Intellectual Property India Publishes Patent Application for A Federated Multi-Node Deep Learning Technique With Cluster Visualized Lda (cvlda) Feature Reduction For High-Performance Breast Cancer Classification Across Distributed Clinical Data Environments

MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202641096794 A) filed by Koneru Lakshmaiah Education Foundation on August 11, 2026, for A Federated Multi-Node Deep Learning Technique With Cluster Visualized Lda (cvlda) Feature Reduction For High-Performance Breast Cancer Classification Across Distributed Clinical Data Environments.

Inventors include Naidu Kirankumar; and Dr T. Santhi Sri.

The application for the patent was published on August 14, 2026, under issue no. 33/2026.

Abstract: ABSTRACT The present invention relates to a privacy-preserving artificial intelligence framework for breast cancer classification across distributed clinical data environments. The invention integrates federated multi-node deep learning with Cluster Visualized Linear Discriminant Analysis (CvLDA) feature reduction. Multiple hospitals, diagnostic centres, laboratories, and healthcare institutions locally process breast cancer imaging and clinical datasets, perform preprocessing and feature extraction, and train local deep learning models without transferring raw patient information. Locally optimized model parameters or updates are securely communicated to a federated aggregation mechanism to generate and iteratively improve a global classification model. The CvLDA mechanism clusters related and correlated diagnostic features, identifies discriminative feature groups, and projects high-dimensional representations into a reduced feature space with improved class separability. The resulting compact representation is supplied to a deep learning classifier for breast cancer classification. The framework includes secure communication, federated aggregation, performance evaluation, feature visualization, validation, and clinical decision-support modules. Evaluation may include accuracy, sensitivity, specificity, precision, recall, F1-score, ROC analysis, AUC, convergence, communication efficiency, computational complexity, and model stability. The invention provides a scalable framework for collaborative healthcare analytics while maintaining patient confidentiality.

Disclaimer: Curated by HT Syndication.