Abstracts

Rationale: Defining a procedure for delineating and naming brain structures is important in the analysis of neuroimaging data. The demarcation of regions in an individual brain is useful in guiding diagnostic and management decisions, for instance, in epilepsy surgery where localization of intracranial EEG electrodes may guide operative resection. Current limitations in brain localization stem from anatomical differences between individuals. Common neuroanatomical terminology is used to communicate brain regions and locations, for instance a gyrus, sulcus or neurovascular structure. The resolution of this brain parcellation is low and defines areas on the cortex to the gyrus, for instance ‘superior frontal gyrus’. This leads to imprecision in the definition of brain regions. In order to more accurately and precisely facilitate communication among researchers we present an alternative approach that parcellates the cortex to the sub-gyral level, meaning that sections of a single gyrus can be defined. Methods: We set out to create a terminology for defining regions on the human cerebral cortex to the nearest centimeter while accounting for interindividual differences in anatomy. We used the MNI ICBM 152 reference brain to identify high fidelity anatomical landmarks. Any structures which were labelled in Duvernoy’s neuroanatomical atlas and were also demonstrable on the MNI brain were eligible for use as landmarks or boundaries, since there was a proven high degree of fidelity in those structures. A hierarchical scheme was used to parcellate the cortical surfaces. This scheme may be used to individualize the parcellation for each person or institution. Gross cortical boundaries were defined, including the Sylvian fissure, central sulcus and occipitotemporal incisure. An outline of our atlas was drawn using 2D images of the lateral, medial and basal brain surfaces. We aimed for an approximate parcel size of 1cm in MNI space, measured in the anterior-posterior dimension. Results: The cerebral cortex was segmented into 487 discrete cortical parcels, 106 gyral areas and 32 lobar regions. The parcellation was presented in 2D images of each cortical surface, with the lateral and medial cortices provided as examples (FIG 1 and 2). Each parcel had its own unique code. Our parcellation methodology was tested on an epilepsy surgery team, consisting of an attending, a fellow and researcher. Team members were asked to independently localize electrode contacts in patient MRI space using the parcellation. 17 trials were performed using 17 electrode contacts. In 11 trials (64.7%), all raters agreed on a single 1cm section of cortex. In 4 trials (23.5%), raters were split between two adjacent 1cm sections of cortex. The maximum inter-rater variance was 3cm, where three adjacent parcels were identified in one trial.  Conclusions: We present a comprehensive cortical parcellation method developed using neuroanatomical landmarks and features which are demonstrable between individuals. It is precise to the nearest centimetre and has high fidelity when used to define locations while taking into account unique features at the individual level. Our parcellation segments single gyri into 487 discrete parcels, each of which is encoded by a unique three- or four-letter code along with a description of the hemisphere and surface of interest. We tested the parcellation for inter-rater agreement. The parcellation may be used to communicate specific areas in the brain of an individual and as a common language for groupwise research. Funding: No funding