A Torn ACL Mapping in Knee MRI Images Using Deep Convolution Neural Network with Inception-v3.

Area of Science:

• Orthopedics and Sports Medicine
• Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Anterior Cruciate Ligament (ACL) injuries are common knee ligament damage, often occurring during sports activities.
• Magnetic Resonance Imaging (MRI) is crucial for diagnosing knee abnormalities, including ACL tears and meniscal tears.
• Accurate and timely diagnosis of ACL tears is essential for effective patient management and treatment.

Purpose of the Study:

• To develop and evaluate a Deep Convolutional Neural Network (DCNN) model for detecting ACL tears using MRI knee images.
• To leverage deep transfer learning (DTL) with the Inception-v3 architecture for enhanced classification accuracy.
• To assess the model's performance against other established deep learning models.

Main Methods:

• Utilized the MRNet database comprising 1,370 knee MRI images.
• Employed a DCNN-based Inception-v3 DTL model for image classification.
• Performed data preprocessing, feature extraction, and classification.
• Compared the proposed model's performance with VGG16, VGG19, Xception, and Inception ResNet-v28.

Main Results:

• The Inception-v3 DTL model achieved a training accuracy of 99.04%.
• The model demonstrated a testing accuracy of 95.42% in performance analysis.
• Evaluated performance using metrics such as accuracy, precision, recall, specificity, and F-measure.

Conclusions:

• The proposed DCNN with Inception-v3 DTL model effectively detects ACL tears from knee MRI images.
• This AI-driven approach shows significant potential for improving the diagnosis of knee abnormalities.
• The model's high accuracy suggests its utility in clinical settings for aiding orthopedic diagnoses.