Early Stage Oral Cancer Detection using Machine Learning and Deep Learning Algorithms

Oral cancer remains a significant global health challenge with high mortality rates. The major cause is late recognition of oral cancer, making treatment less effective. This condition is particularly prevalent in low- and middle-income countries. Early-stage detection and accurate diagnosis are critical for improving treatment outcomes, reducing deaths, and minimizing recurrence. Enabling automation in the identification of malignant cells in the oral cavity could lead to low-cost and early diagnosis of the disease. This study aims to detect early cancer stages accurately using machine learning and Deep learning techniques. A histopathological dataset with cellular-level images of different regions in the oral cavity helps in detecting cancer cells more effectively. Preprocessing and image enhancement techniques are used to leverage spatial and temporal features from imaging data to distinguish damaged cancer cells from normal cells. Wavelet Decomposition and edge detection techniques play a crucial role in identifying damaged regions in images. The ensembling technique with the voting approach independently combines multiple CNN models and binary classifiers providing the highest accuracy of 98% shown in the results section. Ensembling approach produces significant improvements in cancer cell identification and enhancing system robustness.