Enhanced CaffeNet/SVM model for knee meniscus tear classification using modified Gooseneck barnacle strategy

Early diagnosis of knee meniscus tears is important for timely intervention, reduced recovery time, and avoidance of chronic problems. However, MRI-based medical imaging is the best method to visualize multiple planes to identify such tears. Based on this, machine learning and artificial intelligence methods have recently gained enormous traction for meniscus tear detection using MR images. This study presents a new methodology involving deep learning and metaheuristic optimization for the accurate and automated detection of knee meniscus tears. Our approach consists of a high-level feature extraction with CaffeNet and a classification task with a support vector machine. To improve performance, we present an advanced metaheuristic algorithm based on fractional theory for feature extraction and classification, called the Modified Gooseneck Barnacle Optimizer. We evaluate our approach on a publicly available dataset of knee MRI images, and experimental results reveal that our methodology outperforms the state of the art in the detection of knee meniscus tears.