MUMBAI, India, May 1 -- Intellectual Property India has published a patent application (202641051519 A) filed by Srinivasa Ramanujan Institute Of Technology; Mrs. D. Jahnavi; V. Meghana; A. Asha Latha; and D. Akhila, Ananthapuramu, Andhra Pradesh, on April 22, for 'texture driven hybrid deep learning approach for lung nodule classification using ct images.'
Inventor(s) include Srinivasa Ramanujan Institute Technology; Mrs. D. Jahnavi; V. Meghana; A. Asha Latha; and D. Akhila.
The application for the patent was published on May 1, under issue no. 18/2026.
According to the abstract released by the Intellectual Property India: "The primary aim of this invention is to develop an automated and intelligent system for the detection and classification of lung nodules using computed tomography (CT) images. The invention focuses on designing a texture driven hybrid deep learning framework that improves the accuracy and reliability of lung cancer diagnosis. The proposed system analyzes CT scan images through a structured processing pipeline that includes image preprocessing, lung region segmentation, multi-instance texture feature extraction, feature optimization, and deep learning based classification. Preprocessing techniques such as normalization, resizing, noise reduction, and contrast enhancement are applied to improve the visibility of pulmonary structures and ensure consistent analysis. The lung regions are then segmented using thresholding and morphological operations to isolate relevant anatomical structures and remove unnecessary background information. Following segmentation, texture features are extracted using Gray Level Co-occurrence Matrix (GLCM) methods to capture spatial relationships between pixel intensities, including statistical descriptors such as energy, contrast, entropy, and correlation. These features represent the heterogeneity present in pulmonary nodules and support effective discrimination between benign and malignant tissues. To improve computational efficiency and model generalization, feature fusion and dimensionality reduction techniques are applied to remove redundant information. The refined feature set is then processed using a convolutional neural network (CNN) to perform automated classification of lung nodules. By integrating statistical texture descriptors with deep learning techniques, the proposed system improves diagnostic accuracy while reducing false positives. The developed framework serves as a reliable and computationally efficient decision support tool to assist radiologists in early lung cancer detection and clinical decision making."
Disclaimer: Curated by HT Syndication.
