Abstracts

Rationale:
Periventricular nodular heterotopias (PVNHs) are malformations of cortical development that are frequently associated with pharmacoresistant epilepsy. Invasive evaluation with stereotactic-electroencephalography (SEEG) is considered a useful step of the presurgical evaluation for patients with PVNH lesions, for providing an estimation of the epileptogenic zone. Currently, clinical MRI has been used to visualize the PVNH lesions; however, conventional MRI could not manifest any signal differences to distinguish the epileptic PVNH lesions from the non-epileptic ones. In this study, we investigated quantitative T1 and T2 values as potential biomarkers of tissue properties in epilepsy patients with PVNH using a novel quantitative MRI method, magnetic resonance fingerprinting (MRF).
Method:
A whole-brain MRF scan was acquired from 5 patients with PVNH in a 3T Prisma scanner. Quantitative T1 and T2 maps were generated based on direct matching of the data to a predefined dictionary. MRF T1-weighted (T1w) images were synthesized from the T1 and T2 maps generated by MRF. Clinical 3D T1w MPRAGE and CT scans were acquired for localization of SEEG electrode contacts. The grouping of seizure onset zone (SOZ) and non-SOZ of nodule lesions was based on the official consensus at the epilepsy patient management conference as shown in Figure 1 (red dots on SEEG are SOZ). To extract MRF values from SOZ and non-SOZ, the MRF T1w was registered to the clinical T1w using SyN. Then the warping information was applied to the T1 and T2 maps. Masks for PVNH lesions are generated as follows. First, the clinical T1w was segmented to obtain a cerebral spinal fluid (CSF) mask, and a whole-brain mask was subtracted by the CSF mask. Second, to precisely remove voxels of white matter (WM) at unclear boundaries between WM and nodule lesions, histograms of T1 and T2 maps were fit to Gaussian curves to calculate mean and standard deviation of WM.  The mean plus 3 times of standard deviation value was used as a threshold to remove WM voxels in the mask. We then used circular ROIs with a radius of 3 mm around the center of the lesions to represent each PVNH lesion. For statistical analysis, T1 and T2 values in SOZ were compared to those in non-SOZ using a T-test.
Results:
MRF images and SEEG electrodes for five patients are shown in Figure 1, revealing higher T1 values in SOZ (red dots) than non-SOZ (blue dots). Figure 2 shows scatter plots of T1 and T2 values of SOZ and non-SOZ (A) and statistical analysis (B) for five patients. T1 voxels at SOZ and non-SOZ formed different clusters (Figure 2A). T1 is significantly higher in the SOZ voxels than in that of the non-SOZ voxels for all patients (p < 0.001, 14% higher in SOZ) whereas mean T2 is not consistent across patients (Figure 2B). Further technical development is being performed on T2 maps.
Conclusion:
MRF has the potential to provide additional information for epilepsy patients with PVNH lesions, providing lesion characterization for nodules located in the SOZ vs. non-SOZ. The additional information may improve diagnostic accuracy and assist in the surgical planning of this challenging patient cohort.
Funding:
:This study is supported by NIH R01 NS109439.