Abstracts

Rationale: At present, presurgical functional MRI (fMRI) is primarily used for mapping of the eloquent brain to be spared in surgery in patients with intractable epilepsy. Its application as a localization tool to identify paroxysmal areas has been intensively investigated. But often simultaneously acquired EEG for temporal information is required. Therefore, a simpler method that is more readily implementable in the clinical settings is desired. Independent component analysis is a data-driven approach that does not require knowledge of the precise timing of the epileptic events or the exact hemodynamic response function (HRF). However, a straightforward framework to select the independent components that are related to epileptic activity remains to be developed. Methods: To localize hemodynamic changes in patients with focal epilepsy we performed independent component analysis (ICA) on the fMRI data collected during presurgical evaluation. A set of IC selection criteria was developed basing on two aspects 1) Spatial pattern of noisy components: Components with a large portion of signal located outside of the brain is likely to be noise or not of neurological origin. 2) Spatial pattern of epilepsy related components: Epileptic activities in focal epilepsy cases are clinically important, and tend to occur unilaterally or asymmetrically. Therefore, components of hemodynamic changes related to epilepsy are likely to have a unilateral or asymmetrical spatial distribution. Activation maps of identified components from patient data were compared with postoperative MRI and surgical report to access accuracy of the method in the clinical population. Furthermore, to evaluate the proposed method with known cortical activation patterns, we simulated lateralized focal activity using healthy subjects performing hand movement tasks.Results: Nine out of the ten patients studied received resective surgery. In 8 out of the 9, the activation maps of identified ICs were concordant with surgical resection. In the patient who did not receive surgery, the fMRI results showed altered resting state networks which may explain why the patient was not a suitable candidate for surgical treatment. In the simulation study, in all 10 healthy subjects studied, we identified components with concordant spatial and temporal features as expected from unilateral hand motor tasks. Conclusions: These results suggest the lateralization and localization value of fMRI in presurgical evaluation for patients with intractable focal epilepsy. Our proposed IC selection algorithm yielded activation maps revealed areas overlapping with patients surgical resection regions, which was informed by invasive procedures such as intracranial recording. The merits of our method were further demonstrated in simulated focal activities with motor tasks. This method has potential to be incorporated in current presurgical evaluation procedures for localization of epileptic foci.