Artificial-Based Pulse Learning Web Application Intelligence Using Convolutional Neural Networks

Abstract

Studying pulse waveforms in healthcare is crucial as they aid in diagnosing and treating chronic diseases. However, the limited data on pulse waveforms makes it challenging for health education to teach this topic effectively. Practitioners of Traditional Chinese Medicine (TCM) require a significant amount of time to obtain pulse wave data accurately. Additionally, the pulse wave data collected by TCM practitioners exhibit various forms and characteristics. This study aims to integrate web-based pulse waveform learning with Artificial Intelligence (AI) using Convolutional Neural Network (CNN) to enhance effectiveness and efficiency. Pulse waveform data were obtained from Traditional Chinese Pulse Diagnosis and were redrawn to achieve diverse and accurate results. A total of 400 images were generated for each of the five types of pulse waveforms to improve data quality. The redrawn data were then tested to ensure accuracy. Once validated, a comparison of deep learning models using three CNN architectures—VGG16, VGG19, and ResNet50—was conducted, with VGG19 achieving the highest accuracy among the models. Consequently, the VGG19 model was implemented into a web-based pulse waveform learning application using JavaScript, HTML5, and TensorFlow. The results demonstrate that the VGG19 model outperformed other architectures in terms of accuracy. The successful integration of the VGG19 model into the web-based application shows that AI can be used to create an interactive learning platform for pulse waveform education. This study proves that the collaboration between web-based pulse waveform learning and AI can serve as an interactive educational tool for the future.