MR Fingerprinting for Epilepsy

Abstract Approximately 40% of patients with pharmacoresistant focal epilepsy undergoing presurgical evaluation have no identifiable lesions on the MRI. In these patients, detection of subtle epileptic lesion is directly linked to the odds of seizure-free outcome. Focal Cortical Dysplasia (FCD) is frequently the underlying pathology in MRI-negative epilepsies. FCD lesions are often missed on conventional clinical MRI, because they have subtle MRI characteristics and only differ slightly from normal cortex. The limitations of conventional structural MRI to detect FCD lesions can be addressed by MR Fingerprinting (MRF). This novel technology allows highly efficient quantitative MRI scan in a clinically feasible time on commonly used clinical scanners. The tissue property maps generated by MRF are specific to cell structure and micro-environment changes that are particularly relevant for FCD. Our previous R01 project centered on the development of a 3T MRF framework specifically designed for patients with epilepsy and FCD. In this renewal proposal, we aim to address critical steps necessary for full translation of the all-quantitative MRF framework to make a real clinical impact. In Aim 1, we will develop a fully automatic, fast, and traceable imaging pipeline for 3D MRF with enhanced motion robustness and GPU-based online reconstruction within 5-minute post-scan turnaround. This end-to-end workflow from acquisition to radiology reading room will make MRF available to every patient undergoing presurgical evaluation. Assisted by this workflow, we will conduct a neuroradiologist consensus review to assess the diagnostic yield of MRF. In Aim 2, we will develop robust MRF machine/deep-learning (ML/DL) models to predict the epileptogenic zone, utilizing data from patients who have achieved post-operative seizure freedom. We will delineate the epileptogenic zone based on the regions included in surgery shown by coregistered post-operative MRI. Validation will be performed on a withheld data set not used for the model training. Our central hypothesis is that the quantitative MRF framework will detect subtle tissue abnormalities with high epileptogenicity, and the removal of these abnormalities will be associated with favorable postoperative seizure outcomes. Our developments will lead to significant improvements in seizure outcomes for both adult and pediatric patients suffering from disabling pharmacoresistant epilepsy. Taking another leap, in Aim 3, we seek to leverage the unique availability of SEEG and MRF data in the same individual, to clarify the intricate relationship between MRF signals and their electrophysiological underpinnings on SEEG. We will deliver anatomo-functional atlases as an open-source tool, comprising two precisely co-localized layers of information: a quantitative MRF tissue property layer, and an SEEG spectra electrophysiological layer, with interrelationships between these two layers in both normal and FCD-affected brain regions. These developments will advance future imaging-pathological-electrophysiological studies in the normal brain, epilepsy and other neurological conditions.