How to access radiomics features accompanying TotalSegmentator-CT-Segmentations analysis results collection?

Support
1

Thank you so much for making this resource available to the scientific community! Are the radiomic analysis results available through IDC? Within the OHIF v3 viewer I see two structured reports with shape and first order measurements, but attempting to select these gives me an error message ‘Something went wrong.’ Is there an alternate way to access these structured reports?

1 Like
2

@virostko thank you for posting your question in the forum!

Yes, it is rather easy to download SRs from this collection using the idc-index package. See the steps below, or check out this brief colab notebook with all of the steps.

Step 1. Install idc-index: pip install --upgrade idc-index.

Step 2. Instantiate IDCClient provided by idc-index, and select SRs from the TotalSegmentator-CT-Segmentations analysis results collection

from idc_index import index

client = index.IDCClient()

query = """
SELECT 
  PatientID, 
  StudyDate, 
  SeriesInstanceUID, 
  SeriesDescription
FROM
  index
WHERE
  Modality = 'SR' AND 
  analysis_result_id = 'TotalSegmentator-CT-Segmentations'
"""
selection_result = client.sql_query(query)

Step 3: Download the files referenced by SeriesInstanceUID.

sr_series = selection_result.SeriesInstanceUID.values.tolist()
client.download_from_selection(seriesInstanceUID=sr_series,downloadDir = ".")

Note that there are about 250K of those, and you will need about 140GB of disk!

We will have more instructions how to navigate the content of those structured reports - I will add those steps a bit later.

We were also thinking about hosting these measurements as tables attached to the Zenodo descriptor that currently only contains the manifests, but I’ve been struggling a bit with what would be the most convenient representation for those features. With the amount of images and segments we have, we have the total of 267,819,496 radiomics features, and the size of the table, even with just the basic attributes accompanying the features (name of the structure and feature, segmented CT image), would be in gigabytes.

Would such table be helpful to you? Would you have preference over the grouping of the features, if we split them into multiple files (e.g., by timepoint? by anatomic structure?).

Currently, I am thinking that perhaps splitting by anatomic structure would be the most convenient.

Hope this helps, and please do follow up with questions if anything is unclear, or if there’s anything else we can do to help you use this dataset!

3

Thank you for these instructions! I do think a table of features split by anatomic structure would be helpful. I am interested in gallbladder features for a current project.

1 Like
4

@virostko sorry for the delay, but we have now added zip files with the per-structure radiomics features in CSV and Parquet format to the Zenodo record. you can access them here: https://doi.org/10.5281/zenodo.12004521.

Please let us know if you have any feedback! Hope this is helpful.

5

Very helpful - thank you so much!

1 Like