MUMBAI, India, July 24 -- Intellectual Property India has published a patent application (202623065434 A) filed by Pia Toor; Amarjit Singh Toor; and Jingles Toor on May 24, 2026, for Ai-Embeded System And Method For 3d Reconstruction Of Adult Tb Lung From 2d Ventral-Dorsal Radiograph.
Inventors include Pia Toor; Amarjit Singh Toor; and Jingles Toor.
The application for the patent was published on July 17, 2026, under issue no. 29/2026.
Abstract: ABSTRACT Inventors have designed a path breaking, economical, secured, scalable and user-friendly solution using in-house developed artificial intelligence-based model. It is a method and system to re-construct a highly accurate patient-specific 3D lung model showing correspondingly TB infection severity, variable confidence score with respect to quantum of infection by using only a single Posterior Anterior, 2D X ray of Chest in DICOM format with minimal noise, good resolution and preferably of Indian adult patient. It helps the pulmonologist to quickly & precisely ascertain a bird’s eye view in 360 degrees of Tb infected multiple regions of both the lobes of the lung. Further, it enables him to theoretically estimate the Quantum of spread of the potential Tb infection within the specific region of that particular lobe, multiple direction and depth wise interior portion of the lung. Inventor's ecosystem uses multiple sections of CT scan images from reputed open-source datasets, subject them to denoising, segmentation, then binary classification using a multi layered indigenous (inventors own awarded patent number: 550456, priority date: 24/09/2017) Convolution Neural Network (CNN) AI trained, validated and tested model having extremely high accuracy. Further, steps involve carving out lung, masking of the infection zone, segmentation with high confidence level, followed by coloured heat map with severity prediction feature. A learned 3D lung prior conditioned on the 2D view, combined with anatomical landmarks and priors, yields a volumetric lung shape; 2D TB evidence is extrapolated into 3D via depth prior networks and data from anatomical atlases. It is multi-disciplinary novel integration of knowledge skills from machine learning, data analytics, computer vision, image processing, medical instrumentation & health sciences. Inventor’s method has minimal mismatch between the available ground truth either in terms of equivalent CT scan and/or in combination Lateral view TB Chest X ray images. The correctness and or the certainty of the inventor system is quantified using geometric plausibility, 2D–3D segmentation consistency, and clinical validation metrics. Inventors’ disruptive solution enables volumetric TB assessment in reso-limuriceted settings without CT or multi-view, radio isotope rich dosage and expensive imaging solution prevailing in the health sector. Usage of transfer learning technique facilities to use the in- house made in Bharat technology to construct volumetric 3d view of other anatomical parts and plurality of respiratory diseases. Key Words: 3D Lung, X Ray, CT, Reconstruct, Metrices
Disclaimer: Curated by HT Syndication.
