Efficient Anatomical Labeling by Statistical Recombination of Partially Label Datasets

Introduction: Numerous clinically relevant conditions (e.g., degeneration, inflammation, vascular pathology, tramatic injury, cancer, etc.) correlate with volumetric/morphometric changes as observed on MRI. Study of these changes necessitates the ability to label (delineate) voxel-wise classifications for structures of interest. The established gold standard for identifying class memberships is manual voxel-by-voxel labeling by a neuro-anatomist expert, which can be exceptionally time and resource intensive. Furthermore, different human experts often have differing interpretations of ambiguous voxels (on the order of 5-10% of a typical brain structure). Therefore, pursuit of manual approaches is typically limited to either (1) validating automated or semi-automated methods or (2) the study of structures for which no automated method exists. Previously, statistical methods have been presented to simultaneously estimate rater reliability and true labels from complete datasets by several different raters [1-3]. These maximum likelihood/maximum a posteriori methods (i.e., STAPLE) increase the accuracy of a single labeling by combining information from multiple, potentially less accurate raters (so long as the set of raters is independent and unbiased). However, the existing methods require that all raters delineate all voxels, which limits applicability in real research studies where different sets of raters may delineate arbitrary subsets of a population of scans due to the rater availability or the duration of the study. Here, we present Simultaneous Truth and Performance Level Estimation with Robust extensions (STAPLER) to enable use of data with: (1) Missing labels: partial labels sets in which raters do not delineate all voxels; (2) Repeated labels: labels sets in which raters may generate repeated (independent) labels for some (or all) voxels; and (3) Catch trials: label sets in which some raters may have known reliabilities (or some voxels have known true labels). STAPLER simultaneously incorporates all labels from all raters to estimate a maximum a posteriori estimate of both rater reliability and labels.