3D reconstruction of brain section images for creating axonal projection maps in marmosets

The brain of the common marmoset (Callithrix jacchus) is becoming a popular non-human primate model in neuroscience research. Because its brain fiber connectivity is still poorly understood, it is necessary to collect and present connection and trajectory data using tracers to establish a marmoset brain connectivity database.

To visualize projections and trajectories of axons, brain section images were reconstructed in 3D by registering them to the corresponding block-face brain images taken during brain sectioning. During preprocessing, autofluorescence of the tissue was reduced by applying independent component analysis to a set of fluorescent images taken using different filters.

The method was applied to a marmoset dataset after a tracer had been injected into an auditory belt area to fluorescently label axonal projections. Cortical and subcortical connections were clearly reconstructed in 3D. The registration error was estimated to be smaller than 200 μm. Evaluation tests on ICA-based autofluorescence reduction showed a significant improvement in signal and background separation.

Regarding the 3D reconstruction error, the present study shows an accuracy comparable to previous studies using MRI and block-face images. Compared to serial section two-photon tomography, an advantage of the proposed method is that it can be combined with standard histological techniques. The images of differently processed brain sections can be integrated into the original ex vivo brain shape.

The proposed method allows creating 3D axonal projection maps overlaid with brain area annotations based on the histological staining results of the same animal.