Automatic Classification of Anterior Talofibular Ligament Based on 2D Convolutional Neural Network

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Orthopedics

Background:

• Anterior talofibular ligament (ATFL) injuries are common, necessitating accurate diagnostic methods.
• Magnetic Resonance Imaging (MRI) is crucial for visualizing ATFL tears.
• Manual analysis of MR images can be time-consuming and subject to inter-observer variability.

Purpose of the Study:

• To assess the feasibility of 2D Convolutional Neural Networks (CNNs) for automated ATFL classification on MR images.
• To develop and validate AI models for both ATFL segmentation and tear classification.

Main Methods:

• T2-weighted MR images from multiple centers were used to train and validate a YOLOv11 segmentation model.
• A Dice Similarity Coefficient (DSC) was employed to quantify segmentation accuracy.
• A 2D ResNet model was trained to classify ATFL tears (normal, partial, total) on segmented images.
• The integrated workflow was tested on an independent dataset.

Main Results:

• The YOLOv11 segmentation model achieved a median DSC of 0.95.
• The automated workflow demonstrated 92.6% overall accuracy on the test dataset.
• Sensitivity and specificity for abnormal ATFL detection were 95.0% and 93.6%, respectively.
• Automated classification was significantly faster than manual review by a junior radiologist.

Conclusions:

• CNN-based automatic segmentation and classification of ATFL on MR images are feasible.
• This AI approach shows potential for efficient and accurate ATFL injury evaluation.
• The developed models offer a promising tool for augmenting radiological assessment of ATFL injuries.