C2DEEP-OT: Utilizing Multi-Agent Deep Reinforcement Learning Algorithm and Optimized Attentive Transformer Network for Cervical Cancer Detection

Cervical cancer (CC) is the major common cancers among women, and detecting earlier critical for successful treatment. Traditional methods, includes as Pap smear tests, are highly contagious to manual error which paves the way for Artificial Intelligence (AI) solutions for improved detection. Whereas the conventional AI enabled models faced with poor reliability and accuracy respectively. In order to overcome the issue mentioned, this research develops AI enabled model named C2DEEP-OT which is coined as Cervical Cancer Detection through Deep Reinforcement Learning and Optimized Transformers. Our models employ coloscopy and histopathology images for diagnosing the cervical cancer for enabling normalization and noise removal. After that, major features were extracted from Multi Agent Deep Reinforcement Learning (MA-DRL) named Enhanced Deep Q-network (EDQN) that effectively manage the color, contextual, and spectral, and spatial information with better accuracy. In parallel, the extracted features are then provided to the Optimized Attention based Transformer (OAT) which is improved by Rat Swarm Optimization (RSO) for categorize cervical cancer in accurate manner into three classes includes malignant, benign, and normal. From the results, it is seen that C2DEEP-OT gains 98.63% of accuracy which superiors state of the art models.