Automated scoring of patient pain drawings using artificial neural networks: efforts toward a low back pain triage application.

The goal of this research was to examine methods of automatically scoring patient pain drawings. Two hundred and fifty pain drawings were selected from the files of an orthopaedic surgeon who specializes in the treatment of low back pain patients. An artificial neural network was designed to score these drawings. The drawings were segmented into 85 regions following dermatomal mappings and from these regions the percent area in pain in each was computed and used as the neural network input variables. With five outcome categories (scores) we obtained a classification sensitivity of 49%, which is approximately as well as physician experts and discriminant analysis achieved using a subset of the same data. We conclude that an artificial neural network is well suited to automatically score patient pain drawings.