نوع مقاله : علمی-پژوهشی
نویسندگان
1 مهندسی پزشکی، دانشکده مهندسی پزشکی، دانشگاه صنعتی سهند، تبریز، ایران
2 عضو هیات علمی/گروه بیوالکتریک، دانشکده مهندسی پزشکی، دانشگاه صنعتی سهند
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسنده [English]
The occurrence of stroke is due to the sudden degeneration of brain cells, which is caused by the lack of oxygen supply to the cells due to vascular blockage or their rupture and the interruption of blood flow, which can lead to gait impairement. The brain imaging techniques including magnetic resonance imaging, computed tomography and cerebral angiography are the main tools for stroke detection, which may not provide a cost-effective and non-invasive diagnosis. Tthis study aims to propose an automatic, non-invasive and low-cost method for Ischemic stroke detection based on computer-aided analysis of the plantar pressure signals. The proposed method is based on new time-frequency plantar feature extraction based on Tunable Q-factor Wavelet Transform, informative feature selection based on ReliefF and classification based on Support Vector Machine, K-Nearest Neighborhood and Random Forest techniques. The main property of this method is the ability to extract fluctuating components and transient information of non-stationary plantar pressure signal using a new time-frequency method and the possibility of adapting to its time-varying characteristics. In order to evaluate the detection performance, the foot pressure signals of 36 patients afflicted with stroke and 46 healthy controls recorded during walking has been used. The obtained results have shown the high diagnostic capability of the proposed method with an average accuracy rate of 99.77% using 35 simple statistical features. The proposed method is able to provide a trade-off between high diagnostic accuracy and low computational cost using simple statistical plantar features, which seems suitable for practical diagnostic applications.
کلیدواژهها [English]
)