Introduction: In chronic stroke, functional MRI (fMRI) is used to map residual motor networks. Standard preprocessing masks the stroke lesion, excluding both the infarct cavity and adjacent T2 hyperintense tissue to correct for tissue displacement. However, this may eliminate regions with abnormal T2 signal that retain metabolic viability, which could support recovery. We implemented a double masking approach that distinguishes the T2 hypointense infarct core from the T2 hyperintense tissue, and evaluated whether this method reveals task-related activity in the hyperintense region that may need further study. Methods: Participants with chronic ischemic stroke underwent wrist-flexion fMRI before a neuromodulation study. Participants performed 8 trials of visually cued wrist flexion followed by rest (TR = 3 s, 100 volumes, 3.0T Siemens scanner). High-resolution T1-weighted and T2-FLAIR images were manually segmented in ITK-SNAP to delineate the cystic infarct and the surrounding hyperintense area. Preprocessing in SPM25 included realignment, co-registration, segmentation, and smoothing with a 6 mm Gaussian kernel. Lesion-aware confound regressors were incorporated into the first-level GLMs, including signal from the infarct core, six head motion parameters, and volumes exceeding 1 mm framewise displacement. Wrist flexion blocks were modeled using the canonical HRF with a high-pass filter of 128 s. Small-volume correction was applied within the T2 hyperintense region. Results: Task-related activation was observed in the T2 hyperintense area in a subset of participants. A significant activation cluster emerged within the T2 hyperintense region during wrist flexion (pFWE=0.005, qFDR=0.025), detectable only when the cystic infarct core was separated from surrounding tissue using the double masking approach. Conclusions: Despite challenges in interpreting BOLD signals within T2 hyperintense regions due to altered neurovascular coupling and T2* signal changes, further investigation is warranted. Double masking preserves potentially viable tissue for analysis and may improve detection of residual network activity. This approach could inform future strategies for stroke rehabilitation.
Sudarshan et al. (Thu,) studied this question.