Novel method for classifying BOLD response mechanisms in focal epilepsy

Riera Diaz, Jorge

Abstract

For the up to 933,000 epileptic Americans who have intractable epilepsy, neurosurgery is the most feasible cure. Such surgery relies upon accurate localization for precise resection of seizure-onset foci, but current approaches yield success rates of only ~50-60%. Various methods are used, but to increase success rates, improved approaches are needed for localizing these zones. Brain areas generating interictal epileptic discharges (IEDs), known as irritative zones, are associated with seizure origin. IEDs can be detected by EEG, which is a key clinical tool for pre-surgical evaluation, but EEG alone has insufficient spatial resolution for localizing IED origin zones and image-uniqueness problems that preclude discrimination of actual from spurious IED-evoked current sources. By combining fMRI with EEG in concurrent recordings, brain areas exhibiting IED-evoked blood oxygen level-dependent (BOLD) responses, an indirect indicator of neuronal activation, can be localized with high spatial resolution, and an established standard clinical protocol enables localization of seizure-onset foci by this technique. However, a technical problem limits application of this cost- advantageous, widely-available approach. Central to the clinical EEG-fMRI mapping protocol are rigid and non- interpretable hemodynamic response functions (HRFs) that permit characterization of local neurovascular coupling and representation of fMRI by IED-dependent regressors. Currently, HRF computation assumes that IEDs cause local cerebral blood flow increases (BOLD activations), but IEDs also evoke atypical BOLD responses, e.g. deactivations, in certain brain regions, confounding analysis and thus barring effective EEG- fMRI use in many patients. To overcome this barrier and enable increased, more accurate use of EEG-fMRI for seizure-onset zone mapping, our aims are to: 1) identify the biophysical mechanisms of IED-evoked BOLD responses in epilepsy; 2) determine their cellular substrates; and 3) develop an automated image-guided method for classifying them by origin type. Our central hypothesis, supported by our preliminary findings, is that 3 mechanisms, alone or in combination, contribute to the emergence of atypical BOLD responses in the epileptic brain together with classical hyperemic responding areas: 1) unexpected IED-mediated interruptions of the resting state network; 2) blood stealing/leaking effects; and 3) local vascular/metabolic decoupling. Our preliminary findings further indicate that it is feasible to discriminate BOLD response mechanisms by analyzing EEG-fMRI-derived data, which will permit effective localization of foci. To solve the complex neurophysiological and computational problems involved, we established a unique multi- disciplinary/institutional collaboration between Florida International Univ. and Yale Univ. teams, with interactions with 2 South-Florida hospitals. We will use a novel, translationally-relevant rat epilepsy model and innovative strategies for EEG-fMRI analysis. Validating our ideas will permit translation and widespread, field-advancing clinical implementation of cost- effective/noninvasive EEG-fMRI imaging modality, improving surgical outcomes in human epilepsy patients.

Public Health Relevance

Novel method for classifying BOLD deactivation mechanisms in focal epilepsy Focal refractory epilepsy can be treated effectively by neurosurgical ablation, but in many patients, inconsistencies prevent localizing seizure-onset zones to target for removal, precluding successful therapeutic neurosurgery. EEG-fMRI is a cost-effective, advantageous technique for mapping epileptogenic sites, but is often confounded by atypical blood-oxygen-level-dependent (BOLD) responses that occur in many brain areas with high epileptogenic profiles, preventing analysis. We will solve this problem by analyzing the neural-vascular and neural-metabolic mechanisms that cause different types of BOLD responses in brains of epilepsy patients, using advanced mathematical and neurophysiological methods in a novel rat model of chronic epilepsy. We will thereby create a new automated analysis method that permits effective EEG-fMRI use to precisely localize lesions for ablation in epileptic patients.