Generative Modelling of Cortical Receptor Distributions from Cytoarchitectonic Images in the Macaque Brain

Area of Science:

• Neuroscience
• Computational Biology
• Medical Imaging

Background:

• Neurotransmitter receptor densities are crucial for understanding brain region architecture.
• Current methods like autoradiography are time-consuming and costly, limiting whole-brain mapping.
• High-resolution 3D brain maps from light microscopy capture cell density but lack receptor data.

Purpose of the Study:

• To investigate the feasibility of predicting neurotransmitter receptor density distributions from cell-body stainings.
• To address data gaps in creating comprehensive, multi-modal whole-brain atlases.
• To develop a computational method for cross-modality image translation in neuroscience.

Main Methods:

• Utilized conditional Generative Adversarial Networks (cGANs) for image-to-image translation.
• Trained cGANs on aligned sections of macaque monkey cortex (V1 and M1) showing cell-body and receptor distributions.
• Predicted densities of M2 acetylcholine and kainate glutamate receptors based on cell-body stained images.

Main Results:

• Demonstrated the capability of cGANs to predict realistic neurotransmitter receptor density distributions.
• Validated predictions qualitatively and quantitatively, showing preservation of cortical features like laminar thickness and curvature.
• Established a successful mapping between cell-body staining and receptor distribution modalities.

Conclusions:

• Cross-modality image translation using cGANs is feasible for predicting receptor densities.
• This approach can help bridge data gaps in constructing detailed whole-brain atlases.
• The method offers a potential solution for more efficient neuroreceptor mapping.