MUMBAI, India, Aug. 10 -- Intellectual Property India has published a patent application (202641089943 A) filed by Vellore Institute Of Technology on July 24, 2026, for Explainable Counterfactual Lung-Aware Framework For Pneumonia Detection From Chest X-Ray Images.
Inventors include Navamani T M; and Sruthy K G.
The application for the patent was published on July 31, 2026, under issue no. 31/2026.
Abstract: ABSTRACT The present invention discloses an explainable artificial intelligence framework for trustworthy pneumonia detection from chest X-ray images. The framework integrates lung-region-constrained feature learning, transformer-based classification, explainable AI, counterfactual verification, and reliability-aware model refinement. A lung segmentation and ROI extraction module isolates clinically relevant pulmonary regions, which are processed using an ROI-constrained Swin Transformer model for pneumonia classification and confidence estimation. An explainability generation module produces lung-focused visual explanations, which are further utilized for generating explanation-guided counterfactual images. Counterfactual prediction verification evaluates whether identified regions causally influence diagnostic outcomes, while an explanation reliability assessment module determines explanation faithfulness based on prediction consistency and confidence variations. A reliability-weighted counterfactual retraining mechanism improves robustness and reduces shortcut learning. The framework generates transparent diagnostic reports comprising predictions, confidence scores, explanation visualizations, verification outcomes, and trustworthiness indicators for reliable AI-assisted pneumonia diagnosis.
Disclaimer: Curated by HT Syndication.
