MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641074241 A) filed by Koneru Lakshmaiah Education Foundation; Chinaguravaiah Makkena; Pavan Kumar Pagadala; Hari Prasada Raju Kunadharaju; and Ramakrishna Akella on June 15, 2026, for A Context-Shift Convolution–based Deep Learning System And Method For Automated Multi-Class Skin Lesion Classification Using Dermoscopic Images.
Inventors include Chinaguravaiah Makkena; Pavan Kumar Pagadala; Hari Prasada Raju Kunadharaju; and Ramakrishna Akella.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: The present invention relates to CSC-DermNet, a Context-Shift Convolution–based deep learning system and method for automated multi-class classification of dermoscopic skin lesions. The disclosed framework introduces a Context-Shift Convolution mechanism that enhances contextual feature representation through controlled spatial displacement and convolutional fusion of feature channels. Combined with melanoma-weighted focal loss, the proposed architecture effectively addresses class imbalance while improving malignant lesion recognition. Experimental evaluation using HAM10000 and ISIC benchmark datasets demonstrates an overall accuracy of approximately 93.2%, melanoma sensitivity of approximately 88.7%, specificity approaching 95%, and an AUC of approximately 0.97. The invention provides an efficient, robust, and clinically deployable solution for early skin cancer diagnosis and decision support.
Disclaimer: Curated by HT Syndication.
