Intellectual Property India Publishes Patent Application for An Intelligent Artificial Intelligence-Based System And Method For Early Oral Squamous Cell Carcinoma Lesion Detection And Localization Using Yolov8 Deep Learning Analysis

MUMBAI, India, Sept. 22 -- Intellectual Property India has published a patent application (202641109855 A) filed by Vallurupalli Nageswara Rao Vignana Jyothi Institute Of Engineering And Technology on September 13, 2026, for An Intelligent Artificial Intelligence-Based System And Method For Early Oral Squamous Cell Carcinoma Lesion Detection And Localization Using Yolov8 Deep Learning Analysis.

Inventor includes Mrs. R. Jyothsna.

The application for the patent was published on September 18, 2026, under issue no. 38/2026.

Abstract: ABSTRACT [0020] The present invention provides a comprehensive deep learning framework for the automated classification and localization of Oral Squamous Cell Carcinoma from histopathological images. The system evaluates and compares multiple model families including convolutional neural networks such as ResNet50, EfficientNet-B0 and DenseNet121, vision transformers including the Vision Transformer and Swin Transformer, and hybrid architectures that integrate convolutional and attention mechanisms such as ConvNeXt and CvT-13. After uniform preprocessing and hyperparameter tuning on a publicly available OSCC histopathological dataset, the CvT-13 hybrid model achieved the highest classification accuracy of 94.91 percent, outperforming the other evaluated architectures. Following classification of an image as cancerous, Gradient-weighted Class Activation Mapping generates heatmaps that highlight the regions most influential to the model decision. These heatmaps are further refined with OpenCV-based thresholding and contour detection to produce clear visual outlines of suspected areas. Bounding boxes derived from the refined maps are then used to train a YOLO object detection model for real-time localization of cancerous regions. The integrated pipeline delivers accurate classification together with interpretable visual explanations, improving generalization, scalability and clinical utility for early OSCC diagnosis while reducing reliance on purely manual histopathological review.

Disclaimer: Curated by HT Syndication.