Automatic Feature Extraction in Accelerated Plethysmography

Abstract

The heart works as a pumping machine which circulates blood in whole body and nourishes every cell of the body. The pulse generated by heart is sensed with the help of Photo plethysmography technique. Plethysmography is a test which is used to measure the blood clots in arms and legs and how much air one can store in the lungs. The changes in finger blood flow is given by the pulse wave amplitude derived from Finger plethysmography and is analyzed for its diagnostic abilities. The motive of this study is to analyze the features extracted from the second derivative of the pulse waveform. Further, the parameters are to be found which are directly link to the increase level of pitta. Pitta is one of the Ayurvedic dosha which may be the disease causing agent in the human body. It has been analyzed that the pitta level increases due to sunlight and after having mid-day meals. In this dissertation work the second derivative of the finger PPG waveform has been studied and some important features have been extracted from it. The data has been recorded with the help of BIOPAC MP System and Acknowledge software. For this study data of 36 subjects has been acquired. Finger PPG’s of three fingers namely index, middle and ring of both left and right hands has acquired after breakfast, before and after lunch. Eight parameters from each finger has extracted using computer algorithm. From a total of 36 subjects best 25 subjects are being taken. For comparison, two groups have been made, named as group 1(after breakfast and before lunch) and group 2(after breakfast and after lunch). Further analysis is done by applying Fisher discriminant ratio and t-Test on these two groups. Out of these 48 parameters, 13 and 16 such parameters from group 1 and 2 respectively are found that having high Fisher discriminant ratio. After this analysis, the correlation matrices are formed for these 13 and 16 parameters. Parameters having correlation more than ‘0.6’ are further selected, the lower Fisher discriminate ratio parameter is eliminated. These correlation matrices yield 6 and 10 parameters from each group respectively for further investigation. After these 6 and 10 parameters of group 1 and 2 respectively, 4 parameters were found to be common. Those parameters are recommended for further studies on detection of high level of pitta.