A comparative study of methods of processing patient pain drawings for analysis by an artificial neural network

In an effort to aid in the analysis of patient pain drawings, neural networks were used to determine whether diagnostic information acquired from low back patient pain descriptors was best represented by the neurologic dermatomes and gross anatomical areas or by a matrix with each individual, equally weighted, matrix cell representing an area of the pain drawing. The data consisted of 250 patient pain drawings selected from five lumbar spine disorder categories. Over 46000 neural networks were trained using this data. The results show that neurological dermatomes and gross anatomical areas are the best choice for representing diagnostic information acquired from low back patient pain descriptions