Classification of nailfold capillary images in patients with hypertension using non-linear SVM

Hypertension poses a serious atherosclerotic risk as it causes both macro- and micro circulation damage. Nailfold capillaroscopy is a valuable yet simple tool to assess microcirculation of blood capillaries. This technique is important in detecting early occurrences of scleroderma spectrum disorders and evaluating Raynaud's Phenomenon. Here it is used in detecting hypertension in patients. Current methods in Nailfold Capillaroscopy involve processing of high resolution images acquired using a commercial capillaroscope whilst the analysis of low resolution images acquired using a low cost hardware is a challenging task. In this paper we preprocess the low resolution capillary images and use Discrete Wavelet Transform for feature extraction. The nonlinear kernels of Support Vector Machine are used to classify the images into either normal or hypertensive subjects. Polynomial kernel is found to produce better results compared to the other kernels.