Artificial intelligence to diagnose meniscus tears on MRI.

Area of Science:

• Orthopedic imaging
• Artificial intelligence in medicine
• Medical diagnostics

Background:

• Meniscus tears are common knee injuries.
• Accurate detection on MRI is crucial for treatment.
• Current diagnostic methods can be time-consuming.

Purpose of the Study:

• To develop and assess a high-performance algorithm for detecting and characterizing meniscus tears using knee MRI.
• To improve the accuracy and efficiency of meniscus tear diagnosis.

Main Methods:

• A fast-region convolutional neural network (CNN) algorithm was trained on 1123 knee MRI images.
• The algorithm performed three tasks: horn detection, tear presence detection, and tear orientation determination.
• External validation was conducted on 700 images, with performance measured by area under the curve (AUC).

Main Results:

• The algorithm achieved an AUC of 0.92 for horn detection, 0.94 for tear presence, and 0.83 for tear orientation.
• A final weighted AUC of 0.90 was obtained, demonstrating high overall performance.
• The developed algorithm shows significant accuracy in identifying key features of meniscus tears.

Conclusions:

• The fast-region CNN algorithm effectively detects meniscus tears on knee MRI.
• This represents a significant advancement towards AI-powered diagnostic tools in orthopedics.
• The algorithm's performance suggests its potential utility in clinical settings.