Question about SliceThickness values in IDC portal

One of the users made the following observation about the content of IDC portal:

the Slice Thickness pie chart indicates that the units are in millimeters, but this doesn’t make sense given that the values are divided into bins with increments of 200 up to 1000. It seems likely that the units are actually micrometers.

According to the DICOM standard, the value of SliceThickness should always be defined in mm. The values reported in the portal are reflecting the values of SliceThickness that are present in the data. But it is understandable to be confused and question the accuracy of the values reported in the portal. Fortunately, it is relatively easy to quickly understand better what is going on.

We can investigate this a bit further by querying the IDC bigquery-public-data.idc_current.dicom_all metadata table.

You can open Google Cloud BigQuery console, and copy-paste and run the query below to identify all distinct series that have SliceThickness above 100, sorted in descending order by the value of SliceThickness. Note that SliceThickness is recorded in DICOM as a string, thus we have to cast the values to float.

SELECT
  DISTINCT(collection_id),
  SeriesDescription,
  SliceThickness,
  CONCAT("https://viewer.imaging.datacommons.cancer.gov/viewer/",StudyInstanceUID,"?seriesInstanceUID=",SeriesInstanceUID) as viewer_url
FROM
  `bigquery-public-data.idc_current.dicom_all`
WHERE
  SAFE_CAST(SliceThickness AS FLOAT64)>100
ORDER BY
  SAFE_CAST(SliceThickness AS FLOAT64) DESC

As can be observed from the values of SeriesDescription, most of those series correspond to scout scans, which cover the entire body thickness, and large values of SliceThickness make sense.

The last column in the resulting table will contain a URL string that can be opened in the browser to see the series corresponding to a given value of SliceThickness.

If we take a series that has a less readable SeriesDescription (“0,OPL,GE,LSQX,EXPERIMENTAL7,0,na,120,0,0,na”), which is observed in this series: https://viewer.imaging.datacommons.cancer.gov/viewer/1.2.840.113654.2.55.288521325592476656031520643558862749799?seriesInstanceUID=1.2.840.113654.2.55.84618752850688410105301877800324756665, we can confirm it is also a localizer.

We can also use ImageType to help with this investigation, but I am out of time at the moment, and will follow up on this a bit later.