Graph Convolutional Networks for Deciding Acupoints in Acupuncture and Moxibustion Treatment

Acupuncture and moxibustion treatment (AMT) is widely recognized for its efficacy in restoring balance and al-leviating pain by regulating Qi and blood flow. The integration of artificial intelligence (AI) with AMT shows potential for enhancing treatment effects, yet also presents challenges in ensuring accuracy and safety. This study aims to address these challenges by appropriately processing raw data in AMT and exploring effective ways to leverage machine learning methods for personalized acupoints prescriptions. Specifically, we build a database correlating symptoms with acupoints and refined the differentiation of symptom importance. Based on this, we construct a Traditional Chinese Medicine (TCM) graph consisting of disease case nodes and symptom nodes. We employ Graph Convolutional Networks (GCN) to perform supervised learning on the graph with acupoints as labels. This approach enables the provision of personalized acupoints prescriptions for treating patients. Our study reveals promising avenues for improving acupoints prescription accuracy through the proposed techniques, despite current limitations. This research underscores the potential of integrating AI technologies to offer personalized treatment recommendations in AMT.