Intellectual Property India Publishes Patent Application for A Privacy-Preserving Distributed Machine Learning System For Early Prediction And Risk Stratification Of Lung Cancer Using Multimodal Clinical And Imaging Data

MUMBAI, India, Oct. 12 -- Intellectual Property India has published a patent application (202641118934 A) filed by Sr University on October 05, 2026, for A Privacy-Preserving Distributed Machine Learning System For Early Prediction And Risk Stratification Of Lung Cancer Using Multimodal Clinical And Imaging Data.

Inventors include S Varsha Reddy; and Dr. Kabita Thaoroijam.

The application for the patent was published on October 09, 2026, under issue no. 41/2026.

Abstract: ABSTRACT The present invention relates to a privacy-preserving distributed machine learning system and method for early prediction and risk stratification of lung cancer using multimodal clinical and medical imaging data. The system comprises a plurality of healthcare computing nodes configured to retain patient-level clinical records and medical images within respective institutional environments. A data preprocessing and harmonization module standardizes heterogeneous clinical and imaging information, while an imaging feature extraction module and a clinical feature extraction module generate modality-specific representations. A multimodal fusion module combines the representations to generate a unified patient-specific representation. A distributed machine learning module enables collaborative training among the healthcare computing nodes without transferring raw patient data. A privacy protection module applies differential privacy, gradient clipping, secure aggregation, encryption, cryptographic masking, or combinations thereof to protect locally generated model updates. A secure aggregation module combines protected updates to generate an updated global prediction model. The trained model generates an early lung cancer risk score and corresponding risk stratification. An explainability module identifies influential imaging regions and clinical variables, while an uncertainty estimation module determines prediction reliability and generates an alert when uncertainty exceeds a predefined threshold. The invention thereby facilitates collaborative, privacy-preserving, multimodal lung cancer risk prediction across heterogeneous healthcare institutions while maintaining local control over sensitive patient information.

Disclaimer: Curated by HT Syndication.