Colorectal cancer is one of the most common cancers and a leading cause of mortality, particularly among men. Despite therapeutic advancements, treatment resistance, particularly to FOLFOX-based regimens, cetuximab, and panitumumab, remains a major challenge. This study proposes an innovative approach using Deep Learning to classify resistant tumors and predict molecular mutations from histopathological images. We analyzed existing models to develop a Deep Learning framework capable of identifying resistance-associated mutations, such as those of the KRAS gene, from annotated medical images. These images enable the correlation of morphological features with specific mutations, offering a cost-effective alternative to expensive genetic tests. Preliminary results show a significant improvement in accuracy compared to traditional methods. In the Moroccan context, where genetic resources are limited, this approach could revolutionize colorectal cancer diagnosis and treatment. By integrating artificial intelligence into medical practice, we aim to provide a precise and accessible tool for personalized medicine, thereby improving clinical outcomes for patients. This research highlights the transformative potential of Deep Learning in colorectal cancer management and calls for collaboration to create a local image database, bridging the gap in research specific to the Moroccan population. Our work paves the way for promising clinical applications for patients resistant to conventional treatments.