Combining the body size data and the biochemical data can help us to analyze the relationships between 'body size' and 'disease'. After that, a good diagnosis function by correctly preventing disease and high efficiency of diagnosis would satisfied with people who in therapy service. The aim of this study is to find an optimum structure to explain the relationships between 'body size' and 'disease' by data mining techniques. The data measured by the three-dimension (3D) body scanner and the biochemical data of health examination. The data mining method used back-propagation (BP) from neural network of artificial intelligence. BP would reduce some unimportant factors and conduct a new structure. Although the root mean square error (RMSE) is rather inferior to previous one, its module relation level is still more than 90%. As well as we can decrease the cost while collect medicine data.