Machine Learning Based Multi-Class Classification and Grading of Squamous Cell Carcinoma in Optical Microscopy

Abstract

Histopathological tissue grading is critical for disease diagnosis and treatment, but manual grading is labor-intensive and time-consuming, requiring expert pathologists. This study presents an efficient analysis of squamous cell carcinoma (SCC) histopathological images using machine learning (ML) and deep learning (DL) models. Five ML models—support vector machine, Naïve Bayes, decision tree, k-nearest neighbor (KNN), and neural network—were trained with 5-, 7-, and 10-fold cross-validation. Discrete wavelet transform along with gray level co-occurrence matrix and histogram features extracted 360 features per image, and Student's t-test selected 114 key features. Among ML models, KNN with sevenfold cross-validation achieved 98% accuracy. Additionally, a convolutional neural network (CNN) trained achieved 98.23% accuracy in automated classification. These results suggest that combining ML for feature analysis with interpretable DL models can lead to more accurate and efficient SCC grading, reducing reliance on manual pathology.