I am working with EPI d
ata acquired on a Siemens 3T scanner. The data is multiband with no in-plane acceleration. I reconstructed the multiband data using the slice GRAPPA technique.
However, I am facing an issue: the reconstructed image intensity fades from anterior to posterior. In contrast, the corresponding Siemens-reconstructed images appear intensity-corrected. I have completed the reconstruction, but I am unsure what additional post-processing steps are required to match Siemens’ reconstruction. Specifically, I am looking for guidance on how to perform intensity correction on reconstructed single-band data to improve image quality.
In each run (.dat) file, there is a prescan (3D acquisition) and a 2D EPI acquisition (which is reconstructed). The prescan has lower resolution:
Prescan resolution: (32, 32, coils, 64) → (partition, phase, coils, readout)
Body prescan: (32, 32, 2, 64)
Head prescan: (32, 32, 38, 64)
The reconstructed EPI has resolution (36, 64, 38, 64) → (slices, phase, coils, readout). I will attach an image comparing my reconstruction with the Siemens reconstruction for clarity regarding intensity differences and post-processing effects.
Currently, my approach is as follows:
Use affine transforms to map voxels from prescan and EPI into scanner XYZ coordinates.
Transform to EPI space and interpolate to match the prescan and EPI dimensions.
Compute the ratio head_prescan / body_prescan to correct the EPI, applying smoothing to the ratio.
Unfortunately, this approach does not produce the expected intensity correction. It is possible that I am not correctly using the geometry transforms in the Siemens header, or that this approach is not optimal.
Any guidance would be greatly appreciated. In particular, I am interested in:
Correct usage of Siemens geometry information for intensity correction or other approaches that work well
Open-source tools or implementations that can reproduce Siemens-like intensity correction
Thank you very much for your time and assistance.