Patient CT Projection Data Library
Our team is pleased to announce the public availability of data from 299 patient CT exams of the head, chest and abdomen, 150 cases from Siemens scanners and 149 cases from GE scanners. All cases have routine dose and simulated low dose projection data, full dose images, and clinical data identifying all pathology. This library is unique in that it includes, for the first time ever, the x-ray projection data used to create cross-sectional images, data that are critical for the development of advanced reconstruction and denoising techniques, including those using artificial intelligence.
The need for open access to patient projection data was recognized by our group and others for a number of years. Our plan to develop these resources became focused at the NIBIB sub mSv CT workshop held on February 24-25, 2011. In 2012, NIBIB announced a sub mSv CT funding program and this work was one aim of our funded proposal.
We are excited to finally make the database publicly available worldwide. It will be rewarding to see what new science comes from having these data available!
For each patient CT scan, three types of data are provided: DICOM-CT-PD projection data, DICOM image data, and Excel clinical data reports. DICOM-CT-PD is an extended DICOM format that contains CT projection data and acquisition geometry in a vendor-neutral fashion (1). The de-identified patient projection data were decoded with help of the manufacturer and converted into this open standardized format.
CT projection data are provided for both full and simulated lower dose levels. CT image data reconstructed using the commercial CT system are provided for the full dose projection data. For patients scanned on the SOMATOM Definition Flash CT scanner from Siemens Healthcare, CT image data reconstructed using the commercial CT system are provided for the simulated lower dose projection data. All CT images were reconstructed using a filtered back projection method.
The collection comprises 99 head scans, 100 chest scans, and 100 abdomen scans. Fifty cases for each scan type are from a SOMATOM Definition Flash CT scanner (Siemens Healthcare, Forchheim, Germany). Forty-nine head cases, 50 chest cases, and 50 abdomen cases are from a Lightspeed VCT CT scanner (GE Healthcare, Waukesha, WI).
(1) Chen B, Duan X, Yu Z, Leng S, Yu L, McCollough CH. Technical Note: Development and validation of an open data format for CT projection data. Med Phys. 2015;42(12):6964. (doi: https://doi.org/10.1118/1.4935406)
Acquisition and Noise Insertion
All CT scans were acquired at routine dose levels for the practice at which they were obtained using standard clinical protocols for the anatomical region of interest. Poisson noise was inserted into each projection dataset to create a second projection dataset at a simulated lower dose level. Head and abdomen cases are provided at 25% of the routine dose and chest cases are provided at 10% of the routine dose. The simulation technique accurately models the impact of tube current modulation, the bow-tie filter, and electronic noise (2).
Provided projection data are from right before image reconstruction, i.e., after all preprocessing and taking the logarithm. For reconstruction algorithms requiring statistical information, a noise map, expressed as the noise-equivalent incident number of quanta for each data frame, is provided for data from the Siemens CT system.
Several instructional documents are provided to help users extract needed information from the DICOM-CT-PD files, including a dictionary file for the DICOM-CT-PD format, a DICOM-CT-PD reader, and a user manual.
(2) Yu L, Shiung MM, Jondal D, McCollough CH. Development and validation of a practical lower-dose-simulation tool for optimizing computed tomography scan protocols. J Comput Assist Tomogr. 2012;36(4):477-87. (doi:10.1097/RCT.0b013e318258e891)
Citation of these data should use the following: McCollough, C.H., Chen, B., Holmes, D., III, Duan, X., Yu, Z., Yu, L., Leng, S., Fletcher, J. (2020). Low Dose CT Image and Projection Data [Data set]. The Cancer Imaging Archive. https://doi.org/10.7937/9npb-2637
"Thank you so much, thank you for providing us with relevant data about CT. With these data we can carry out research in related fields. Your behavior makes us feel that science knowledge no borders and understands the academic spirit. If we have published a paper, we will emphasize the contribution of you and your team. Although you and I are thousands of miles apart, we are fortunate to be linked by academic exchanges."
-A grateful investigator
Funding for this work was provided by the National Institute of Biomedical Imaging and Bioengineering under grants EB 017095 and EB 017185.
We gratefully acknowledge the following individuals who assisted in converting their company’s proprietary data format into DICOM-CT-PD:
We would also like to thank our radiologist colleagues at Mayo Clinic Rochester and Mayo Clinic Arizona for collecting, evaluating and annotating the patient cases, as well as many members of our CT Clinical Innovation Center team. This work would not have been possible without their skill, perseverance, and can-do attitude, for which we are sincerely grateful.
Finally, we are grateful to the many patients who have allowed their imaging data to be shared for research purposes.
A list of full acknowledgements can be found here.