Predicting Severe Knee Arthritis Based on Two Inertial Measurement Unit Sensors as a Dynamic Coordinate System Using...

Area of Science:

• Biomedical Engineering
• Orthopedics
• Wearable Technology

Background:

• Musculoskeletal disorders, particularly knee osteoarthritis (KOA), pose a significant healthcare challenge due to aging populations.
• Current KOA diagnosis relies on radiographs, which involve ionizing radiation and have drawbacks.
• There is a need for non-invasive, low-cost diagnostic methods for KOA.

Purpose of the Study:

• To evaluate the efficacy of a wearable device in distinguishing between healthy individuals and those with severe KOA (grade 4).
• To explore the potential of wearable sensor data for KOA diagnosis.

Main Methods:

• A wearable device with two inertial measurement unit (IMU) sensors (thigh and lower leg) was utilized.
• New features were extracted in a dynamic coordinate system from 1433 IMU signals from 15 healthy and 15 severe KOA individuals (aged >45).
• Four classifiers (naive Bayes, KNN, SVM, random forest) were evaluated using 10-fold cross-validation.

Main Results:

• The K-nearest neighbors (KNN) classifier achieved the highest accuracy at 93.71 ± 1.1%.
• KNN also demonstrated high precision at 93 ± 1.31%.
• The proposed algorithm showed improved sensitivity compared to existing methods.

Conclusions:

• Novel features derived from a dynamic coordinate system and the KNN model effectively diagnose between healthy individuals and KOA patients.
• The wearable device shows potential as an auxiliary tool for arthritis diagnosis.
• Further validation is needed, as results are specific to severe KOA (grade 4) and may differ for other grades.