In order to enhance the diagnostic accuracy, feature selection for colorectal cancer identification attempts to extract the most insightful features from medical images. This study presents a new method for detecting colorectal cancer using hybrid method that combines Ant Lion Optimizer (ALO) with mutual information for feature selection and classification based on neural networks. To standardize and improve the dataset, the process starts with preprocessing medical images, which includes resizing, normalization, and data augmentation. Local Binary Patterns (LBP) are used for textural feature extraction as it can enclose significant amount of textural information pertinent to cancerous tissues. The ALO, enhanced with mutual information, is used to choose the most informative features through assessing their relevance to cancer classification. The optimized feature set is then utilized by a neural network to perform the classification. The integrated approach capitalizes on the advantages of metaheuristic optimization and modern machine learning technologies that collectively lead to an improved efficacy compared with existing methods for colorectal cancer detection. Experimental results with respect to significant features and classification performance validate the effectiveness of our approach, suggesting its potential in improving diagnostic accuracy using medical imaging.