Medical image analysis is important in disease diagnosis and treatment planning, demands highly precise and effective computational models. The current research assesses the performance of multi-headed models like CNN, RNN and Transformers, also we included algorithms like Transformers, RNNs, and LightGBM. CNN is used for extraction of spatial features, RNN is used for sequential dependencies, Transformers are used for long-range dependencies and LightGBM is used to improve decision making with gradient boosting. Experimental results show that integration of Transformers, RNN and lightGBM has achieved highest accuracy of 94% when compared with traditional deep learning models. These findings highlight the potential of multi-headed architectures in advancing for text classification.