Classification of cardiac doppler signals by using artificial neural network and NEFCLASS

In this work, cardiac Doppler signals recorded from aorta valve of 60patients were transferred to a personal computer by using a 16 bit soundcard. The fast Fourier transform (FFT) analysis was applied to the recordedsignal from each patient in order to obtain systole, diastole, resistive index,pulsality index and systole/diastole ratio values. Further these values wereclassified by using multi layer perception neural network and NEFCLASSneuro fuzzy classifier. Thus, an additional diagnosis tool is developed for theaid of expert medical staff. It was obtained that, 96,67% classificationsuccess rate from multi layer perception neural network, and NEFCLASS.On the other hand, power spectrum density curves of cardiac Dopplersignals which recorded from mitral valve were obtained. According to thecurves formed after power spectrum density analysis, 10 power spectrumdensity values which are corresponding to the increases by 500 Hz have beenapplied as input vectors and classified for multi layer perception neuralnetwork and NEFCLASS. It is seen that, 93,33% classification success ratesfrom multi layer perception neural network, and 90% that of NEFCLASSwere obtained. Furthermore, by the linguistic terms in the NEFCLASSvineuro fuzzy classifier, the diagnosis given by the classifier became easilyunderstood and interpreted by the doctor and the patient.Science Code : 626.06.01Key Words : Neuro Fuzzy Classifier, multi layer perception neuralnetwork, NEFCLASS, cardiac Doppler, Fast FourierTransform (FFT), aorta valve, mitral valve.Page Number : 74Advisers : Prof. Dr. nan GÜLER, Assist. Prof. Dr. Fırat HARDALAÇ