r20240729

This release contains the SCIsegV2: A universal tool for the segmentation of intramedullary lesions and the spinal cord in SCI. The orientation of the input image is expected to be RPI (right-posterior-inferior). The nnUNet output structure is kept intact, thus allowing it to be used in the 3DSlicer software as well.

This work was accepted at the MICCAI AMAI 2024 workshop. The preprint can be found at https://arxiv.org/pdf/2407.17265

r20240716

This release contains the updated SCIsegV1 model for lesion and spinal cord segmentation in traumatic/ischemic SCI. The orientation of the input image is expected to be RPI (right-posterior-inferior).

Corresponding preprint: https://www.medrxiv.org/content/10.1101/2024.01.03.24300794v2

r20240130

This release contains the latest SCIseg model for lesion and spinal cord segmentation in traumatic SCI. The orientation of the input image is expected to be RPI (right-posterior-inferior).

Corresponding preprint: https://www.medrxiv.org/content/10.1101/2024.01.03.24300794v1

r20231108

A release of the SCISeg model submitted to ISMRM 2024. Contains data conversion, nnUNet training/testing, lesion metrics evaluation scripts. Model was trained on Zurich sagittal and Colorado axial T2w images with heterogeneous resolutions.

What's Changed

• New preprocessing script with centerline extraction by @naga-karthik in #30
• Add masks{Tr, Ts} folders to nnUNet dataset format by @valosekj in #32
• add zurich & spine-generic data conversion script by @naga-karthik in #34
• Generate sagittal QC for all subjects by @valosekj in #42
• Generate new lesion by @valosekj in #31
• Add script to segment SC for sci-paris by @valosekj in #47
• Minor improvements of anima script by @valosekj in #48
• Add a scripts for conversion of all datasets by @valosekj in #50
• Update preprocess_data_sci-zurich.sh script by @valosekj in #35
• packaging nnunet for inference by @naga-karthik in #40
• Get orientation for each file by @valosekj in #53
• Add script to compare our nnUNet model with other methods by @valosekj in #55
• Fetch sequence parameters by @valosekj in #58
• Run inference on sci-zurich axial T2w images by @valosekj in #57
• Add script for cropping the images around the SC by @valosekj in #62

New Contributors

• @valosekj made their first contribution in #32

Full Changelog: v0.1...r20231108

sci_seg v0.1

Release sci_seg v0.1

First release of the sci_seg package containing ivadomed models for automatic segmentation of lesions in acute SCI. Contains preprocessing scripts and configuration files for reproducing preliminary results on both datasets.