SMARTPHONE-BASED DIAGNOSTIC MODEL FOR HYPERTENSION USING FEATURES FROM PHOTOPLETHYSMOGRAM

Abstract

Photoplethysmography (PPG) is an optical technique which measures blood volume changes in the arterial blood using red and IR LEDs of wavelengths 660[Formula: see text]nm and 940[Formula: see text]nm, respectively. This paper proposes a methodology to measure the pulse rate from the video signal obtained using an LETV-LE MAX 2 mobile phone’s camera and also to evaluate hypertension. The Android smartphone records the intensity of light reflected from the index finger. The recorded video is separated into red, green and blue frames. Since the red video frames returned useful plethysmographic information, they are filtered using Butterworth band-pass filter and power spectral density analysis was performed on them. The immediate peak gives the pulse rate of the respective subject. Fifteen features of pulse waveform are extracted and by performing the feature selection process, seven features are selected and they undergo classification process using a neural network. The feature selection process is performed by using the eigenvalues of the principal component analysis method. The eigenvalues obtained from this method show the degree of variation present in the data. The eigenvalue that is near or close to zero gives the principal components. The features that are selected by the feature selection process of principal component analysis method are peak interval, settling time, rise time, normalized PPG, peak-to-peak amplitude, first derivative and second derivative. While performing the classification process using a neural network, the accuracy of prediction was calculated for both the normal and hypertensive subjects.