Epilepsy is a chronic disabling disorder affecting over 1% of the population who suffer higher rates of physical, mental and social disability and associated costs of over $10,000 per year. Neurosurgery is an invaluable treatment that can often be curative and superior to alternatives (Wiebe et al., 2001). It can have cognitive risks, however, including language and memory difficulties. My lab's research mission is to develop standardized, evidence-based protocols to accurately predict patients’ real-world functioning after epilepsy surgery.
The ability to communicate through expressive and receptive verbal language is one of the most important human functions. The risk of language decline has historically been estimated using Wada testing, which is highly invasive. Functional Magnetic Resonance Imaging (fMRI) can predict individuals' risk of language decline, and is entirely non-invasive, but requires highly standardized procedures. Anecdotally, the clinical fMRI protocols across surgical programs vary markedly. In a pair of survey studies we provided the first detailed pictures of –
The questions clinicians ask of language fMRI, and how they interpret the findings (open).
The tasks, training, and analytic approaches that are most often used for presurgical fMRI (open).
These data show that the interpretation and execution of fMRI in the clinic varies, which will lead to varying accuracy in predicting decline. Separately, they suggest that many clinicians already use fMRI maps to guide surgical margins and that the majority recognize two language areas when they do so, while at least four further language regions are identifiable based on findings from fMRI and direct cortical stimulation data.
We defined a new approach to estimate the location of six known language areas, including Broca’s & Wernicke’s areas, Exner’s Area, Supplementary Speech Area, Angular Gyrus, and the Basal Temporal Language Area (Benjamin et al. 2017). This method relies on a set of three different language tasks and a trained clinician selecting and combining data from the tasks to identify these regions. In 22 epilepsy patients, this method was reliable when used by different clinicians (78% overlap) and identified the Wada-defined language-dominant hemisphere with accuracy equivalent to the best published methods (85% of cases). Further, activation consistent with all six language regions was identified consistently, and more often than in an automated analysis of the same data.
An ongoing set of studies in our lab are focused on better understanding the properties of these language regions, the strengths and weaknesses of its ability to predict decline, and this approach's comparability to existing approaches. The tasks themselves, as well versions of other excellent, validated tasks (e.g., Bonelli et. al., 2012; Sabsevitz et al. 2003) are available for download here.
Vision: Mapping the optic radiations
Another goal of surgical planning is the mapping of the visual system's optic radiations. This vital structure is extremely hard to map with MRI due to its circuitous route. Researchers have largely focused on improving MRI acquisition methods to better map the structure. This study showed that a prominent idea in the field, that the original work of Meyer in 1908 showed that the radiations include three structures, is incorrect. It also showed that existing methods for finding this structure are highly variable, and showed that a combination of these works more effectively. This work was supported by, and completed under the guidance of, Professor Simon Warfield, PhD within his Computational Radiology Laboratory.
Memory: The temporal lobe, time and space
A further goal in planning surgical treatment in epilepsy is understanding which brain structures in the temporal lobe support memory, and how well they are functioning. This allows the effects of different forms of surgery to be predicted during planning. The key to this has been seen as understanding which hemisphere’s mesial temporal lobe (MTL) is critical in verbal memory. Curiously, we remain unable to map memory for this purpose using fMRI. In this study we studied which temporal lobe structures are important in encoding the "space" and "time" of experience; for instance, so that you can remember the order of events that occurred over a few minutes, and where everything (people, objects, etc.) where during this time. The results from this functional MRI study provided further evidence that a key function of the temporal lobe is gluing information together for storage in memory, and suggest that different areas in the middle portion of the temporal lobe ("medial temporal lobe") achieve this in different ways.
In current work we are focused on understanding how different profiles of verbal memory performance relate to memory changes and real-world functioning after surgery.