Pulse waveform analysis for pregnancy diagnosis based on machine learning

Measuring and feeling the pulse waveform of radial artery at patient's arm wrist has been used in traditional Chinese medicine for thousand years as a diagnostic tool for pregnancy detection. However, in traditional Chinese medicine practice, the pulse waveform was measured by doctors' fingertips, and it is too subjective to quantify and standardize the complex pulse waveform. Its accuracy has been an open question from modern technology perspective. Machine learning is an emerging avenue to transform the complex pulse waveform to a quantitative data which can make diagnosis of pregnancy more precisely. In this study, we used four algorithms of machine learning to analyze digital pulse waveform signal for recognition of pregnancy. Pulse signals from 495 women were measured and collected by a modern digital pulse waveform device. This device converted the pulse waveform into a one-dimensional time domain signal. Both time domain and wavelet domain features were extracted for further waveform analysis. Then these features were sent to different supervised learning models for pattern recognition. Adaptive boosting showed the best detection accuracy among them. Meanwhile, the group of women in the second and third trimester of pregnancy demonstrated higher recognition rate than the group that in first trimester of pregnancy. Our results indicate that pulse waveform can be recorded and analyzed objectively, and machine learning is an effective tool for pregnancy diagnosis in clinical practice.