K-NN based algorithm for degree of stenosis classification using dual non-invasive photoplethysmography system

Abstract

Abstract Haemodialysis (HD) patients who undergo long-term treatment are very susceptible to arterial stenosis. In this study, we propose two main features taken from patients undergoing the dialysis process, namely: Rising Slope – RS and Falling Slope - FS. These features are yielded from a photoplethysmography signal extraction on the hand that is used to create vascular access called HD hand. Eleven dialysis patients with the arteriovenous fistula (AVF) method were the object of this study. The feature data was taken twice, before and after the dialysis process. Utilizing the t-test on the variance of RS features on the HD hand showed a statistically significant value of 0.0211 (p <0.05). Furthermore, these RS features are used as input for KNN classifiers to classify degrees of stenosis in patients undergoing HD. Four patient data with RS features from the HD hand that was not previously known became this classifier test data. From the experimental results showed that K-NN with Euclidean and Minkowski distance could classify the degree of stenosis well. The percentage of misclassification of the system against unknown data was 18 percent (82 percent accurate) based on the cross-validated classification accuracy.