Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

From Diagnostic CT to DTI Tractography labels: Using Deep Learning for Corticospinal Tract Injury Assessment and Outcome Prediction in Intracerebral Haemorrhage

Olivia N Murray, Hamied Haroon, Paul Ryu, Hiren Patel, George Harston, Marieke Wermer, Wilmar Jolink, Daniel Hanley, Catharina Klijn, Ulrike Hammerbeck, Adrian Parry-Jones, Timothy Cootes

TL;DR

This study addresses the challenge of CST injury assessment in intracerebral haemorrhage when diffusion MRI is unavailable by training an nnU-Net to predict diffusion-based CST labels from diagnostic CT. The model uses paired CT and high-directional DTI data to achieve a mean $DSC=57\%$ relative to diffusion tractography and is applied to the MISTIE III trial dataset to derive CST-integrity metrics that significantly predict motor outcomes in both acute and chronic stages. The approach enables CT-only CST assessment, offering prognostic value and potential for enriching surgical trials in settings lacking diffusion imaging. Overall, the work demonstrates a practical path to non-invasively assess white matter injury from routine CT and to inform treatment decisions where advanced imaging is scarce.

Abstract

The preservation of the corticospinal tract (CST) is key to good motor recovery after stroke. The gold standard method of assessing the CST with imaging is diffusion tensor tractography. However, this is not available for most intracerebral haemorrhage (ICH) patients. Non-contrast CT scans are routinely available in most ICH diagnostic pipelines, but delineating white matter from a CT scan is challenging. We utilise nnU-Net, trained on paired diagnostic CT scans and high-directional diffusion tractography maps, to segment the CST from diagnostic CT scans alone, and we show our model reproduces diffusion based tractography maps of the CST with a Dice similarity coefficient of 57%. Surgical haematoma evacuation is sometimes performed after ICH, but published clinical trials to date show that whilst surgery reduces mortality, there is no evidence of improved functional recovery. Restricting surgery to patients with an intact CST may reveal a subset of patients for whom haematoma evacuation improves functional outcome. We investigated the clinical utility of our model in the MISTIE III clinical trial dataset. We found that our model's CST integrity measure significantly predicted outcome after ICH in the acute and chronic time frames, therefore providing a prognostic marker for patients to whom advanced diffusion tensor imaging is unavailable. This will allow for future probing of subgroups who may benefit from surgery.

From Diagnostic CT to DTI Tractography labels: Using Deep Learning for Corticospinal Tract Injury Assessment and Outcome Prediction in Intracerebral Haemorrhage

TL;DR

This study addresses the challenge of CST injury assessment in intracerebral haemorrhage when diffusion MRI is unavailable by training an nnU-Net to predict diffusion-based CST labels from diagnostic CT. The model uses paired CT and high-directional DTI data to achieve a mean relative to diffusion tractography and is applied to the MISTIE III trial dataset to derive CST-integrity metrics that significantly predict motor outcomes in both acute and chronic stages. The approach enables CT-only CST assessment, offering prognostic value and potential for enriching surgical trials in settings lacking diffusion imaging. Overall, the work demonstrates a practical path to non-invasively assess white matter injury from routine CT and to inform treatment decisions where advanced imaging is scarce.

Abstract

The preservation of the corticospinal tract (CST) is key to good motor recovery after stroke. The gold standard method of assessing the CST with imaging is diffusion tensor tractography. However, this is not available for most intracerebral haemorrhage (ICH) patients. Non-contrast CT scans are routinely available in most ICH diagnostic pipelines, but delineating white matter from a CT scan is challenging. We utilise nnU-Net, trained on paired diagnostic CT scans and high-directional diffusion tractography maps, to segment the CST from diagnostic CT scans alone, and we show our model reproduces diffusion based tractography maps of the CST with a Dice similarity coefficient of 57%. Surgical haematoma evacuation is sometimes performed after ICH, but published clinical trials to date show that whilst surgery reduces mortality, there is no evidence of improved functional recovery. Restricting surgery to patients with an intact CST may reveal a subset of patients for whom haematoma evacuation improves functional outcome. We investigated the clinical utility of our model in the MISTIE III clinical trial dataset. We found that our model's CST integrity measure significantly predicted outcome after ICH in the acute and chronic time frames, therefore providing a prognostic marker for patients to whom advanced diffusion tensor imaging is unavailable. This will allow for future probing of subgroups who may benefit from surgery.
Paper Structure (11 sections, 3 figures, 2 tables)

This paper contains 11 sections, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Data
  4. Diffusion tensor image processing
  5. Model training
  6. Clinical utility
  7. Results
  8. Discussion
  9. Conclusion
  10. Acknowledgments.
  11. Disclosure of Interests.

Figures (3)

  • Figure 1: Probabilistic tractography maps of the CST, superimposed onto the b0 volume of the DTI images for A) a patient with a haematoma involving the CST and B) a patient with a haematoma not involving the CST.
  • Figure 2: Three test cases for the CST segmentation model. The ground truth DTI CST tractography is shown in purple, the model prediction is shown in cyan. A) A patient with a haematoma not involving the CST and a high DSC (DSC = 71%) B) A patient with a haematoma not involving the CST (DSC = 52%) C) A patient with a haematoma involving the CST (DSC = 66%).
  • Figure 3: Three patients from the MISTIE III dataset with predicted CST and haematoma labels. A) A patient with a haematoma not involving the CST B) A patient with CST haematoma overlap but no split tract C) A patient with no CST haematoma overlap, but a split tract.