Development of convolutional neural network model for diagnosing meniscus tear using magnetic resonance image

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Orthopedics

Background:

• Deep learning (DL) is an advanced machine learning technique increasingly applied in medical image analysis.
• Convolutional Neural Networks (CNNs) are a type of DL model particularly effective for image recognition and classification tasks.
• Meniscal tears are common knee injuries requiring accurate diagnosis and classification.

Purpose of the Study:

• To develop a CNN model for detecting the presence of meniscal tears using knee MRI.
• To develop a CNN model for classifying different types of meniscal tears.
• To evaluate the diagnostic performance of the developed CNN models.

Main Methods:

• Retrospective collection of 1048 knee MRI cases (599 with meniscal tears, 449 without).
• Development of two CNN models: one for tear presence detection and another for tear type classification.
• Utilized a dataset including horizontal, complex, radial, and longitudinal tear types for classification model development.
• Data was split into 70% for training and 30% for testing to measure model performance.

Main Results:

• The CNN model achieved high performance in detecting medial meniscal tears (AUC 0.889), lateral meniscal tears (AUC 0.817), and combined tears (AUC 0.924).
• For tear type classification, the model demonstrated strong performance for complex (AUC 0.850) and longitudinal tears (AUC 0.858), with moderate performance for horizontal tears (AUC 0.761) and lower performance for radial tears (AUC 0.601).

Conclusions:

• The developed CNN model demonstrates significant potential for the accurate detection of meniscal tears.
• The CNN model shows capability in differentiating between various types of meniscal tears.
• This AI-driven approach could aid clinicians in diagnosing and classifying meniscal tears more effectively.