Anterior Cruciate Ligament (ACL) tears are prevalent injuries in sports and physical activities that necessitate prompt and precise diagnosis for optimal treatment and re-habilitation. Conventional diagnostic techniques like physical examination and MRI, may be subjective and protracted. This study proposes a new efficient technique for detecting tears of ACL based on the integration of a Convolutional Neural Network (CNN) and an improved version of Human Evolutionary Algorithm (IHEA). The purpose of the suggested IHEA is to enhance the hyperparameters of the CNN to improve its performance in detecting ACL rupture from MRI scans. The suggested technique has been validated by assessing it on a standard case study and comparing its results with some other advanced methods, including the Convolutional Neural Network (CNN), Generative Adversarial Network (GAN), Generative Adversarial Network (GAN2), Gated Recurrent Unit combined with Flexible Fitness Dependent Optimizer (GRU/FFDO), and GRU optimized by Hybrid Tasmanian Devil Optimization (GRU/HTDO). Final results showed the superiority of the proposed model in diagnosing of the ACL tear.