Introduction
Scattered Light Imaging (SLI) is a novel neuroimaging technique that allows to explore the substructure of nerve fibers, especially in regions with crossing nerve fibers, in whole brain sections with micrometer resolution (Menzel et al. (2020)). By illuminating histological brain sections from different angles and measuring the transmitted light under normal incidence, characteristic light intensity profiles (SLI profiles) can be obtained which provide crucial information such as the directions of crossing nerve fibers in each measured image pixel.
This repository contains the Scattered Light Imaging ToolboX (SLIX) – an open-source Python package that allows a fully automated evaluation of SLI measurements and the generation of different parameter maps. The purpose of SLIX is twofold: First, it allows to transform the raw data of SLI measurements (SLI image stack) to human-readable parameter maps that can be used for further analysis and interpreted by researchers. To this end, SLIX also contains functions to visualize the resulting parameter maps, e.g. as colored vector maps. Second, the results of SLIX can be processed further for use in tractography algorithms. For example, the resulting fiber direction maps can be stored as angles or as unit vectors, which can be used as input for streamline tractography algorithms (Nolden et al. (2019)).
The figure belows shows the different steps, from the SLI measurement to the generation of parameter maps:
The sample is illuminated from different angles, with constant polar angle and different equidistant azimuthal angles
(starting on top and rotating clock-wise), see figure (a). A camera behind the sample records an image of the transmitted light under normal incidence for each direction of illumination, yielding a series of images (b).
The SLI image stack is used as input for SLIX (.nii or .tiff are accepted). The software assumes that the measurement has been performed with equidistant angles over a full range of 360°. The number of images defines the illumination angles (e.g. when using 24 images as input, the software assumes that the images were recorded with = 0°,15°,...,345°).
Each pixel in the SLI image stack contains a light intensity profile (SLI profile ), which is characteristic for the brain tissue structure at this point (see Menzel et al. (2020) for more details). With SLIX, it is possible to automatically extract and evaluate the SLI profiles for all image pixels.
The peak positions are computed with scipy.signal.find_peaks, taking the 360° periodicity of the signal into account. To account for inaccuracies introduced by the discretization of the SLI profile, the determined peak positions are corrected by calculating the geometric center of the peak tips with a height corresponding to 6% of the total signal amplitude. The value of 6% turned out to be the best choice to obtain reliable fiber orientations (see Menzel et al. (2020), Appx. B), but can be changed by the user. Figure (c) shows an SLI profile with 15° discretization and the corrected peak positions as vertical lines.
To avoid that peaks caused by noise or details in the fiber structure impair the computed fiber direction angles, only prominent peaks are used for further evaluation. The peak prominence (scipy.signal.peak_prominences) indicates how strongly a peak stands out from the background signal and is defined by the vertical distance between the top of the peak and the higher of the two neighboring minima (see figure (c), in red). If not defined otherwise by the user, peaks with a prominence above 8% of the total signal amplitude (max-min) are considered as prominent peaks. The value of 8% turned out to be the optimal choice for the generation of reliable fiber orientations (best compromise between correctly and wrongly detected peaks for regions with known fiber orientations, see Menzel et al. (2020), Appx. A).
The peak width (see figure (c), in dark blue) is determined as the full width of the peak at a height corresponding to the peak height minus half of the peak prominence.
The in-plane fiber direction angles are computed from the (corrected) mid positions of prominent peak pairs with a pair-wise distance of (180 +/- 35)°, see figure (c) in green/magenta. The range of possible distances has been limited to ensure that the out-of-plane angles of the nerve fibers are not too large (the direction angles of inclined crossing fibers cannot be reliably determined).
By evaluating the SLI profiles of each image pixel, SLIX generates different parameter maps, which provide various information about the investigated brain tissue:
- The average map shows the overall scattering of the tissue; maximum and minimum can be used to get an idea of the signal amplitude and the signal-to-noise ratio.
- The number of non-prominent peaks and the number of prominent peaks indicate the clarity of the signal (regions with indefinite scattering signals, such as background or regions with a small number of nerve fibers, show a higher number of non-prominent peaks); the average peak prominence indicates the reliability of the peak positions.
- The average peak width and the peak distance correlate with the out-of-plane angle of the nerve fibers (in-plane fibers show two narrow peaks with a large distance of about 180°).
- The direction maps show the in-plane direction angles of the nerve fibers for up to three different crossing fibers. The fiber directions can be represented by a colored vector map, as described in the tutorial below.
With SLIXLineplotParameterGenerator, it is possible to evaluate individual SLI profiles and compute characteristics such as the number of peaks, their positions, and in-plane fiber direction angles. For a given SLI image stack, SLIXParameterGenerator computes the desired parameter maps for all image pixels.