Initial-impression diagnosis using low-back pain patient pain drawings

Abstract

Patient pain drawings were blindly selected from five lumbar spine disorder categories. The drawings were classified by low-back physicians, discriminant analysis, and several computerized artificial neural network configurations. The purpose was to determine the reliability of the patient pain drawing when diagnosing low-back disorders and to delineate the pain mark patterns particular to each disorder by comparing physicians with computerized methods. The physicians averaged 51% accuracy with individual preferences for certain disorder groups. The computerized methods demonstrated comparable accuracy (48%) and more agreement in classification. Associations were found between the predicted pain patterns for each diagnostic group made by an expert and the patterns generated by computerized methods. Variances in these associations are instructive to clinicians for making accurate predictions of diagnosis from pain drawings.