Ileana Ozana Jelescu

Conventional connectome edge weights, such as number of streamlines (NOS) or diffusion tensor imaging (DTI) metrics, lack specificity to microstructural details which may hold relevance for macroscale brain organisation. Since biophysical diffusion modelling offers greater specificity to microstructure, we investigated whether parameters from the Standard Model of white matter diffusion provide informative alternatives for connectome weights – namely the intra-axonal signal fraction (f) and perpendicular extra-axonal diffusivity (De⟂). Using diffusion MRI data from healthy adults, we constructed structural networks at four parcellation scales, weighted by f, De⟂, NOS, fractional anisotropy (FA) and radial diffusivity (RD). While all weights reproduced expected small-world properties and hub structure, only De⟂ and normalised NOS captured non-random properties of local organisation at all spatial scales. We then correlated each weighted connectome with resting-state fMRI functional connectivity and intracranial measurements of conduction velocity. At both the whole-brain and regional level, NOS gave strong coupling with fMRI functional connectivity. Connectomes weighted by De⟂, FA and RD gave weak but significant coupling with fMRI functional connectivity and conduction velocity. Notably, De⟂ and RD gave highest consistency in regional structure-function coupling across spatial scales. Thus, connectome weights derived from De⟂ capture meaningful aspects of brain network organisation with functional relevance.

Children with type C hepatic encephalopathy (HE) suffer from irreversible neurological damage even after liver transplantation. The metabolic and structural mechanisms underlying the vulnerability of the developing brain are still poorly understood. We used here the unique combination of diffusion magnetic resonance spectroscopy (dMRS) and diffusion magnetic resonance imaging (dMRI) to probe cell-type specific microstructural changes in vivo, non-invasively and longitudinally, in the cerebellum of a rat model of early onset type C HE. Glial and neuronal metabolites, present intracellularly only, exhibited increased diffusivities and decreased kurtosis in HE rats compared to controls, indicating reduced intracellular microstructure complexity. Water, present both intra and extracellularly, exhibited higher diffusivity both in white and grey matter and, through biophysical modelling, probed increased intra-fiber water diffusivity, decreased fiber fraction, increased soma radius and tentatively, decreased water intra-to-extracellular exchange rate in HE rats versus controls. These results suggest that early HE causes irreversible reduction of astrocytic and neuronal network complexity and arborization in the cerebellum, as confirmed by immunohistochemistry of the granular layer, Bergmann glia and Purkinje neurons in the same animals. These observations reorient the current and restrictive hypothesis of astrocytes swelling towards a broader one of complex multi-cellular microstructure alterations in type C HE.

The non-invasive estimation of intra- and extracellular microstructural parameters using biophysical models has been a major focus in brain microstructure imaging with MRI. The Standard Model (SM) of diffusion in white matter (WM) provides a unifying framework for various modelling approaches, representing axons as impermeable sticks embedded within a locally anisotropic extra-axonal space. However, the SM relies on simplifying assumptions that may not hold in realistic WM tissue, as they do not take into account axonal undulations, beading, the presence of glial cells, or membrane permeability. In this work, we investigate how SM parameter estimates, derived using the Standard Model Imaging (SMI) framework, behave when the model is applied to realistic numerical WM substrates generated by the CATERPillar tool. Specifically, we vary (i) axonal morphological features such as beading and undulations, (ii) axonal packing density, (iii) orientation dispersion, (iv) membrane permeability of axons and astrocytes separately, (v) myelin volume fraction, and (vi) diffusion time. In each part of the analysis, different noise levels are introduced. Our findings enable users to identify potential biases introduced by varying conditions and to adjust their interpretations accordingly.

Biophysical diffusion MRI models such as Neurite Exchange Imaging (NEXI) are essential for probing gray matter microstructure, estimating compartment diffusivities (Di, De), neurite fraction (f), and exchange time (tex). However, NEXI multi-shell, multi-diffusion-time requirements lead to prohibitively long acquisitions. Leveraging the Connectome 2.0 ultra-high gradient scanner, we developed a time-efficient protocol using an Explainable AI (XAI) framework. Combining XGBoost, SHAP, and Recursive Feature Elimination trained on synthetic signals, XAI identified an optimal 8-feature subset, halving scan time from 27 to 14 minutes. Validated in vivo in 7 healthy participants, the XAI protocol was benchmarked against the full 15-feature acquisition, a Cramér-Rao Lower Bound (CRLB) theoretical optimum, and 2 expert-designed heuristics (“Mid-Range” and “Corner”). The XAI protocol robustly reproduced parameter estimates and maintained full protocol test–retest reproducibility. Remarkably, the data-driven XAI selection converged to the theoretical CRLB optimum. This validates XAI mathematical optimality while highlighting its primary advantage: achieving gold-standard optimization without requiring complex analytical Jacobians, making it effortlessly adaptable to numerically defined models or complex noise distributions where CRLB becomes intractable. Furthermore, XAI demonstrated superior in vivo robustness compared with heuristics: “Mid-Range” sampling yielded biased tex estimates due to insufficient temporal diversity, while “Corner” sampling yielded unstable Di estimates (fivefold higher coefficient of variation) due to noise sensitivity. Ultimately, this robust 14-minute NEXI protocol accelerates exchange-sensitive microstructural mapping, establishing a generalizable, model-agnostic optimization framework adaptable to future ultra-high gradient systems (e.g., Siemens Cima.X or GE MAGNUS) and existing clinical scanners.

NEXI is a gray matter (GM) microstructural model designed to probe brain tissue microstructure in vivo using diffusion MRI. NEXI describes GM as two exchanging Gaussian compartments—neurites, modeled as randomly oriented, infinitely long sticks, and the extracellular space—allowing the estimation of biophysically interpretable parameters related to neurite microstructure and intercompartmental exchange. While modeling cell processes as sticks and each compartment as Gaussian are common assumptions for brain biophysical models of diffusion, neurite structural irregularities and the presence of somas, particularly in GM, may violate them and bias NEXI parameter estimates. Furthermore, the barrier-limited exchange assumed in the Kärger model that underlies NEXI may also be violated in realistic conditions. Therefore, in this work, we evaluate NEXI’s accuracy in numerical substrates that contain GM features such as neurite beading, undulation, orientation dispersion, and the presence of somas across a range of membrane permeabilities. Diffusion signals were generated with Monte Carlo simulations of water diffusion and subsequently fitted with NEXI. Overall, NEXI captured trends of exchange time across permeability levels and showed good ability to disentangle exchange effects from other microstructural features. The largest bias in tex was introduced by somas, and in Di by permeability, somas and beading. Otherwise, the structural irregularities evaluated in this study had a relatively modest impact on NEXI estimates. Most bias in NEXI estimates comes from finite SNR. Future work should aim to further improve substrate realism to better represent the complexity of GM morphology.

Early human cortical development involves rapid cytoarchitectural changes that are difficult to study non-invasively. Diffusion MRI can probe developing microstructure and membrane water exchange, but model assumptions are difficult to validate in human tissue. Cortical organoids provide a tractable platform for relating diffusion measurements to tissue architecture. We scanned six fixed cortical organoids at 9.4 T, three grown under directed and three under undirected differentiation protocols, at three diffusion times and b-values up to 7000 s/mm ² . Center-to-periphery profiles of diffusion kurtosis imaging (DKI) metrics were reproducible across directed organoids but more variable across undirected organoids, consistent with nuclear staining. Unexpectedly, DKI-derived apparent mean diffusivity (MD) increased with diffusion time in more than 80% of tissue voxels in both batches, opposite to previous organoid measurements. Across 12,348 analytically simulated two-compartment substrates and Monte Carlo diffusion simulations in sphere packings, this trend emerged in permeable substrates with a marked contrast between extracellular and apparent intracellular diffusivities. Detailed Monte Carlo trajectory analysis of a representative substrate showed that fitted MD increased despite decreases in displacement-derived and second-cumulant diffusivities, revealing an exchange-sensitive finite-b effect involving higher-order signal contributions. These findings support using cortical organoids to probe early microstructure, test diffusion models, and investigate potential emerging diffusion biomarkers.

Brain cell morphology plays a crucial role in function and pathology. Biophysical models of diffusion MRI (dMRI) quantify cell morphology in vivo , enabling the design of novel biomarkers. These models represent cells by simplified geometries, such as spheres and randomly oriented cylinders, for which analytical signal expressions exist. However, dMRI signals are sensitive to morphological features, such as branching, tapering, undulation, beading, and protrusions, unaccounted for in most models, as they often render analytical solutions impossible. Simulations of the dMRI signal in synthetically generated cells offer a powerful tool to explore how microstructural morphology impacts the dMRI signal. Nevertheless, no tool to generate digital replicas of brain cells is available openly. To address this gap, we introduce the OCTOPUS toolbox, which generates cells featuring all geometrical features described above. OCTOPUS, provided via the Python interface OCTOpool, enables accessible, efficient creation of cells with complex geometries. We recreated histologically reconstructed neuronal and glial cells, including pyramidal, GABAergic, and glutamatergic neurons, and astroand microglia. To illustrate the plausibility of OCTOPUS-generated cells, we reproduced established properties of dMRI signals from brain tissue, such as the signatures of short-range disorder, branching and protrusions, and a high-b-value power law. By comparing the geometries and dMRI signals of generated and original cells, we found different growth strategies adequate for more isotropic and more anisotropic cells. We anticipate that realistic cell substrates created by OCTOPUS will help validate biophysical models, design dMRI sequences sensitive to fine-grained cell morphology beyond analytical models, and generate realistic numerical substrates of brain tissue. Highlights: OCTOPUS is an open-source tool generating realistic gray matter cells in seconds. Cells have tunable features: branching, tapering, undulation, beading, and spines. A Python wrapper facilitates fast, flexible cell replication with OCTOPUS. dMRI decays in cells showed signatures of disorder, branching, and spines. OCTOPUS is a powerful tool to explore microstructure imaging techniques.

The coupling between brain excitatory activity and positive blood oxygen level‐dependent (BOLD) responses is well‐established. Although often associated with inhibition, negative BOLD remains partially understood. Moving away from neurovascular coupling, apparent diffusion coefficient (ADC)‐fMRI provides a more direct measure of excitatory activity, possibly mediated by transient cellular deformations. Diffusion‐weighted fMRI (dfMRI), from which ADC‐fMRI derives, combines vascular and microstructural contributions. While decreases in ADC align with positive BOLD, the possible translation of negative BOLD into positive ADC and the ability of ADC‐fMRI to capture inhibitory activity remain unexplored in humans. In this study, we used transcranial magnetic stimulation (TMS)‐fMRI on the right primary motor cortex (M1) to selectively induce contralateral negative BOLD responses, whilst acquiring interleaved fMRI. TMS was applied at 5 Hz, 90% resting motor threshold, for both BOLD‐fMRI and ADC‐fMRI contrasts, in n = 12 and n = 11 healthy participants, respectively. We replicated previously reported negative BOLD clusters in the contralateral M1 and primary somatosensory cortex (S1). This was accompanied by a negative dfMRI response, but no ADC‐fMRI response, indicating minimal microstructural fluctuations. In ipsilateral M1/S1, no BOLD response was detected while dfMRI revealed a positive cluster, suggesting different sensitivity to the excitatory/inhibitory balance. Overall, combining the findings from BOLD‐fMRI and ADC‐fMRI provides new insights into the vascular and neuronal responses underlying subthreshold TMS and negative BOLD.

Purpose: Biophysical models of diffusion tailored to characterize gray matter (GM) microstructure are gaining traction in the neuroimaging community. NEXI, SMEX, SANDI, and SANDIX represent recent efforts to account for different microstructural features, such as soma contributions and inter‐compartment exchange, in the diffusion MRI (dMRI) signal. The purpose of this work is to provide a comparative evaluation of these four GM diffusion models. Methods: A comparative analysis of NEXI, SMEX, SANDI, and SANDIX was performed using a single, publicly available in vivo human dataset, the Connectome Diffusion Microstructure Dataset (CDMD), acquired with two diffusion times. Cortical microstructure metrics were estimated in 26 healthy subjects using the open‐source Gray Matter Swiss Knife toolbox, and goodness of fit, anatomical patterns, and consistency with previous studies were evaluated. Results: CDMD data yielded GM parameter estimates consistent with values reported in previous studies across all four models. NEXI and SMEX produced similar cortical anatomical patterns, with consistent regional distributions across diffusion times. Goodness of fit varied across models, with NEXI showing the best fit, followed by SMEX, and finally SANDIX. SANDI parameter estimates showed strong dependence on diffusion‐time selection and fitting algorithm. Conclusion: This retrospective cross‐model analysis establishes the feasibility of estimating exchange models from only two diffusion times and highlights trade‐offs in biological specificity, model complexity, and fitting robustness, which are critical considerations when selecting a model for future clinical and research applications.