Am I right to assume that you put the ADC image into samseg? I'm not sure I've seen that before. If you have an anatomical, I would run samseg on that, then use mri_coreg (or bbregister if you ran it through recon-all) to register the anat to the ADC, then use mri_vol2vol to map the seg to the ADC space, then use that as for your ROI analysis. I'm not sure about your bias field correction question. If you run the analysis as I suggest, then there is not the option, which is probably the cleanest way to do it. If you want to let samseg do the bias correction, you will need to undo the scaling (I think there is a text file called scalefactor.txt or some such)

On 5/27/2025 11:43 AM, Taghizadeh Mortezaei, Nima wrote:

Hello everyone,

I am currently working with ADC values extracted from inhomogeneity bias-corrected ADC MRIs obtained through SAMSEG and have encountered a few issues that I would greatly appreciate your insights on.

1. Units for ADC Values: When extracting ADC values for my ROIs, using 3D Slicer, I noticed that no unit was provided for the ADC values, neither in the software nor in the imaging visualization tools (PACS, VISAGE). I also checked the DICOM metadata for ADC and DWI MRIs but could not find any relevant tags related to ADC values, rescaling, or intersection. As a result, I am uncertain about the unit or scale used for the ADC values.
2. Discrepancy in ADC Value Range: The mean ADC values extracted from the bias-corrected image for my ROIs were in the range of 150-250. These values seem far away from the normal average ADC values for these structures. I suspected that the bias correction might have influenced these values, so I also extracted the values from the original ADC image. In this case, the values ranged from 1200-1500 for ROIs. However, I still could not identify any unit or rescaling factor.

Considering that the average normal ADC values for my ROI are around 1000 to 1500 x 10^-6  mm²/s, I have two questions:

For extracting radiomics features (including ADC values) from the ADC image, would you recommend using the bias-corrected image or the original ADC image as the scalar volume?

Based on the values I found, there must be a rescaling factor (like 10^-6 ). Would you happen to know where I might find the rescaling factor, or how I could confirm it?

I would greatly appreciate any guidance you could provide on these issues.

Thanks, Nima

---

Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer