Classification of Oral Cancer into Pre-Cancerous Stages from White Light Images Using LightGBM Algorithm

This study presents a novel approach for classifying oral cancer into pre-cancerous stages using white light images and the Light Gradient Boosting Machine (LightGBM) algorithm. Initially, input images are resized to 264 × 264 pixels and transformed from the Red Green Blue (RGB) color space into four alternative color spaces: Hue Saturation Value (HSV), Hue Saturation Lightness (HLS), YCbCr, and XYZ. This transformation facilitates a comprehensive analysis of color information, which is crucial for accurate classification. Feature extraction is conducted to obtain both color and texture features that are pivotal in distinguishing between various pre-cancerous conditions. The extracted features are then utilized to train the LightGBM model, known for its efficiency and high performance in handling large datasets. The model is evaluated for its ability to classify three specific pre-cancerous stages: leukoplakia, erythroleukoplakia, and erythroplakia. The results demonstrate that the proposed methodology effectively enhances the accuracy of oral cancer diagnosis by providing a robust and efficient classification system, thereby contributing to improved clinical decision-making in the management of oral health.

Index Terms— Binary and multi-class classification, color spaces, early detection, feature importance, oral cancer, white light images.