Portal de Periódicos da CAPES

Anterior cruciate ligament tear detection based on deep belief networks and improved honey badger algorithm

2023; Elsevier BV; Volume: 84; Linguagem: Inglês

10.1016/j.bspc.2023.105019

1746-8108

Junjie Sun, Lijuan Wang, Navid Razmjooy,

Infrared Thermography in Medicine

The Anterior Cruciate Ligament (ACL) tear is a common injury among athletes who participate in extreme sports such as basketball, football, American football, and skiing. When an ACL tear is suspected, doctors usually take X-Rays of the patient's knee to identify the injury. MRI can often be used to help with diagnosis. This study proposes a novel hierarchical approach for more accurate ACL injury detection. The method starts by applying preprocessing techniques to improve image quality, then using Co-occurrence Matrix (GLCM) and Discrete Cosine Transform (DCT) in combination, and features from the images are retrieved. The features are then sent into a Deep Belief Network (DBN) which has been trained for classification and is further optimized using a new metaheuristic method known as the “Improved Honey Badger Algorithm”. Results are compared with methods like Euclidean Distance and Neural Networks (ED/NN), Random Forest (RF), Fuzzy and Convolutional Neural Networks (CNN) and it is seen that the proposed method achieves 96% accuracy, 98% sensitivity, and 80% specificity, proving highest efficiency than all other methods.