Study on Quantification and Classification of Acupuncture Lifting-thrusting Manipulations on the Basis of Motion Video and Self-organizing Feature Map Neural Network / 上海针灸杂志

ABSTRACT

Objective To obtain the classification and characteristics of acupuncture manipulations by analyzing teacher's lifting-thrusting manipulation parameters derived from Germany-made Simi Motion 3D image analysis system using self-organizing feature map neural network (SOM). Method Thirty acupuncture teachers were selected for the study. Unilateral Quchi (LI11) was used as a point for acupuncture. A motion video of lifting-thrusting even reinforcing-reducing, reinforcing and reducing manipulations was made. The video was analyzed using Simi Motion 3D image analysis system to obtain original motion parameters at thumb tip tracking mark point. Teacher's manipulation classification parameters were analyzed using SOM. Result Acupuncture manipulation parameters showed nonnormal distribution with larger dispersion. Even reinforcing-reducing and reinforcing manipulations could be classified into four types and reducing manipulation, into five types. The characteristics of every type showed general concentration degree. Conclusion Acupuncture manipulation parameters overall showed the diversity characteristics of nonnormal distribution and are typical artificial control curve. They have the characteristic of lower concentration degree of neural network classification. Their classification is mainly based on curve shape and cycle length and related to the degree of difficulty of manipulation. This technique can be applied to the quantitative analysis of various manipulationsand the study of technical inheritance and provide reference for the standardization of acupuncture manipulation.