TB Portals Published Clinical Data contains the following data sets:

• TB Portals Patient Cases – Master file of patient case data. Each row represents a unique patient case with “condition_id” as the unique identifier and contains all distinct data points that were ever recorded for the patient case (shown in a comma-delimited list if more than one result). Domain-specific data files are also included separately when there is more than one result possible per patient case. Columns “condition_id” through “organization” are clinical/socioeconomic attributes, “image_body_site” through “totalcavernum” are CT study annotations, “overall_percent_of_abnormal_volume” through “highdensitycalcifiedtypicallysequella” are manual CXR annotations, “culture” through “genexpert_rifampicin” are DST results, “period_start” through “specimen” are treatment and clinical/socioeconomic attributes, “lineage” through “genexpert” are pathogen genomic attributes, and “qure_bluntedcp” through “qure_cavity” are automated CXR annotations by collaborator Qure.ai
• TB Portals Regimens – Detailed breakdown of the treatment dates (relative to registration date in days) and regimens for each patient case. Data joins to the Patient Cases file using “condition_id”.
• TB Portals DSTs – Detailed breakdown of the drug susceptibility tests (DST) and results by patient and specimen for microscopy, culture, MGIT Bactec, First- and Second- Line Probe Assays, Lowenstein-Jensen, and GeneXpert. The results of the DST tests are shown aggregated as well as captured individually based on the test for each drug. Data joins to the Patient Cases file using “condition_id”.
• TB Portals Biochemistry – Detailed breakdown of the Biochemistry lab tests and results. Data joins to the Patient Cases file using “condition_id”.

TB Portals Published Genomic Data contains the following data sets:

• TB Portals Genomics – Detailed breakdown of the pathogen genomic variance data. It captures the sequences, SRA metadata, and gene variants by patient and specimen. It also includes the gender, age of onset, and type of resistance for each patient case. Data joins to the Patient Cases file using “condition_id.
• TB Portals Sequenced DST - Detailed breakdown of the drug susceptibility tests (DST) and results by patient and specimen for microscopy, culture, MGIT Bactec, First- and Second- Line Probe Assays, Lowenstein-Jensen, and GeneXpert. The results of the DST tests are shown aggregated as well as captured individually based on the test for each drug. Data joins to the Patient Cases file using “condition_id”. Note: this DST dataset is limited to sequenced specimens only. Additional DST results are available for other specimens in the full clinical dataset.

TB Portals Published Imaging Data contains the following data sets:

• TB Portals CXRs – Detailed breakdown of the Chest X-Ray (CXR) image dates (relative to registration date in days). It also includes the gender, age of onset, and type of resistance for each patient case. Data joins to the Patient Cases file using “condition_id”. Data joins to the CXR images using “series_instance_content_url” as filepath.
• TB Portals Published CXR Manual Annotations – Detailed breakdown of CXR imaging annotations added by a radiologist. Data joins to the CXR images using “imagingstudy_ID”.
• TB Portals Published CXR Qure Annotations – Detailed breakdown of CXR imaging annotations generated by Qure.ai. Data joins to the CXR images using “imagingstudy_ID”.
• TB Portals CTs – Detailed breakdown of the CT series image dates (relative to registration date in days). It also includes the gender, age of onset, and type of resistance for each patient case. Data joins to the Patient Cases file using “condition_id”. Data joins to the CT series images using “patient_id / uid / series_uid” as filepath.
• TB Portals Published CT Annotations – Detailed breakdown of the CT imaging annotation reports. Data joins to the CT images using “imagingstudy_ID”.
• TB Portals Published CXRs - The Chest X-Ray (CXR) images tie to patient cases using the parent folder “patient_id” and TB Portals CXRs data file. A patient can have zero to many images. Please note the image filepaths are very long and can cause issues for Windows users if downloaded into a file directory with a long path name (greater than 260 characters). On Windows 10, enabling file paths longer than 260 characters require administrator privileges.
  • The CXR images were provided by multiple institutions and in many cases are digitized versions of film x-rays. Digitization is performed using either a flatbed scanner or a digital camera. The DICOM modality tag (0008, 0060) value for images digitized with a camera is XC, the value for all other images is CR.
  • The collection includes both color and grayscale images. Grayscale images come in two forms, low dynamic range images with intensities in [0,255] and high dynamic range images with intensities in [0, 2^16-1]. Color images are the result of digitizing with a camera. To correctly interpret these images, one needs to use the photometric interpretation (0028,0004) DICOM tag. Extracting a valid intensity image depends upon your DICOM reader, whether it keeps the three channels in the original color-space specified by the photometric interpretation or if it converts it to the common RGB color-space.
• TB Portals Published CTs - The CT series images tie to patient cases using the parent folder “patient_id” and TB Portals CTs data file. A patient can have zero to many images. Please note the image filepaths are very long and can cause issues for Windows users if downloaded into a file directory with a long path name (greater than 260 characters). On Windows 10, enabling file paths longer than 260 characters require administrator privileges.
• Image Quality - The Image Quality script traverses a file system starting at a user specified directory and outputs the modality, image size, and pixel quality data for DICOMs in a CSV file. You can run the script on the TB Portals CXRs and CTs. Optional inputs to the script include the prefix to the output file (which will be a csv) and a program that verifies compliance with the DICOM standard (e.g. the dciodvfy verification tool from David Clunie, https://www.dclunie.com/dicom3tools.html). NOTE: If the image could not be read (corrupted or not DICOM) only the file name will be present.

NIAID provides researchers three ways to access TB Portals data:

(1) File-based download of TB Portals data via Aspera File Share

(2) Programmatic access via the TB Portals Data API

(3) Cohort creation, analysis, and export via TB DEPOT

Each method has the same underlying data, but different purposes and download formats. We encourage everyone to review TB Portals Data Sharing – Data Download Options (nih.gov) for more information.