A Comparative Study of CNN Architectures in TCM Hesitant Pulse Classification

Traditional Chinese Medicine (TCM) is an alternative medicine practice that originated in China. TCM is based on the concepts of Qi, Yin, and Yang and aims to restore balance in the body. In TCM, the pulse rate is divided into two categories: hesitant and moderate. A hesitant pulse is irregular and has inconsistent intervals between beats, while a moderate pulse falls within the healthy and normal range for adults. Previous research has used deep learning for pulse classification, but none has utilized VGG16, AlexNet, and LeNet. This study aims to improve the accuracy of pulse classification analysis and minimize errors. The research employed a quantitative method and utilized three models. The study used 48 hesitant data and 80 moderate data. The f1-score values for the VGG16 model were 0.0 for hesitant data and 0.76 for moderate data. The AlexNet model achieved f1-scores of 0.90 for hesitant data and 0.94 for moderate data. The LeNet model achieved f1-scores of 0.95 for both hesitant and moderate data. Based on the f1-scores, the LeNet model demonstrated high performance. These results could assist healthcare professionals in improving their detection and analysis of pulse data.