Fiber enhancement and 3D orientation analysis in label-free two-photon fluorescence microscopy

Abstract 3D fluorescence microscopy can be profitably used in combination with tissue clearing and myelin labeling techniques for investigating the human brain myeloarchitectonics with sub-micrometric resolution, and for providing detailed histological reference data for the validation of diffusion magnetic resonance imaging (dMRI) tractography. Differently from state-of-the-art polarimetry-based neuroimaging modalities such as 3D polarized light imaging and 3D polarized-sensitive optical coherence tomography, the quantitative histological analysis of cortical and white matter fiber architectures from fluorescence microscopy images requires the estimation of fiber orientations via image processing techniques, as Fourier or structure tensor analysis. However, these can be error-prone as they are not able to detect myelinated fibers in surrounding brain tissue. Here we present a novel image processing pipeline built around a 3D Frangi filter able to enhance fiber structures of varying diameters in tiled brain section reconstructions of arbitrary size, and to produce accurate 3D fiber orientation maps in both grey and white matter regions, while preserving the native resolution of the employed microscopy system. The developed software tool also features the estimation of fiber orientation distribution functions which, in view of their widespread use in the neuroimaging community, may favor a multimodal comparison with brain connectivity data obtained from the aforementioned label-free optical modalities, and the validation of modern dMRI-based tractography. The proposed image processing pipeline was applied to brain samples treated according to a novel label-free preparation technique capable of enhancing the autofluorescence of myelinated fiber axons, thus demonstrating the feasibility of single fiber orientation analyses in fluorescence microscopy without the need for exogenous staining.