Early and noninvasive screening of common cardio vascular related diseases such as diabetes and cerebral infarction using photoplethysmograph signals

Abstract

It is significant to detect the inability of the heart at its early stage, because most of the deaths in all over the world is due to cardiovascular diseases (CVD) (Atlas of Diabetes, 2012). This paper presents a novel learning technique; utilizing Kmeans clustering strategy, Fine Gaussian Multi class Support Vector Machine (FGMSVM) with one-against-one and maximum votingtechnique, to classify the samples to its respective classes, for noninvasive screening of common CVD related diseases. These diseases are investigated by the measurement of Photoplethysmogram (PPG) signals acquired from the subjects with various kinds of diseases. It is seen that the parameters used in this work show clear changes in their values, thus exhibiting their adequacy in recognizing distinctive heart related diseases. The support vectors, which are critical for classification, are obtained by learning from the training samples. Capability of different kernel functions in classification is also demonstrated.

Keywords

• Cardiovascular risk assessment
• FGMSVM
• CVD related diseases
• Cerebralinfarction
• Photoplethysmogram PPG