Abstract

Mouth cancer remains a significant health concern, and early recognition of symptoms is essential for improving treatment outcomes and strengthening public awareness. This research presents an integrated approach that combines medical image analysis with numerical clinical data to support more accurate mouth cancer detection. Image samples were first enhanced using a mean filter to reduce noise and highlight suspicious regions, allowing clearer segmentation of the tumor-affected Region of Interest (ROI). Two deep learning models, Artificial Neural Network (ANN) and Convolutional Neural Network (CNN), were applied to analyse both image-based features and patient numerical attributes. CNN was used for segmentation and visual classification, while ANN processed symptom patterns and clinical indicators such as pain, persistent mouth sores, abnormal tissue patches, and difficulty in chewing or swallowing. Both models were also evaluated on a combined dataset to predict cancer stage and estimate risk levels. The fused analysis provided higher diagnostic accuracy than individual data sources, demonstrating the value of multimodal learning. This study emphasises the importance of early symptom identification, increased public health awareness, and prompt medical treatment supported by advanced computational techniques.