These models are NOT intended for 3D Printing. STLs are not recommended.
Model files include .fbx and .glb models for timepoints 1,2 and 3. Images were capture from TB-in-3D viewer where visibility of lung structures has been lowered to 20%.
The following models were generated for the TB in 3D viewer in NIAID's TB Portals as part of World TB Day 2021.
To create the case model, we performed manual and artificial intelligence (AI) segmentation of the lungs and lesions using the case’s three chest computed tomography (CT) images gathered by the TB Portals as part of its efforts to collect complete, multidimensional TB data.
These CTs come from the case of a 45-year-old male, a smoker, who was initially diagnosed with TB (first timepoint for the model - initial case baseline) and empirically treated for drug-sensitive TB due to negative microscopy and culture results. Two years later, the disease relapsed as a multidrug-resistant TB case with negative microscopy, negative BACTEC culture, and negative GeneXpert MTB/RIF at baseline, followed by positive LJ culture (second timepoint). The patient was treated with drugs consistent with a MDR TB regimen and improved significantly after just one month of treatment (third timepoint). He was cured over the course of 2-year therapy.
The CTs—together with the patient’s clinical, pathogen genomic, microbiological and anonymized socioeconomic data—can be found here as part of the TB Portals collection.
The AI algorithm aims to predict the location and extent of lesions (foci, caverns, fibrosis, pleuritis, atelectasis, and pneumothorax). It is currently a work-in-progress and was utilized by the TB Portals model developers for the “AI” viewing mode. The algorithm was developed by the TB Portals consortium members using 500 CTs from the TB Portals collection. We would like to acknowledge the contributions of the United Institute of Informatics Problems of the National Academy of Sciences of Belarus and the Republican Research and Practical Centre for Pulmonology and Tuberculosis of the Belarus Ministry of Health.
Comparison of the manual and AI results illustrates the power of machine learning in medicine and exemplifies the TB Portals efforts to make high quality algorithmic predictions available to wide area of doctors and researchers all over the world.
