Enhanced Skill Optimization Algorithm and Stacked Long Short-Term Memory with Sech Activation Function for Gastrointestinal Disease

Gastrointestinal (GI) disease significantly affects the quality of human life, leading to an increased death rate. Deep learning (DL) is an emerging tool that helps analyze medical images for disease diagnosis, prediction, and treatment planning. However, existing DL methods fail to accurately detect rare GI diseases due to insufficient feature representation, irrelevant feature selection, and high positive rates, which limit diagnostic reliability. This research proposes the Enhanced Skill Optimization Algorithm and Stacked Long Short-Term Memory with Sech Activation Function to efficiently select the relevant features and capture complex patterns in GI disease diagnosis. In traditional SOA, the Gorilla Troop Optimizer is used as an enhanced strategy in the exploration phase to select relevant features by diversifying the search space, which prevents premature convergence, minimizes dimensionality, and enhances generalization. The Stacked LSTM enhances the model’s ability to capture intricate temporal dependencies, whereas the Sech activation function provides smooth gradient flow that minimizes vanishing and exploding gradients during training. Advanced preprocessing and deep feature extraction methods are used to enhance the model’s ability to accurately capture significant patterns. Hence, the proposed method achieves superior accuracies of 99.60% and 99.88% on the Kvasir V1 and V2 dataset, respectively, compared with existing methods such as SK-Net, demonstrating high potential for accurate GI diagnosis and offering a robust solution for clinical image analysis.