Request for Co-registration Code

I am currently working with raw data and attempting to warp results into NCCT space. However, I am unable to access existing co-registration codes or transformation parameters, making it difficult to perfectly restore the pre-provided derivative data even if I attempt to construct individual matching codes. Therefore, I am requesting the release of a co-registration code or entire preprocessing code to ensure compatibility across dataset space.

Thank you.

Hi,

We don't aim to release the preprocessing scripts. If you want to use raw data, then you can preprocess it (including co-registration) by yourself, using your own strategy. Alternatively, if you don't want to preprocess the images yourself, you can use the already preprocessed images- so that there is no need to take care of co-registrations.

Best, Ezequiel.

Hi, have you validated the registration with experts? it seems to me there are some errors e.g. the segmentation interacts with the bone in some images and outside the voxel data in others, see the attached figures.

Hi,

Short answer- yes. Scans are qc-ed by radiologists.

Long answer:

• Some modalities do not provide full head coverage -as the CTP images and parameter maps you showed in the first figure. NCCT and CTA, in most cases, do cover the full head, as follow-up MR scans. That's why lesion masks might fall outside voxel areas.

• Registration might sometimes be sub-optimal given:

1) The inherent cross-modality registration. 2) The fact that longitudinal scans are being co-registered. Thus, acute and sub-acute images are being aligned, with existing pathological changes in between. 3) Interpolation - MR has way lower voxel resolution than e.g. NCCT or CTA. 3) In some cases, there is visible edema in the follow-up MR scans, not appearing at the acute stage. This generates e.g. compression of the CSF cavities, and consequently, lesions might fall in non-brain tissue areas in acute modalities.

Hope this helps. Ezequiel.