Researchers at INM‑1 have developed a new method to visualize the fine structure of the brain without elaborate laboratory procedures. With the help of artificial intelligence (AI), they can now virtually show how nerve cells are distributed and how they connect with nerve fibers. The study was published in Imaging Neuroscience.
To fully capture how nerve cells (cytoarchitecture) and nerve fibers (myeloarchitecture) are arranged in brain tissue, different imaging techniques must be combined. Traditionally, stained tissue sections are analyzed under a microscope to study cells, while nerve fibers can be made visible using the method of 3D Polarized Light Imaging (3D‑PLI). This technique provides three‑dimensional orientations of nerve fibers with microscopic resolution—without the need for staining.
Excerpts from the primary motor cortex (left column) and the cornus ammonis of the hippocampus (right column). The first row shows fiber tracts in both brain regions, determined using 3D-PLI. Second row: cell bodies stained with cresyl violet. Last row: distributions of cell bodies predicted from the 3D-PLI data.
Copyright:— Alexander Oberstrass et al., Imaging Neuroscience 2026; 4 IMAG.a.1079. https://doi.org/10.1162/IMAG.a.1079
However, to directly compare cells and fibers in the same tissue sections, the tissue has so far been stained afterward. This leads to unavoidable distortions and is technically very demanding. As a result, only a few samples can be analyzed in this way.
For the study, the scientists therefore used artificial intelligence methods. Deep learning models learn what a certain staining—in this case using the dye cresyl violet—would look like and generate a “virtual staining” directly from the 3D‑PLI images. Special mathematical algorithms correct small misalignments between image data so that the cellular structures appear precisely in their proper locations.
The result: the AI can realistically extract the distribution and shape of nerve cells in the brain without the samples actually having to be stained.
The study lays the foundation for future work aiming to further improve 3D PLI analysis. A particular focus is the development of specialized cell detection methods that allow individual cell bodies to be identified directly from 3D PLI data.
At the same time, the results show how important it is to obtain additional training data in order to further refine virtual cresyl violet staining and extend it to more sections, brains, and species. Although the current model is designed to replicate classic cresyl violet staining, the approach can also be applied to other types of staining, provided that the model is retrained for this purpose. Future research should therefore focus in particular on examining how well the method can be transferred to other data sets, staining protocols, and brains.
Original publication:
Alexander Oberstrass, Esteban Vaca, Eric Upschulte, Meiqi Niu, Nicola Palomero-Gallagher, David Graessel, Christian Schiffer, Markus Axer, Katrin Amunts, Timo Dickscheid; From fibers to cells: Fourier-based registration enables virtual Cresyl violet staining from 3D polarized light imaging. Imaging Neuroscience 2026; 4 IMAG.a.1079. doi: https://doi.org/10.1162/IMAG.a.1079