<p>Heart disease is a leading cause of death globally, and continuous efforts are being made to treat this challenging condition. One of the promising combats that could advantage many patients is stent placement, which is a small mesh tube inserted into coronary arteries to maintain blood flow in patients with heart disease. Accurately deciding whether a stent is needed from coronary angiogram datasets is challenging due to the complex nature process of decision making. Although many metaheuristic and optimization algorithms have been proposed in this determination process, their outcomes have been less than encouraging in identifying stent necessity. One of the primary objectives of this study is establishing a significant indicator for stent necessity in hearth disease infected by analyzing an image dataset that includes stent and non-stent cases. Moreover, the study aimed to enhance the accuracy of stent necessity assessments in these patients using ResNet-50 for feature extraction integrating with six optimization algorithms. Five evaluation metrics: accuracy, specificity, sensitivity, precision, and the confusion matrix were utilized to assess the performance of these methods. Among those algorithms, Learner Performance-Based Behavior with Simulated Annealing (LPBSA), when integrated with ResNet-50 in this study and referred to as LPBSA-(ResNet-50), obtained 100% across the evaluation metrics. In comparison, particle swarm optimization with ResNet-50, denoted as PSO-(ResNet-50), demonstrated competitive performance throughout the evaluation process, except the sensitivity metric, obtaining 75%.</p>