification accuracy with 5 fold cross validation. These results showed that the Fuzzy Dis- crete Hidden Markov Model (FDHMM) method is effective for classification of internal carotid artery Doppler signals. Ó 2010 Elsevier Ltd. All rights reserved. 1. Introduction Carotid artery is a disease affecting the vessels going to the head and brain. The symptoms of the carotid artery disease are stenosis and occlusion in the interval carotid artery. Internal carotid artery plaques cause these symptoms (Baker, 1985). Angiography and blood tests can be used for the diagnosis of the disease. However, being invasive, these techniques are not favorable. Doppler tech- niques are non-invasive and do not include any risk (Wright, Gough, Rakebrandt, Wahab, & Woodcock, 1997). Therefore this technique is usually used for demonstrates the flow characteristic of carotid arteries. As a medical application of Doppler principle, an ultrasonic wave is sent to the body and the shift of the wave’s frequency is measured detecting the reflected wave from the red blood cells in the body. The Doppler signal frequency difference between the incoming and reflected waves is directly proportional to the blood velocity. The scattering caused by the reflected incoming waves from the red blood cells exhibit a frequency spectrum. Conclusions can be drawn as to the blood flow by the inspection of the changes in this frequency spectrum. Based on the analysis of the spectral form and the parameters, it can be determined whether the sign belongs to a healthy or diseased artery (Evans, McDicken, Skidmore, & Woodcock, 1989; Übeyli & Güler, 2003a). The power spectral density (PSD) estimates can be used to track the velocity distribution, since the velocity components will be proportional to the frequency shifts (Güler, Kiymik, Akin, & Aklan, 2001; Übeyli and Güler, 2003b, 2004). Spectrum analysis tech- niques were used to obtain the Doppler PSD estimates of carotid arterial Doppler signals. Recently, making use of various spectrum analysis methods in the processing of internal carotid artery Doppler signals (Kara & Dirgenali, 2007; Özs ßen, Kara, Latifog ˘lu, & Günes ß, 2007; Übeyli & Güler, 2004a, 2004b, 2008). AR method is a common practice based on the fact that solving linear equations makes it easy to estimate the AR parameters. For that reason, AR method is widely used as a spectrum analysis technique (Özs ßen et al., 2007; Übeyli & Güler, 2004b, 2003a, 2003b). Yule–Walker, Burg, covariance, least squares, and maximum likelihood estimation are examples to the various estimation methods for the parameters of AR method (Stoica & Moses, 1997; Kay, 1988). Burg AR method stands as a powerful calculation method in terms of computa- tional efficiency, and resulting in consistent estimates, is fre- quently used in PSD estimates of internal carotid artery Doppler signals. In our study, Burg AR method has been employed for the spectrum analysis of the internal carotid artery Doppler sig- nals. The PSD values obtained from the Doppler signals by the Burg AR method were applied as input to the FDHMM based clas- sification system, which is used to classify the healthy and the unhealthy subjects. In the literature, there are various studies on classification of the carotid artery Doppler signals. Özs ßen et al. extracted the features with AR method and classified these features using a new Artificial Immune Systems classifier (Özs ßen et al., 2007). Cey- lan et al. used Complex Valued Artificial Neural Network (CVANN) structure to classify carotid artery Doppler signals using Principal Component Analysis and Fuzzy c-means Clustering (FCM) as fea- ture extraction methods before the CVANN classifier (Ceylan, Ceylan, Dirgenali, & Özbay, 2007). Özbay & Ceylan used Fast Fou- rier Transform, Hilbert Transform, and Welch Method with differ- ent window types. They investigated effects of window types on classification of carotid artery Doppler signals (Özbay & Ceylan, 2007). Polat et al. used Support Vector Machine (LSSVM) with a 0957-4174/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.eswa.2010.12.084 ⇑ Corresponding author. Tel.: +90 332 223 19 26; fax: +90 332 241 06 35. E-mail address: harun_uguz@selcuk.edu.tr (H. Ug ˘uz). Expert Systems with Applications 38 (2011) 7407–7414 Contents lists available at ScienceDirect Expert Systems with Applications journal homepage: www.elsevier.com/locate/eswa