The Cancer Imaging Archive (TCIA) now offers three new data collections for cancer research. These include data from:

• glioblastoma multi-parametric magnetic resonance imaging (mpMRI),
• a glioblastoma-based MRI Digital Reference Object (DRO), and
• colorectal digital biopsy slides.

All these data sets have been deidentified and readied for immediate use. For more information on the latest additions, see below.

 

Glioblastoma multi-parametric mpMRI scans and accompanying information

Source

• University of Pennsylvania Health System

What’s Included?

• Whole brain segmentations as well as whole distinct tumor sub-regions, including intensity, volumetric, morphologic, histogram-based, and textural parameters.
• Patient demographics, clinical outcomes (e.g., overall survival, genomic information, tumor progression).

Advantages

• Manually corrected/approved by expert board-certified neuroradiologists.
• Researchers without computational backgrounds will be able to use this collection to look for associations with molecular markers (radiogenomic biomarker research), as well as clinical outcomes, treatment responses, and other endpoints.

Format used

• DICOM images. Includes a panel of radiomic features, along with their corresponding co-registered mpMRI volumes in NIfTI format. Clinical data and radiomic features are provided as CSV files.

DRO trained using DSC-MRI data from a patient database of 23 glioblastomas

Source

• Barrow Neurological Institute

What’s Included?

• The DRO data are separated into two collections based on magnetic field strengths (3T and 1.5 T).
• Each collection includes folders with corresponding repetition times (TR) and contrast agent dosing schemes. Sub-folders show flip angles and echo times.
• Single slice DSC-MRI signal time series samples at corresponding TR values for four regions of interest.

Advantages

• Offers a validated MRI signal computational approach.
• Useful for investigating how DSC-MRI acquisition and post-processing methods influence cerebral blood volume accuracy.
• Serves as a benchmark for perfusion analysis algorithms, potentially improving consistency in managing patients and clinical trials.

Format used

• DICOM images

Digital slides from colorectal biopsies

Source

• Archived from the 2nd Department of Pathology, Semmelweis University, Budapest

What’s Included?

• Two hundred digital whole-slide images obtained from hematoxylin-eosin stained biopsy slides captured at the highest available magnification (40x).

Advantages

• This single center data set, with homogeneous data and consistent processing and methodology, offers a useful source for training an artificial neural network to detect pathological conditions.

Format used

• Raw MIRAX (mrxs) formatted data