Quantitative Analysis of Facial Paralysis Based on TCM Acupuncture Point Identification

Traditional methods for recognizing facial palsy suffer from factors such as reliance on professional doctors and high subjectivity. Existing intelligent methods for detecting facial palsy mainly extract facial features and base their judgments on face asymmetry, but they cannot perform quantitative analysis of the degree of facial palsy. To address these limitations, we propose a novel approach to diagnose and quantitatively analyze facial palsy based on facial acupuncture point recognition.Our method involves identifying three sets of acupuncture points related to facial palsy by accessing an improved Resnet model after detecting, cropping, and pre-processing the input image. We then measure the intersection angle of the line connecting these points and the vertical line of the middle symmetry line of the face to achieve quantitative analysis and grading of the degree of facial nerve damage.The aim of our proposed model is to quantify and grade the degree of facial nerve damage. We compared our model with VGG and MobileNet frameworks, and the experimental results demonstrate that our proposed model is effective in acupoint recognition and has the lowest error in quantitative analysis, and the average angle error can be reduced to 4.5 degrees. By using facial acupuncture point recognition to diagnose and grade facial palsy, our method provides a more objective and accurate approach to assessing this condition.