Knee Meniscus Tear During Exercise Identification Using Capsule Neural Network Optimized by Adapted Seagull Optimizer

Accurate, timely diagnosis of knee meniscus tear remains a clinical challenge due to the limitations of traditional diagnostic methods and the suboptimal performance of conventional deep learning models in capturing spatial and pose information from MRI images. To address these gaps, this study proposes a novel diagnostic model that combines the Capsule Neural Network (CapsNet) with an Adapted Seagull Optimization (ASO) algorithm, forming an integrated framework for enhanced feature extraction and automated hyperparameter tuning. Unlike existing methods—such as classic deep learning and convolutional neural networks—which often rely on manual parameter selection and are prone to local minima, our approach leverages the global search capability of ASO to optimize CapsNet configurations more effectively. The model is rigorously evaluated on a curated clinical MRI dataset and benchmarked against state-of-the-art techniques to demonstrate its superior accuracy and generalization. Results indicate that the proposed CapsNet/ASO model offers a robust decision-support tool for clinicians, substantially improving diagnostic performance over existing approaches.