Automated weight-bearing foot measurements using an artificial intelligence-based software.

Area of Science:

• Radiology
• Artificial Intelligence
• Medical Imaging Analysis

Background:

• Weight-bearing foot radiographs are crucial for diagnosing conditions like hallux valgus and assessing foot alignment.
• Accurate measurements of specific angles are essential for diagnosis and treatment planning.
• Manual measurements can be time-consuming and subject to inter-observer variability.

Purpose of the Study:

• To evaluate the accuracy of the BoneMetrics (Gleamer) artificial intelligence (AI) software.
• To assess the AI's capability for automated angle measurements on weight-bearing forefoot and lateral foot radiographs.
• To compare AI measurements against a ground truth established by expert radiologists.

Main Methods:

• Retrospective collection of weight-bearing forefoot and lateral foot radiographs from three institutions.
• Independent annotation of key points by two senior musculoskeletal radiologists to define ground truth measurements.
• Statistical analysis using Mean Absolute Error (MAE), Bland-Altman analysis for bias, and Intraclass Coefficient (ICC) for reliability.

Main Results:

• Low Mean Absolute Error (MAE) and bias were observed for all measured angles between AI predictions and ground truth.
• Excellent Intraclass Coefficients (ICC) were found for most measurements, indicating high agreement.
• Specific MAE values ranged from 0.7° to 1.2° for forefoot angles and 1° to 3.9° for lateral foot angles.

Conclusions:

• The AI software demonstrated significant potential for accurate and automated measurements on weight-bearing foot radiographs.
• AI-driven analysis shows promise in improving efficiency and consistency in foot imaging assessments.
• Further validation may support the integration of AI tools into routine clinical practice for foot radiography.