Authors :
Presenting Author: Sydnie Hom, BA – Cooper Medical School of Rowan University
Alex Prusky, BA – Sidney Kimmel Medical College
Shaghayegh Poursabbagh, MA – Rutgers University
Mahdi Alizadeh, PhD – Thomas Jefferson University
Ron Gefen, MD – Cooper University Healthcare
Khuram Kazmi, MD – Cooper University Healthcare
Islam Fayed, MD – Cooper University Health Care
Joseph Ifrach, DO – Cooper University Healthcare
Todd Siegal, MD – Cooper University Healthcare
Evren Burakgazi-Dalkilic, MD – Cooper Univeristy Hospital
Melissa Carran, MD – Cooper University Healthcare
Eric Nagele, DO – Cooper University Healthcare
Rationale: In epilepsy, neurovascular uncoupling (NVU) can cause a mismatch between neuronal activity and the BOLD fMRI signal, potentially impairing presurgical language cortex localization. Prior research has validated an amplitude of low frequency fluctuations (ALFF) based correction of resting state fMRI (rsfMRI) to correct for NVU in brain tumors, but this correction has yet to be applied to task fMRI or epilepsy (Agarwal et al., 2019). Here, we apply an ALFF correction to task fMRI in epilepsy patients undergoing presurgical planning for temporal lobectomy.
Methods: fMRI data from seven patients with refractory epilepsy (5 female, 2 male; age range: 18-48 years old) referred for temporal lobectomy planning were included. Seizure types included focal seizures with impaired awareness (n = 4) and focal to bilateral tonic-clonic seizures originating in the temporal lobe (n = 3). To localize the language cortex, each patient underwent rsfMRI and two repetitions of a sentence-completion language task. SPM12 was used to complete preprocessing and generate Z-score activation maps from the general linear model (GLM) comparing task and rest. ALFF maps were computed from bandpass-filtered (0.01-0.08 Hz) resting-state data. Candidate voxels were defined as temporal lobe voxels with Z-scores >25% of the activation mapping as a percentage of local excitation (AMPLE) maximum. For each candidate voxel, corrected Z-scores were computed using a correction factor derived from normalizing ALFF values:
Zcorr = Zmeas × (mALFF⁄ iALFF),
where Zmeas is the original Z-score, Zcorr is the corrected Z-score, mALFF is the mean ALFF in the contralateral temporal ROI (from the Automated Anatomical Labeling atlas), and iALFF is the ALFF value of the candidate voxel. The number of voxels exceeding 50% of the AMPLE maximum were compared pre- and post-correction.
Results: ALFF corrected activation maps showed a reduction in suprathreshold voxels in the ipsilesional temporal lobe during the task (Fig. 1). In the right temporal lobe, average voxel count decreased significantly from pre-correction to post-correction (mean difference = 180.1 voxels, 95% CI [108.3, 251.9], p = 0.00088). In the left temporal lobe, there was also a significant decrease in suprathreshold voxels before and after correction (mean difference = 122.5 voxels, 95% CI [45.2, 199.8], p = 0.0082.). Figure 2 shows pre- and post-correction suprathreshold voxels in the seven patients.
Conclusions: When applied to task fMRI of patients with epilepsy, the ALFF correction revealed significant differences in voxel activation in the temporal lobe. Integration of an ALFF correction may improve functional localization in regions at risk for NVU and assist in surgical planning.
Funding: No funding to report.