SGDWOA: A Novel Approach in Whale Optimisation for Accurate Cell Classification in Oral Squamous Cell Carcinoma Using Machine Learning

Abstract

Oral squamous cell carcinoma (OSCC), a common head and neck cancer, is often unnoticed but can be identified early. Diagnosing this heterogeneous tumour requires extensive human experience, and artificial intelligence can help to improve diagnosis. This study used novel methodologies based on feature selection and classification in an attempt to obtain good findings for the early detection of OSCC. By using cutting-edge hybrid strategies to extract features and improve classification, this work seeks to bridge the gap among deep learning and machine learning procedures. Initially, preprocessing is done to address artifacts in the OSCC dataset. The first method uses SMOTE oversampling and feature scaling in conjunction with Resnet 50 and Efficientnet B5 models for feature extraction. In the second method, the best feature set is chosen using the Statistic gain Dynamic Remodelled Whale Optimization Algorithm (SDRWOA), and the Random Forest Classifier is then employed to classify cancer types into poor, moderate, and well categories. The finding shows that the proposed model beats the other classifiers by attaining the maximum overall accuracy, recall and F1-score of 98% and precision of 97.6%. In conclusion, the suggested approach advances the development of extremely precise and effective OSCC diagnosis techniques.

Keywords

• oral cancer
• oral squamous cell carcinoma
• histopathological images
• AI-based system
• machine learning
• data preprocessing