Machine learning-based bioimpedance assessment of knee osteoarthritis severity

Area of Science:

• Biomedical Engineering
• Medical Diagnostics
• Osteoarthritis Research

Background:

• Knee osteoarthritis (KOA) is a prevalent degenerative joint disease.
• Accurate and accessible diagnostic tools for KOA severity are needed.

Purpose of the Study:

• To develop and validate a multiclass model for classifying knee osteoarthritis (KOA) severity.
• To assess the efficacy of bioimpedance measurements in conjunction with specific exercises for KOA grading.

Main Methods:

• Utilized bioimpedance measurements with an 8-electrode setup and an AD5933-based device.
• Collected data from 37 participants (25 healthy, 13 with KOA) performing five knee tasks.
• Trained a multiclass random forest model on 15 experimental setups (3 measurement types x 5 exercises).

Main Results:

• Achieved 100% average sensitivity and specificity for all four KOA severity grades.
• Demonstrated high accuracy during knee extension, cyclic motion, and gait tasks.
• The model effectively distinguished between healthy knees and different KOA grades.

Conclusions:

• Bioimpedance measurements combined with specific physical tasks offer a promising, non-invasive method for KOA severity classification.
• The proposed approach can function as an effective screening tool, identifying individuals requiring further imaging like X-rays or MRI.
• This technology could streamline the diagnostic pathway for knee osteoarthritis.