Non-invasive Knee Osteoarthritis Diagnosis via Vibroarthrographic Signal Analysis

Crepitus is a prominent sign during the knee joint undergoing a range of motion which can be detected to form the vibroarthrographic (VAG) signals. These signals can be used as useful indicators of osteoarthritis (OA) status for supplementing conventional X-ray imaging in the diagnosis of knee OA. In this study, a non-invasive knee OA diagnosis system was conducted via VAG signal analysis.This system included a goniometer to provide an analog reference signal for positioning, an electronic stethoscope to detect knee VAG signals and an FPGA signal processing system as a system kernel. Power spectra of the signals provided by the Fourier transform were obtained and partition indices calculated. Discriminant functions were built with distribution parameters constructed from the partition indices. The processed VAG signals were compared with X-ray images in OA diagnosis to explore the differences between predicted and observed results. System model performance was evaluated using measures for discriminative ability, including the area under the receiver operating characteristic curve (AUC). The negative predictive value (NPV) was checked for the rate of correctly predicting the absence of OA. The positive predictive value (PPV) is the proportion of patients with OA who are correctly diagnosed.System performance was validated using the receiver operating characteristic curve. The classification result illustrated a sensitivity of 89.52% and a specificity of 67.50%, with a total accuracy rate of 81.52%. The AUC values obtained were 0.68 (95% CI 0.61-0.74). The rates of correctly predicting the lack of OA were approximately 78.6% (NPV) and the proportions of patients with OA who are correctly diagnosed were approximately 82.8% (PPV), respectively.VAG analysis in OA diagnosis provides an economic alternative to X-ray examination in osteoarthritic patients.