Authors :
Presenting Author: Zheng Ding, MS – Cleveland Clinic Foundation
Ting-yu Su, MS. – Epilepsy Center, Neurological Institute – Cleveland Clinic Foundation; Siyuan Hu, BS. – Biomedical Engineering – Case Western Reserve University; Joon Yul Choi, PhD – Epilepsy Center, Neurological Institute – Cleveland Clinic Foundation; Ken Sakaie, PhD – Imaging Institute – Cleveland Clinic Foundation; Hiroatsu Murakami, MD, PhD – Epilepsy Center, Neurological Institute – Cleveland Clinic Foundation; Ingmar Blümcke, MD – Neuropathology – University of Erlangen; Stephen Jones, MD, PhD – Imaging Institute – Cleveland Clinic Foundation; Imad Najm, MD – Epilepsy Center, Neurological Institute – Cleveland Clinic Foundation; Dan Ma, PhD – Biomedical Engineering – Case Western Reserve University; irene Zhong Wang, PhD – Epilepsy Center, Neurological Institute – Cleveland Clinic Foundation
Rationale:
The ‘tip of the iceberg’ phenomenon of radiographic lesions is a major cause of recurring seizures after epilepsy surgery. Tissue surrounding radiographic lesions can be epileptogenic but are not visible on MRI (Rosenow 2001). T1w/T2w is a semi-quantitative image analysis technique proposed for mapping myelin content (Glasser 2011). Magnetic resonance fingerprinting (MRF) offers a novel approach for quantitative multiparametric tissue property measurements not achievable by conventional MRI (Ma 2013). In this study, we aim to use these two approaches to provide multiparametric quantitative measures to depict pathological gradients of FCD non-invasively.
Methods:
Fifteen patients (nine male/six female, mean age 28 ± 15 years) were included. FCD subtypes included seven Type IIb, four Type IIa, and four mMCD according to updated ILAE classification guidelines (Najm 2022). 3D T1w MPRAGE and T2w SPACE sequences were acquired on a Siemens 3T Prisma scanner with a 32-channel head coil at 0.47×0.47×0.94 mm resolution. A 3D whole-brain MRF sequence was acquired, and dictionary-based reconstruction was performed. A 3D region of interest (ROI) was created for each lesion. T1w/T2w images were generated by dividing the T1w images by the T2w images, then applying a binary brain mask thresholded at the average of the entire T1w images. MRF T1 and T2 images, as well as lesion label ROIs, were registered to the T1w images using Advanced Normalization Tools. To investigate the gradient of each FCD lesion, lesion label ROIs were dilated isotopically to include cortical GM and WM around the lesion. Dilation was performed twenty times and the net expansion volume after each dilation was considered as a voxel group at a certain distance away from the epicenter (Figure 1, red plots). The lesion label was also eroded inward isotopically, and the eroded volume in each iteration was calculated for ROI analysis (Figure 1, blue plots). GM and WM voxels were distinguished using probability threshold of 0.5 on FSL segmentation maps and analyzed separately. To summarize trends across subtypes, we grouped the voxels by volume percentage of the lesion ROI. T1w/T2w, MRF T1, and MRF T2 were normalized to 0~1 by the max/min values of each measure for each patient.
Results:
As shown in Figure 2, for both GM and WM, T1w/T2w was lowest at the lesion center likely reflecting demyelination, and showed a consistent increasing trend moving away from the lesion center. MRF T1 and MRF T2 were highest at the lesion center and showed a decreasing trend from the center towards outside the lesion label, likely reflecting the gradient of tissue/neuronal damage. The trend stabilized beyond the lesion ROIs, which could provide quantitative support of the "tip of iceberg" phenomenon. Different subtypes IIa, IIb, and mMCD showed different rates of change in T1w/T2w, MRF T1, and MRF T2 before stabilizing, with mMCD demonstrating a slower rate of change, consistent with the pathological severity of the subtypes.
Conclusions:
Multiparametric quantitative mapping using T1w/T2w and MRF features may be used to aid the identification of FCD subtype and determination of surgical resection margin.
Funding:
NIH R01 NS109439