Abstract

SpasticHemiplegia(SH)isatypeofcerebralpalsyaffectingtheupperandlowerlimbson the same side of the body. Dr. Gage has suggested that this pathology could be classified into at least four groups considering the kinematic pattern on three planes (sagittal, transverse and co - ronal). Traditionally, this classification of pathological patients is done by specialized physi - cians based on physical study and the clinical gait analysis of patients, which uses complemen - tary tests reported on kinetic, kinematic and electromyographical records. The application of automatic classification techniques using computerized methods is a support for this task, a di - agnostic support tool, not a replacement for the specialist. This paper analyzes different auto - matic classification algorithms using supervised learning on gait records from a population of 255 patients collected over six years. Logistical regression trees combined with meta-classifiers have demonstrated good efficacy in the classification task. This effectiveness is evaluated quan- titatively by measuring specificity, number of correct classifications, false positives and nega- tives, among other measurements. Validation with specialized physicians makes it possible to compare results obtained by the automatic models with real expert diagnoses