Related Experiment Video
Updated: Mar 27, 2026
10:25
Brain Infarct Segmentation and Registration on MRI or CT for Lesion-symptom Mapping
Published on: September 25, 2019
49.7K
Comparison between voxel-based and subtraction methods for measuring diffusion-weighted imaging lesion growth after
Wajih Ben Hassen1, Marie Tisserand1, Guillaume Turc2
1Department of Radiology, Centre Hospitalier Sainte-Anne, Université Paris Descartes, Paris, France Centre de Psychiatrie et Neurosciences, INSERM S894, DHU Neurovasc, Paris, France.
Summary
Simple subtraction of diffusion-weighted imaging lesion volumes underestimates true infarct growth. Voxel-based coregistration provides a more accurate measure of infarct growth and better predicts patient outcomes after stroke treatment.
Area of Science:
- Neurology
- Radiology
- Medical Imaging
Background:
- Infarct growth (IG) is a critical surrogate endpoint in stroke therapeutic trials.
- Current methods often use simple subtraction of diffusion-weighted imaging (DWI) lesion volumes, which can underestimate true infarct growth due to lesion reversal.
Purpose of the Study:
- To quantify the difference between true infarct growth measured by voxel-based coregistration and approximated infarct growth using volume subtraction.
- To assess the impact of these methods on predicting three-month unfavorable outcomes.
Main Methods:
- Retrospective analysis of 322 stroke patients undergoing MRI before and after i.v. thrombolysis.
- Defined voxel-based infarct growth (IGvoxel-based) as DWI2 signal changes not overlapping with DWI1.
- Compared IGvoxel-based with simple volume subtraction (IGsubtracted) and their predictive accuracy for unfavorable outcomes (mRS ≥ 2).
Main Results:
- IGsubtracted significantly underestimated IGvoxel-based (9.52 mL vs. 16.98 mL).
- In 27% of patients with apparent lesion stability or shrinkage by subtraction, voxel-based analysis revealed significant infarct growth (≥5 mL).
- IGvoxel-based demonstrated superior prediction of unfavorable outcomes compared to IGsubtracted (c-statistic 0.86 vs. 0.82).
Conclusions:
- Simple volume subtraction of DWI lesions masks significant infarct growth in a notable percentage of patients early after thrombolysis.
- The voxel-based method, though more time-consuming, offers a more accurate assessment of infarct growth and has implications for clinical trial endpoints.
