Hanging protocol optimization of lumbar spine radiographs with machine learning

Area of Science:

• Radiology
• Artificial Intelligence
• Medical Imaging

Background:

• Lumbar spine radiograph interpretation requires precise positioning and orientation.
• Optimizing radiograph hanging protocols is crucial for efficient workflow and accurate diagnosis.
• Current methods for radiograph analysis can be time-consuming and prone to human error.

Purpose of the Study:

• To evaluate the efficacy of machine learning (ML) algorithms in automating lumbar spine radiograph analysis.
• To determine if ML can accurately label radiograph views, detect orthopedic hardware, and correct image rotation.

Main Methods:

• A dataset of 6988 lumbar spine radiographs from 1727 patients was analyzed.
• Radiograph views, hardware presence, dynamic positions, and correctional rotation were manually annotated.
• ML models were trained to classify views, detect hardware, and predict rotational correction degrees.

Main Results:

• Categorical ML models achieved high accuracy with Area Under the Curve (AUC) values ranging from 0.985 to 1.000.
• Rotation correction models demonstrated effectiveness, with mean absolute differences as low as 0.610 degrees in specific datasets.

Conclusions:

• Machine learning algorithms show significant potential for optimizing lumbar spine x-ray hanging protocols.
• ML can accurately automate key aspects of radiograph analysis, including view labeling, hardware detection, and rotation correction.
• Implementation of these ML techniques can enhance efficiency and accuracy in diagnostic imaging.