Quantification and visualization of three-dimensional inconsistency of the globus pallidus internus in the Schaltenbrand-Wahren brain atlas.

The major shortcomings of the Schaltenbrand-Wahren (SW) brain atlas include 3-dimensional (3D) inconsistency and spatial sparseness. This work quantifies and visualizes 3D inconsistency of the globus pallidus internus (GPi), a stereotactic target for the treatment of Parkinson's disease, dystonia and Huntington disease. The GPi 3D models 3D-A, 3D-C and 3D-S are reconstructed from the SW axial, coronal and sagittal microseries, respectively, by applying a shape-based (Nonuniform Rational B Splines) method. All three 3D models, placed in the SW coordinate system, are compared quantitatively in terms of location (centroids), size (volumes), shape (normalized eigen values), orientation (eigen vectors) and mutual spatial relationships (overlaps and inclusions). The analysis is done in 3D within each orientation and across them. The reconstructed 3D GPi models substantially differ in location, size and inclusion rate. The centroid of 3D-C is located more medially (15.6 mm) than those of 3D-A (17.5 mm) and 3D-S (18.2 mm), and that of 3D-A more ventrally (-2.3 mm) than those of 3D-C (-0.1 mm) and 3D-S (-0.4 mm). 3D-S has the smallest volume (347.3 mm3); 3D-A is 1.18 and 3D-C 1.85 times larger. The highest inclusion rate is for 3D-S (54.3 and 56.3%) and the lowest for 3D-C (28.8 and 30.6%). A smaller variability is observed in shape, orientation and overlap size (196.8, 196.1 and 185.5 mm3). To get a better correspondence between 3D-C and 3D-S, the coronal microseries were scaled laterally by 1.1667. This results in a substantial improvement of the inclusion rate of 3D-S (87.9%), though raising the volume mismatch to 2.16. The GPi in the SW atlas has a substantial 3D inaccuracy within each orientation and across them. Therefore, absolute and direct reliance on the original atlas is unsafe, and this atlas has to be used with great care and understanding of its limitations. As matching various SW microseries by global scaling is not feasible, we propose the target-dependent scaling based on structure centroid matching.