Classification of knee arthropathy with accelerometer-based vibroarthrography.

Area of Science:

• Biomedical Engineering
• Orthopedics
• Signal Processing

Background:

• Osteoarthritis is a common knee joint disorder causing progressive cartilage and bone degeneration, primarily affecting the elderly.
• Current diagnostic methods for knee osteoarthritis are often complex, expensive, invasive, or lack sensitivity for early changes.
• Vibroarthrography offers a novel, non-invasive approach by analyzing mechanical vibratory signals from the knee.

Purpose of the Study:

• To develop and evaluate a vibration-based classification system for differentiating healthy knee joints from those with osteoarthritis.
• To assess the efficacy of machine learning classifiers, specifically k-nearest neighbors (KNN) and support vector machines (SVM), using vibroarthrographic data.
• To identify key time-frequency features for accurate classification of knee joint conditions.

Main Methods:

• Collected knee joint vibration signals from 19 healthy and 20 osteoarthritis volunteers.
• Utilized a dataset comprising 92 healthy and 120 arthritic knee joint signal segments.
• Employed k-nearest neighbors and support vector machine classifiers, evaluating performance based on time-frequency features.

Main Results:

• The k-nearest neighbors classifier achieved the highest accuracy of 89.8% using only 6 time-frequency features.
• Achieved a precision of 88.3%, recall of 92.4%, and an f-measure of 90.1% with the KNN classifier.
• Demonstrated that vibroarthrography can effectively differentiate between healthy and arthritic knee joints.

Conclusions:

• Vibroarthrography presents a promising, non-invasive, and low-cost tool for screening knee joint disorders like osteoarthritis.
• The study highlights the potential of using a limited set of time-frequency features with KNN for accurate classification.
• Further refinements in data collection and analysis could enhance the capabilities of vibroarthrography for clinical application.