Combining MR Fingerprinting with Morphometric MRI Analysis to Reduce False Positives for Focal Cortical Dysplasia Detection
Abstract number :
3.25
Submission category :
5. Neuro Imaging / 5A. Structural Imaging
Year :
2022
Submission ID :
2205040
Source :
www.aesnet.org
Presentation date :
12/5/2022 12:00:00 PM
Published date :
Nov 22, 2022, 05:27 AM
Authors :
Zheng Ding, MS. – Cleveland Clinic; Siyuan Hu, BS. – Biomedical Engineering – Case Western Reserve University; Ting-Yu Su, MS. – Epilepsy Center, Neurological Institute – Cleveland Clinic; Joon Yul Choi, PhD – Epilepsy Center, Neurological Institute – Cleveland Clinic; Xiaofeng Wang, PhD – Quantitative Health Science – Cleveland Clinic; Ken Sakaie, PhD – Imaging Institute – Cleveland Clinic; Hiroatsu Murakami, MD – Epilepsy Center, Neurological Institute – Cleveland Clinic; Hans-Juergen Huppertz, PhD – Swiss Epilepsy Clinic – Klinik Lengg AG, Zurich; Ingmar Blümcke, PhD – Neuropathology – University of Erlangen; Stephen Jones, MD, PhD – Imaging Institute – Cleveland Clinic; Imad Najm, MD – Epilepsy Center, Neurological Institute – Cleveland Clinic; Dan Ma, PhD – Biomedical Engineering – Case Western Reserve University; Zhong Irene Wang, PhD – Epilepsy Center, Neurological Institute – Cleveland Clinic
Rationale: Focal cortical dysplasia (FCD) is one of the most common pathologies underlying MRI-negative epilepsy. Detection of FCD can often be aided by voxel-based morphometric analysis for an increased yield (Huppertz et al, Epilepsia 2008). However, false positives (FP) are commonly seen in post-processing results. Interpretation of these results, therefore, still requires substantial efforts and expertise. MR fingerprinting (MRF) is a novel technique to measure quantitative tissue properties not attainable in conventional MRI, and has recently demonstrated effectiveness in distinguishing epileptic lesions. In this study, we aimed to improve automatic FCD detection by combing MRF with morphometric MRI analysis to reduce FP findings, with the hypothesis that true positive (TP) clusters would have significantly higher MRF T1 and T2 values than FP clusters, due to the underlying pathology.
Methods: 3D whole-brain MRF scans were acquired from 47 healthy controls (HCs) and 19 patients with pathologically confirmed FCD (8 type IIb, 6 type IIa and 5 mMCD) on a 3T SIEMENS Prisma scanner (1 mm3 isotropic voxels, Ma et al, JMRI 2019). MRF T1 and T2 maps were reconstructed by matching the signal evolution pattern to a pre-defined dictionary. 3D Lesion labels were created on synthetic T1w images generated from MRF T1 maps. All images were registered into the MNI space. Normalized T1 and T2 were calculated using values from co-registered voxels across all HCs. MAP18 was used to perform morphometric analysis and to generate lesion clusters on FCD probability maps (David et al, Epilepsia 2021), based on T1w MPRAGE from the same scan. The average MRF T1 and T2 values were calculated in each cluster for GM and WM separately. Clusters that overlapped with the manual lesion labels were considered TP. Two-sample t-tests weighted by volume were performed to compare the MRF signal characteristics between TP and FP clusters. A shallow neural network was built to separate TP and FP clusters, using MRF data and cluster volume as input.
Results: Figure 1 shows example images from a patient with FCD IIb. As depicted in Figure 2a-b, in GM, the average T1 was significantly higher for TP clusters (mean±SD=1444±126ms) than FP clusters (1256±189ms, p< 0.001). The average T2 was also significantly higher for TP (60.5±7.9ms) than FP clusters (51.6±12.4ms, p< 0.001). In WM, the average T1 was significantly higher for TP (1005±96ms) than FP clusters (916±94ms, p< 0.001). The average T2 was also significantly higher (60.5±7.9ms) for TP than FP clusters (51.6±12.4ms, p< 0.05). Normalized T1 and T2 at the homotopic cortices showed consistent findings (Figure 2c-d). The 10-fold cross-validation of neural network classification yielded an accuracy of 90.5% for separating TP/FP clusters._x000D_
Neuro Imaging