The YeB Atlas is a three-dimensional and deformable atlas of the basal ganglia, built from a post-mortem specimen that went through MR imaging and histological stainings. It has been created at the Pitié-Salpêtrière Hospital by a multidisciplinary team between 2002 and 2006. The project was lead by Jerome Yelnik (neuroanatomist), Carine Karachi (neurosurgeon) and Eric Bardinet (neuroimaging engineer), and was originally developed with the help of a Medtronic Inc. research grant.

Motivation to build this atlas came from the field of Deep Brain Stimulation (DBS), where there was a need to be able to accurately localise targets in the basal ganglia (subthalamic nucleus or STN, VIM nucleus of the thalamus, internal globus pallidus or GPi,...). These structures were barely visible on MR images, and even if MR imaging has evolved in the last 10 years (sequences, magnets), complete visualisation of the STN, the VIM or the GPi is still challenging. Moreover, the need for accuracy has evolved, and today the neurosurgeon not only asks for the structures, but also for functional territories inside these structures, leading to a millimetric accuracy need.

Therefore, we decided to build an atlas that would comprise MR imaging and histology of the same specimen, as histology allows direct observation of brain tissue, and staining that can reveal specific properties.

We also had to design a deformation strategy that would allow the atlas to be adapted to any brain by MR registration between the atlas and patient’s MR data. This strategy was developed in order to take into account the histological information embedded in the YeB atlas.

Even if the YeB atlas has been developed for DBS, it offers a unique representation of the basal ganglia, including functional subterritories (limbic, associative and sensorimotor), that can be of interest for many applications.

Atlas deformation on patients The atlas can be adapted to any subject through a dedicated deformation strategy based on global and local linear registrations. The method is described in (Bardinet et al, JNeuroSurg 2009). Accuracy has been évaluated by various means, including qualitative standardized visual checking on T1w and T2w MR images of the patient (some structures of the atlas are at least partially visible on MR, such as the caudate nucleus and putamen on T1, as well as partially the thalamus or optic tract, and STN, SN and Red Nucleus as hyposignals on T2, that do not exactly correspond to real anatomy). A quantitative evaluation has also been conducted by comparing anatomical labeling of electrophysiological recording sites annoted while micro-electrodes step down to the chosen target, and position of the corresponding structures given by the atlas adapted to the patient (e.g. electrophysiological STN sites vs. atlas-based STN).