We investigate how our newly introduced paradigm¹, combining advanced modeling approaches with intracellular metabolites diffusion measurements at long diffusion times (t_d), compares qualitatively and quantitatively with histological data. Diffusion compartments for supposedly astrocytic (myo-inositol Ins and choline compounds tCho²) and neuronal metabolites (NAA and glutamate Glu) are reconstructed from diffusion-weighted spectroscopy (DW-MRS) up to ultra-long t_d (2 seconds). Synthetic histological slices are then extracted, allowing direct comparison with real histological slices. Quantitative evaluation of cellular morphology by Sholl analysis yields very similar results for real astrocytes and for synthetic cells reconstructed from Ins and tCho. Interspecies comparison is also performed using our method: we are able to measure major interspecies difference regarding synthetic astrocytes (echoing the most recent knowledge about astrocytes), while dendritic organization (gleaned from NAA and Glu diffusion) appears better conserved throughout species. This work strongly suggests that the time dependence of metabolite apparent diffusion coefficient (ADC), when adequately modeled, allows quantitatively characterizing neuronal and astrocytic morphologies non-invasively.We have reported elsewhere how we measured metabolite ADC in the macaque brain up to very long t_d using a stimulated echo approach^1,3. In a similar fashion, here we measured metabolite ADC as a function of t_d in eight healthy C57/BL6 mice, on a 11.7 T Bruker scanner equipped with a cryoprobe. Spectra were acquired at b=0 and 3000 s/mm² for t_d=52, 352, 502, 652, 1002 and 2002 ms (Fig. 1). Post-processing consisted in scan-to-scan phasing, eddy current correction and subtraction of experimental macromolecule spectrum. Spectra were analyzed with LCModel to estimate the ADC of total N-acetyl aspartate (tNAA=NAA+NAAG), total creatine (tCr), tCho, Glu, Ins. ADC(t_d) was analyzed for each metabolite using our recently introduced simulation-fitting pipeline¹. Briefly, the intracellular diffusivity D_intra and the morphometric statistics (mean and S.D. of segment length L_segment and SD_Lsegment, and mean and S.D. of embranchments along processes N_branch and SD_Nbranch) describing branched cells (like neurons and astrocytes) are iterated until the ADC simulated for particles diffusing in corresponding cells matches experimental ADC. The number of processes leaving the soma, as well as fiber diameters, were set to realistic values for visual representation, but this does not affect any of the quantitative results given below. Synthetic tissues were built by positioning synthetic cells in 3D in a realistic manner. We primarily considered Ins and tCho compartments to tentatively obtain a synthetic tissue mainly representative of astrocytes, comparable with tissues stained for glial fibrillary acidic protein (GFAP) observed by confocal microscopy. 2500 cells were generated according to the morphometric statistics of Ins and tCho compartments, for both mouse and previously acquired macaque data¹. Only cells intersecting a slice of the same thickness as the histological slices (40 μm) were considered for comparison. The 2D synthetic slices and the histological slices of mouse brain were binarized and used to perform Sholl-based analysis using Fiji toolbox integrated in ImageJ.The morphometric statistics resulting from the best fit for all five metabolites are reported in Table 1. Experimental ADC(t_d) and simulation-fitting results, together with some synthetic cells corresponding to the best fit are reported in Fig.2. The increased size and complexity of synthetic astrocytes in primate compared to mouse can be visually assessed. We find that the overall diameter of synthetic astrocytes is 2.52±0.10 times larger in macaques, similar to the 2.55±0.05 ratio recently obtained from real histology (Human versus rodent⁴). Conversely, dendritic organization appears better conserved throughout species, and neuronal size (~500 µm overall diameter) agrees very well with the typical extension of the dendritic tree of many neuronal types. Comparison between real and synthetic astrocytes is reported in Fig.3, while the corresponding Sholl analysis for mouse astrocytes is reported in Fig.4. In our opinion, the fact that Ins and tCho share very similar diffusion compartments, combined with the very good agreement between Sholl analysis on synthetic and real astrocytes (Fig.4C), firmly supports the notion that Ins and tCho are mainly compartmentalized in astrocytes, and provides a strong argument in favor of the validity of our experimental and modeling strategies. Note that in this work we focused on astrocytes because they represent a relatively homogenous population (compared to neurons), and their domains do not overlap (at least in the mouse brain), facilitating Sholl analysis.The “non-invasive histology” paradigm that we have introduced, although in its infancy and demanding for additional validation, already yields striking results, and should stimulate future methodological and theoretical developments. The possibility to extract quantitative information about long-range cellular structure might find applications in preclinical and clinical research.