MUMBAI, India, July 7 -- Intellectual Property India has published a patent application (202641079476 A) filed by P Venkata Krishna on June 28, 2026, for Hybrid Graph-Based Structural Features With Histogram Of Oriented Gradients (hog) For Interpretable Classification Of Colon Histopathology Images.
Inventors include M. Siva Parvathi; T. Sudha; Shaheen Begum; N. V. Muthu Lakshmi; T. Sukeerthi; P. V. Anand Krishna; Bhavani Kumar Yellapragada; and P Venkata Krishna.
The application for the patent was published on July 03, 2026, under issue no. 27/2026.
Abstract: Reliable classification of colon histopathology images is essential for early detection and effective management of colorectal cancer. Manual microscopic examination is labor-intensive and prone to observer variability. This study presents an interpretable hybrid computational framework that integrates Histogram of Oriented Gradients (HOG) with graph-theoretic structural descriptors derived from nuclear spatial organization. Multiple graph constructions, including Delaunay triangulation, Minimum Spanning Tree, k- Nearest Neighbor, and Voronoi diagrams, are systematically investigated. Structural and texture features are fused and classified using a Random Forest model. Experiments conducted on an augmented dataset containing approximately 5,000 images per class demonstrate consistent improvements in accuracy, stability, and cancer detection sensitivity compared to individual descriptors. The proposed framework achieves a favorable balance between predictive performance, interpretability, and computational efficiency, supporting its potential application in routine clinical workflows.
Disclaimer: Curated by HT Syndication.
