Automated Identification of Mod-Sev TBI Lesions Banner
Join

Automated Identification of Mod-Sev TBI Lesions

PLEASE NOTE, AIMS-TBI IS NOT CURRENTLY ACCEPTING SUBMISSIONS

Thank you for submitting to the AIMS-TBI segmentation challenge! Please start by requesting access to the training data: https://forms.gle/2dmoTGeksyA4X1tT8

BACKGROUND

Moderate to Severe Traumatic Brain Injury (msTBI) is caused by external forces (eg: traffic accidents, falls, sports) causing the brain to move rapidly within the skull, resulting in complex pathophysiological changes. Multiple primary, secondary, and surgery related processes has the potential to cause structural deformation in the brain. Each patient with msTBI has a unique accumulation of these structural changes, contributing to extremely heterogeneous lesions, considered a hallmark of msTBI (Covington & Duff, 2021). These lesions differ from other common brain pathologies (stroke, MS, brain tumor) in that they can be both focal or diffuse, varying in size, number and laterality, extending through multiple tissue types (GM/WM/CSF), and can also occur in homologous regions of both hemispheres. Lesions such as these can complicate image registration, normalization, and are known to introduce both local and global errors in brain parcellation (Diamond et al., 2020; King et al., 2020). While multiple tools exist to compensate for lesions in neuroimaging preprocessing (HD_Bet (Isensee et al., 2019), VBG (Radwan et al., 2021), FastSurfer), many require the time consuming manual creation of lesion masks and subsequent manual quality assessment. Furthermore, in our experience, methods that have been developed for lesions of different etiologies (e.g. stroke, tumors [Henschel et al., 2020]) do not perform well in TBI. Whilst a handful of TBI specific algorithms exist, they require either multiple image types (Kamnitsas et al., 2017) (T1, T2, FLAIR, GE & PD) or can run on only CT images (Jain et al., 2019). However, the necessity for multiple image types limits the ability of large scale consortia to aggregate common MRI scans across sites and there is a larger variability in scanning sequence parameters in other MRI modalities (such as diffusion MRI). Therefore, this challenge will focus on identifying lesions in T1 weighted MRI data only as it is the most common MRI scan across our ENIGMA TBI consortium. Advances in lesion segmentation and the implementation of an accurate lesion mask resulting from the lesion segmentation into the next image processing and analyses (such as parcellation, functional connectivity analyses, connectomics, fixel based analysis) will allow for a more accurate prognostication and may improve long term outcomes for patients.

TASK

Segment lesions in T1-weighted MRI data from moderate-severe traumatic brain injury.

Inputs

  • T1-weighted MRI images
  • Tabular demographic and clinical data

Outputs

  • Binary TBI lesion segmentation mask.

SCHEDULE

  • April 22: Training data released
  • May 15: Validation data will be released
  • June 30: Validation phase closes
  • July 1: Final model submission opens
  • August 15: Deadline for submission dockers