Three-Dimensional Reconstruction and Analysis of Laminar Microstructure In Young Autistic Males

Background: A consistent pattern of aberrant neurobiology has yet to be uncovered in autism. Previous studies using postmortem tissue have yielded potential theories for the etiological origin and symptom diversity of the disorder but no conclusive evidence for a specific cytological defect. These studies are severely limited by the small number of young cases available and the intrinsic structural heterogeneity of cerebral cortex.  With such limits in place, optimal analysis of the available tissue is critical. Advances in postmortem tissue labeling using in situ hybridization (ISH) now enable laminar-based analyses using multiple genes. This information, combined with advanced image processing algorithms, can generate rich descriptions of laminar organization and three-dimensional microstructure.

Objectives: To characterize laminar organization of young male autistic tissue using novel algorithms applied to ISH data. Further, we aim to visualize the laminar microstructure using volume reconstruction of registered ISH stacks.

Methods: Prefrontal cortical tissue was obtained from the NIHCD Brain and Tissue Bank for Developmental Disorders for 8 young typically-developing and autistic cases. The tissue was sectioned into ten sub-regions and sliced into a set of 30 slices, each 25µm thick.  ISH was performed using 14 laminar-specific genes with nissl-stained slides for spatial reference. Visible light and expression images were captured for each slice at high resolution.

To calculate laminar distribution, ISH images were first aligned to their reference nissl. Image noise was removed using a HSV filter, a fixed-intensity threshold, and gap-fill step to produce a gap-free binary image. Bezier curves were fit to pial and gray-white contours to create the sampling domain. Intensity profiles were sampled following the curves to produce an aligned intensity array from which mean intensity profiles were obtained.

To generate the volume reconstructions of laminar architecture, filtered ISH images from the intensity calculation were processed using a particle filter to generate a point distribution. A two-dimensional Gaussian convolution was applied to generate a weighted density figure for each laminar marker. The sections were aligned and interleaved to create a single image volume. Volumes for multiple ISH markers were then rendered in a single environment to produce a complete three-dimensional reconstruction of the laminar microstructure.

Results: Using the intensity-based analysis, we were able to identify regions of aberrant organization in multiple layer-specific markers in several autistic cases. Further, we found spared regions of autistic cortex with normal laminar distributions. Finally, from volume reconstruction of regions identified to be aberrant, we were able to visualize a potential structural defect in the laminar microstructure in some autism cases that was not apparent in two-dimensional slice analyses.  

Conclusions: Here we present a sophisticated method to characterize laminar disorganization using intensity and density-based algorithms. With this approach we show differences in laminar-specific organization in several autistic cases compared to controls. Further, from the volume reconstructions we have identified a potential structural feature in some cases that can only been captured using a three-dimensional analysis. We will present this finding and discuss potential implications with respect to normal development and cortical migration defects.