Acute Knee Injury Detection with Magnetic Resonance Imaging (MRI)

• Published in: International journal of computers, communications & control communications & control, 2024-10, Vol.19 (5)
• Main Authors: Mahmood, Mahmood A., Alsalem, Khalaf, Elbashir, Murtada, Abd El-Ghany, Sameh, Abd El-Aziz, A.
• Format: Article
• Language: English
• Subjects: Classification; Data augmentation; Datasets; Deep learning; Genetic algorithms; Injury prevention; Knee; Ligaments; Machine learning; Magnetic resonance imaging
• ISSN: 1841-9836; 1841-9844
• DOI: 10.15837/ijccc.2024.5.6648

The anterior cruciate ligament (ACL) is a major ligament in the knee that helps to stabilize the joint and prevent excessive forward movement of the shinbone. An ACL tear is a common injury, especially among athletes who participate in sports that involve pivoting and sudden changes in direction. This paper proposes an ensemble model, which includes three deep learning models (EfficientNet-B7, ResNet-152V2, and DenseNet-201) and a genetic algorithm, to detect and classify ACL tears using knee magnetic resonance imaging (MRI). The ensemble model was trained on the KneeMRI dataset, which comprises labeled MRI images. The deep learning models can learn to identify subtle changes in ligament structure and signal intensity that are associated with ACL tears and the genetic algorithm is used to find the optimal prediction. The proposed ensemble model was evaluated using the KneeMRI dataset. The dataset was preprocessed using data augmentation techniques. Then, the ensemble model was applied to the KneeMRI dataset, evaluated, and compared with previous models. The accuracy, recall, precision, specificity, and F1 score of our proposed ensemble model were 99.68%, 98.73%, 99.52%, 99.62%, and 98.94%, respectively. Thus, our ensemble model has an unrivaled perceptive outcome and could be used to accurately identify and classify ACL tears, improving patient outcome.