Skip to content

Is CT-Style Normalization Appropriate for MRI in BRATS Test Code? #4

@xwumed

Description

@xwumed

Hi,

I am running the segmentation pipeline on my own glioma MRI dataset using the code in BRATS23/test.py. I noticed that the normalization parameters (a_min=-175.0, a_max=250.0, etc.) are typical for CT preprocessing. However, BRATS and my data are both MRI, and MRI images do not have standardized intensity ranges like CT.

I am observing poor segmentation performance on my glioma MRI data. I suspect this might be related to the normalization method, since z-score normalization (zero mean, unit variance) is usually recommended for MRI images.

Could you please confirm whether using these CT-specific normalization parameters is intentional for BRATS, or would it be more appropriate to use z-score normalization for MRI images in this case?

Thank you so much in advance.

Best
Xuewei

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions