Biometry extraction and probabilistic anatomical atlas of the anterior Visual Pathway using dedicated high-resolution 3-D MRI

Abstract

Anterior Visual Pathway (aVP) damage may be linked to diverse inflammatory, degenerative and/or vascular conditions. Currently however, a standardized methodological framework for extracting MRI biomarkers of the aVP is not available. We used high-resolution, 3-D MRI data to generate a probabilistic anatomical atlas of the normal aVP and its intraorbital (iOrb), intracanalicular (iCan), intracranial (iCran), optic chiasm (OC), and tract (OT) subdivisions. We acquired 0.6 mm3 steady-state free-precession images from 24 healthy participants using a 3 T scanner. aVP masks were obtained by manual segmentation of each aVP subdivision. Mask straightening and normalization with cross-sectional area (CSA) preservation were obtained using scripts developed in-house. A probabilistic atlas (“aVP-24”) was generated by averaging left and right sides of all subjects. Leave-one-out cross-validation with respect to interindividual variability was performed employing the Dice Similarity Index (DSI). Spatially normalized representations of the aVP subdivisions were generated. Overlapping CSA values before and after normalization demonstrate preservation of the aVP cross-section. Volume, length, CSA, and ellipticity index (ε) biometrics were extracted. The aVP-24 morphology followed previous descriptions from the gross anatomy. Atlas spatial validation DSI scores of 0.85 in 50% and 0.77 in 95% of participants indicated good generalizability across the subjects. The proposed MRI standardization framework allows for previously unavailable, geometrically unbiased biometric data of the entire aVP and provides the base for future spatial-resolved, group-level investigations.