segmentation in some hard conditions

[Jiayue2000]

Hello!
We are trying to use SynthSeg to segment some undersampled images. Based on my currently results, SynthSeg seems not very suitable for this task. Below is my result. It is a undersampled T1w image with acceleration factor =6. I would like to know if the result is right or if I can try some other things to make it better. Thank you very much!
my command: python ./scripts/commands/SynthSeg_predict.py --i ./test_data/gaussian1d_R6_T1/T1w.nii.gz --o ./test_data/gaussian1d_R6_T1/T1w_SynthSeg.nii.gz --resample 0.8,0.8,0.8

[BenBillot]

Hi,
first of all, thanks for the interest in our work.
I see two issues there:

1. First, your not using the resample flag correctly. Your input input image will be internally resampled to 1mm isotropic resolution, because that's the resolution of the voxel SynthSeg has been trained on. the --resample flag let's you save the resampled image, such that you have access to it. So the --resample flag should be used with the path where you want to save the resampled image.
2. Your brain is badly oriented, which seems to indicate that your affine matrix in the image header doesn't correspond to the actual image. I'm pretty sure that if you manage to fix that you'll obtain a good segmentation.

[BenBillot]

Hi, could please close the issue of you think your problem has been resolved? I lost access to my previous github account, so I cannot close issues anymore. Thanks! :)