Palpation localization of radial artery based on 3-dimensional convolutional neural networks

Abstract Palpation localization is essential for detecting physiological parameters of the radial artery pulse diagnosis Traditional Chinese Medicine (TCM). Detecting signal or applying pressure at wrong location can seriously affect measurement waves and result in misdiagnosis. In this paper, we propose an effective high accuracy regression model using 3-dimensional convolution neural networks (CNN) processing near-infrared picture sequences to locate upon radius wrist. Comparing with early studies 2-dimensional models, 3Dcnn introduces temporal features third dimension leverage pulsation rhythms, had achieved superior performance as 0.87 within 50 pixels testing resolution 1024 × 544. Model visualization shows that additional highlights dynamic changes image sequences. This study presents great potential our constructed be applied real wrist palpation scenarios bring key convenience diagnosis.