Are there any contrast-enhanced CT images of forearms in IDC? Or even someplace else? The closest I found is Soft-tissue-Sarcoma, namely STS_008 and STS_019. But they don’t have contrast, and I am aiming to segment the few biggest blood vessels.
Such images do exist, as there are some on radiopaedia, such as this. But radiopaedia has non-commercial clause, and does not offer a convenient way to download a 3D image.
I only need one such image. Higher resolution is preferable, but something is better than nothing.
I don’t know of any off the top of my head, but I’ll bet there are some included in other scans by chance. I have found that posing general questions like this to an LLM and asking it to review the collection summaries can be helpful. Generally they are well informed about the collections and can usually tell from the descriptions what kind of images are included based on the the disease and other factors. I had reasonable luck with this finding images that were likely to have enlarged lymph nodes for instance.
I don’t have the answer, but in case you have not found it, we also have this dashboard, which gives you access to the contrast related attributes (contrast agent and route). I didn’t find anything, but maybe this can help.
I don’t think we ever had contrast in the portal.
The underlying BigQuery query for that dashboard is below
SELECT
SeriesInstanceUID,
any_value(PatientID) as PatientID,
any_value(StudyInstanceUID) as StudyInstanceUID,
ANY_VALUE(ConvolutionKernel) AS ConvolutionKernel,
ANY_VALUE(series_aws_url) AS series_aws_url,
ANY_VALUE(ContrastBolusAgent) AS ContrastBolusAgent,
ANY_VALUE(SeriesDescription) AS SeriesDescription,
ANY_VALUE(ContrastBolusRoute) AS ContrastBolusRoute,
ANY_VALUE(collection_id) AS collection_id,
sum(instance_size) as series_size,
ANY_VALUE(ROUND(SAFE_CAST(PixelSpacing[SAFE_OFFSET(0)] AS float64), 2)) AS PixelSpacing,
ANY_VALUE(round(SAFE_CAST(dicom_all.SliceThickness AS float64),2)) AS SliceThickness_rounded,
ANY_VALUE(dicom_curated.BodyPartExamined) AS BodyPartExamined_curated,
ANY_VALUE(Manufacturer) AS Manufacturer,
COUNT(DISTINCT(dicom_all.SOPInstanceUID)) AS num_slices_cnt,
CONCAT("https://viewer.imaging.datacommons.cancer.gov/v3/viewer/?StudyInstanceUIDs=",ANY_VALUE(StudyInstanceUID),"&SeriesInstanceUID=",seriesInstanceUID) AS viewer_url
FROM
`bigquery-public-data.idc_current.dicom_all` AS dicom_all
JOIN
`bigquery-public-data.idc_current.dicom_metadata_curated` AS dicom_curated
ON
dicom_all.SOPInstanceUID = dicom_curated.SOPInstanceUID
WHERE
Modality="CT"
GROUP BY
SeriesInstanceUID
HAVING
ContrastBolusAgent IS NOT NULL
AND num_slices_cnt > 10
AND SliceThickness < 2
I tried relaxing the SliceThickness constraint, but still didn’t find any obvious ones that have whole body or arm, but you can search harder if you like.
Also if you ask an LLM it can come up with a pretty good heuristic for guessing the contrast status of a scan based on the study and series descriptions.
With all due respect to LLMs, I am not convinced they can come up with heuristic to detect whether a series with a blank SeriesDescription (there are 98 of those), or SeriesDescription that have something like “Reformatted” or “CT” has contrast or not.
At the same time, ContrastBolusAgent and ContrastBolusRoute are often not blank for those (as you can see in the dashboard I linked earlier), and are designated means to indicate contrast administration.
Once I have made a selection, how do I download it? Clicking on “Download manifest” in the bottom right does nothing. Right clicking offers “Download page as PDF”, which does not contain the full links.
Thank you. I missed that each section of that dashboard page has its own separate export process, exporting only its specific part of the data presented.
