Detection and Grading of Knee Joint Cartilage Defect using Multi-Class Classification in Vibroarthrography

Vibroarthrography describes the detection of joint pathologies by analysis of vibrations emitted during joint movement. In our study, 30 healthy volunteers and 39 patients with various degrees of chondromalacia or osteoarthritis were selected and accelerometers and piezoelectric sensors were placed on prominent bone structures of patients’ knee joints (patella, lateral and medial tibial plateau) in order to measure the structure-borne noise during active extension and flexion of the joint. After semi-automatic signal segmentation had been applied to isolate flexion and extension cycles, features based on relative high-frequency components were generated. Using machine learning with a linear support vector machine, these signals were classified as healthy, exhibiting chondromalacia °II-IV or osteoarthritis. 84% of healthy subjects were identified correctly, while the classification accuracy for individual stages of chondromalacia or osteoarthritis ranged from 11% (CM °II) to 50% (CM °III). In order to make results easily interpretable without resorting to machine learning techniques, we propose a normalized score between 0 and 1 and show that this "v-score" for flexion and extension significantly correlates with the achieved multi-class classification. Vibroarthrography may qualify as potent screening tool for the detection and grading of joint cartilage defects and aid physicians in the choice and estimation of urgency of further diagnostic and therapeutic decisions. Health Sciences EPiC Series in Health Sciences Volume 2, 2018, Pages 6–11 CAOS 2018. The 18th Annual Meeting of the International Society for Computer Assisted Orthopaedic Surgery W. Zhan and F. Rodriguez Y Baena (eds.), CAOS 2018 (EPiC Series in Health Sciences, vol. 2), pp. 6–11