Using support vector machines to optimally classify rotator cuff strength data and quantify post-operative strength in rotator cuff tear patients.

Abstract

Shoulder strength data are important for post-operative assessment of shoulder function and have been used in diagnosis of rotator cuff pathology. Support vector machines (SVM) employ complex analysis techniques to solve classification and regression problems. A SVM, a machine learning technique, can be used for analysis and classification of shoulder strength data. The goals of this study were to determine the diagnostic competency of SVM based on shoulder strength data and to apply SVM analysis in efforts to derive a single representative shoulder strength score. Data were taken from fourteen isometric shoulder strength measurements of each shoulder (involved and uninvolved) in 45 rotator cuff tear patients. SVM diagnostic proficiency was found to be comparable to reported ultrasound values. Improvement of shoulder function was accurately represented by a single score in pairwise comparison of the pre-operative and the 12 month post-operative group (P < 0.004). Thus, the SVM-based score may be a promising metric for summarizing rotator cuff strength data.