Time-Frequency Approach for Human Brain Ischemic Stroke Region Detection in CT Images

Abstract

A method of detection of human brain ischemic stroke region in CT slices is described in this article. This method is based on joint time-frequency transform, the discrete Haar wavelet transform, allowing evaluation of frequency and time domain properties of ischemic stroke simultaneously. It was shown that region of ischemic brain stroke may be approximately identified by six time transformed CT images and then corrected by using time domain operations. The time-frequency filter algorithm is developed by combining low pass filtering, application of threshold functions and morphological operations. Experimental results show that method is able to accurately detect ischemic stroke region in most images, especially when ischemic stroke is continuous and relatively large. It also allows to adjust an accuracy of detection and to increase the speed of processing. Ill. 3, bibl. 8 (in Lithuanian; summaries in Lithuanian, English and Russian).