A sin laryngeal carcinoma is the most common kind of head and neck cancer to damage the soft tissues of the larynx. To prevent further medical difficulties and to provide better patient care, early stage laryngeal cancer identification is essential. The primary objective of this research is to provide computer-aided cancer diagnosis driven by deep learning. In order to do this, we create a brand-new Optimization-Based Cascaded Deep Neural Network (OCDNN) that distinguishes between good and unhealthy/cancerous tissues. Additionally, we also provide cascaded patch-based CNN (CP-CNN) picture segmentation. The models are assessed using accepted measures that are unique to the biomedical field. When compared to other models, the performance of our suggested OCDNN architecture was exceptional. Additionally, it fared better than other cutting-edge CNN models applied to medical domain challenges. The findings suggest that the suggested method is quite successful in identifying early indications of laryngeal cancer. The results demonstrate that the suggested strategy may help doctors identify laryngeal carcinoma early and precisely. The suggested OCDNN is discovered to attain 98.4% accuracy, 94% precision, 97.5% recall, and 98% F1-score.gle paragraph of about 100 words to give a brief introduction to your work.