Predicting the outcome of brace treatment for scoliosis using conditional fuzzy clustering

Adolescent Idiopathic Scoliosis is a spinal deformity sometimes treated using a brace. Brace treatment is successful when it prevents progression (worsening) of the spinal curve during adolescence. However, physicians´ understanding of what causes progression is unclear, so a method for predicting a braced patient´s risk of progression could assist physicians in planning treatment. Some such models have been proposed, but most are ill-suited for clinical use. We applied conditional fuzzy clustering to a dataset of past Scoliosis patients, to generate prototypes representative of various treatment outcomes. New patients´ outcomes were predicted by comparing them to the prototypes. This model is 76-81% accurate, highly interpretable, and easily integrated into the clinic´s workflow, making it a potentially valuable support tool.