머신 러닝 기반 혈류 진동 데이터의 당뇨병 분류 특성 중요도 분석

Abstract

In this paper, we demonstrated the diabetes classification potential of a machine learning algorithm for blood flow oscillation data. The blood flow oscillation data was extracted using wavelet transform by measuring the blood flow signal of rats with a Diffuse Speckle Contrast Analysis (DSCA) system. The test analysis of the blood flow reactivity demonstrated that additional experiments are not required to classify diabetes. In addition, feature importance analysis showed that blood flow oscillations of cardiac and respiratory activities play an important role in classifying diabetes.