Possible issue with 2D extraction from DECT images

Community Radiomics
1

Dear all,

We are currently working on a project that involves obtaining pyradiomics features from dual-energy CT images which are in a normal DICOM format and can be loaded into 3DSlicer without any problems. We used the Pyradiomics 3DSlicer extension to compute results from 2D circular regions of interest (ROIs) in different organs. However, in each of the different kinds of images as well as in each of the ROIs we measured, we are getting ‘odd looking’ results for some of the examinations.

In the screenshot I attached, you see results obtained from an ROI in the lower spleen. Each row is one (different) CT examination. We are uncertain because there are examinations for which a lot of the features give the same numbers (often 0/1) without decimals – is this likely to be due to an issue with the extension and/or image or may these results actually be sound?

Thank you very much for your help - happy to provide additional information, if needed.

Simon

image
image2933×305 40.6 KB

2

Hi - It would probably help to share screenshots of the corresponding ROIs.

3

Thank you very much for your reply. I attached the screenshots of the ROIs that resulted in the first 10 rows of data.

Row1
Row1718×643 38 KB
Row2
Row2749×619 32.8 KB
Row3
Row3745×659 62 KB
Row4
Row4476×603 31.3 KB
Row5
Row5481×591 39.1 KB
Row6
Row6471×650 44 KB
Row7
Row7605×612 50.9 KB
Row8
Row8613×664 41.6 KB
Row9
Row9602×642 50.4 KB
Row10
Row10477×661 43.7 KB

4

Hmm, I don’t see anything obvious in the images. Maybe @fedorov or @JoostJM has suggestions?

5

The difference between minimum and maximum seems quite small, did you use the default bin width of 25 for your images? Then the discretization will probably result in too few gray values, which may in part explain your results. Other than that I don’t know, for that I would need the results of the diagnostic features and/or the log from the extraction (available through slicer’s report a bug option).