neuropoly / paradim-segmentation-brain

Repository for implementing pre-existing brain segmentation models trained with deep learning in PARADIM
MIT License
2 stars 0 forks source link

Deal with dicom to niftii conversions #3

Open NathanMolinier opened 7 months ago

NathanMolinier commented 7 months ago

Description

The objective of this issue is to discuss about dicom to niftii and niftii to dicom conversions.

Functions to explore

Nilser3 commented 7 months ago

I find interesting the approach of rt-utils, especially its facility to save the input in a new array.

But I had issues to keep the orientation of the inputs and outputs (see image). also, as it is a mesh and not voxels, I observe interpolation problems in very irregular contours.

image

I will explore pydicom-seg approaches: #2

Orientation issues reported

https://github.com/qurit/rt-utils/issues/102 https://github.com/qurit/rt-utils/issues/84

NathanMolinier commented 6 months ago

Nifti and DICOM are using a different spacial coordinate system

NathanMolinier commented 6 months ago

Very interesting issue about DICOM orientation: https://gist.github.com/agirault/60a72bdaea4a2126ecd08912137fe641