Introduction: The Acute Stroke Imaging Roadmap III identifies structural distortion due to vasogenic edema and hemorrhage as a research priority for defining final infarction. Non-linear registration (NLR) of a follow up scan to an undistorted presenting scan could correct for distortions due to edema, hemorrhage or atrophy, achieving this goal. In addition, the difference between the volume of infarction following NLR and the volume following a rigid body registration (RBR) reflects the degree of anatomical distortion. In this study we evaluate this technique to correct for subacute edema at different timepoints, and generate a metric to quantify brain swelling at these times. We determine whether early edema at 24 hours predicts edema at 1 week.
Methods: Patients with non-lacunar ischemic stroke were recruited into a MRI study. Patients had structural T1-weighted, T2-weighted FLAIR and diffusion-weighted imaging (DWI, b=1000/0) at presentation, 24hrs, 1wk and 1mo. Infarction was defined manually at 24hrs using DWI, and at 1wk and 1mo using FLAIR image by 2 raters. To quantify edema, both NLR warps and RBR matrices were generated between the T1 images at each timepoint to the presenting T1 scan. Infarct masks were transformed to presenting image space using RBR and NLR, and the relative difference in volumes used to quantify the Edema Metric (EM).
Results: 34 patients were recruited into the study. NLR corrected for distortions due to edema and hemorrhagic transformation at the 24hr and 1wk timepoints. The EM at 24 hours, 1 week and 1 month were 17.7% (p=0.009), 26.5% (p=0.02), and 7.1% (p=0.05) respectively for the manually defined infarct masks. EM at 24 hours predicted edema at 1 week (r2=37%, p=0.009), but not at 1 month (r2=3%, p=0.6).
Conclusions: NLR provides an opportunity to correct for edema at subacute timepoints and by comparing infarct volumes to those following RBR provides a measure of edema. The EM quantifies the contribution of edema at 24hrs and 1wk, and potentially allows the selection of patients at 24hrs who are likely to develop significant swelling at 1 week. The EM may also be useful in stroke trials to quantify the effect sizes of treatments aimed at minimizing edema in stroke.