Detection of Abnormal Segments in Finger Tapping Waveform using One-class SVM.
Journal:
Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual International Conference
Published Date:
Jul 1, 2019
Abstract
We have developed a finger-tapping device with magnetic sensors, UB2, for the early detection of dementia. The goal of the present study is to develop a method for detecting abnormal segments in the finger tapping waveform in an objective way using machine learning and to evaluate the method in comparison with a human visual assessment. Fifteen-second right-hand finger tapping waveforms of 228 healthy volunteers were measured and cut into one-cycle taps. Fifteen features representing the properties of the one-cycle taps were extracted. As a result of applying a one-class support vector machine (SVM) with an outlier rate of 0.08, 1032 one-cycle taps (8.0%) were detected as abnormal among all 12,898 one-cycle taps. Among these abnormal ones, the features including many outliers (>30%) were the instances of freezing (small fluctuations) and the tap interval. These features correspond to those of which distribution were markedly biased. The visual assessment was likely to overestimate abnormality concerning the instances of freezing and the tap interval (>10%) and conversely underestimate abnormality concerning amplitude of distance/velocity or motion quantity (<; -10%).
