In vivo neuropil density from anatomical MRI and machine learning.

Brain energy budgets specify metabolic costs emerging from underlying mechanisms of cellular and synaptic activities. While current bottom-up energy budgets use prototypical values of cellular density and synaptic density, predicting metabolism from a person's individualized neuropil density would be ideal. We hypothesize that in vivo neuropil density can be derived from magnetic resonance imaging (MRI) data, consisting of longitudinal relaxation (T1) MRI for gray/white matter distinction and diffusion MRI for tissue cellularity (apparent diffusion coefficient, ADC) and axon directionality (fractional anisotropy, FA). We present a machine learning algorithm that predicts neuropil density from in vivo MRI scans, where ex vivo Merker staining and in vivo synaptic vesicle glycoprotein 2A Positron Emission Tomography (SV2A-PET) images were reference standards for cellular and synaptic density, respectively. We used Gaussian-smoothed T1/ADC/FA data from 10 healthy subjects to train an artificial neural network, subsequently used to predict cellular and synaptic density for 54 test subjects. While excellent histogram overlaps were observed both for synaptic density (0.93) and cellular density (0.85) maps across all subjects, the lower spatial correlations both for synaptic density (0.89) and cellular density (0.58) maps are suggestive of individualized predictions. This proof-of-concept artificial neural network may pave the way for individualized energy atlas prediction, enabling microscopic interpretations of functional neuroimaging data.

A 3D atlas of functional human brain energetic connectome based on neuropil distribution.

The human brain is energetically expensive, yet the key factors governing its heterogeneous energy distributions across cortical regions to support its diversity of functions remain unexplored. Here, we built up a 3D digital cortical energy atlas based on the energetic costs of all neuropil activities into a high-resolution stereological map of the human cortex with cellular and synaptic densities derived, respectively, from ex vivo histological staining and in vivo PET imaging. The atlas was validated with PET-measured glucose oxidation at the voxel level. A 3D cortical activity map was calculated to predict the heterogeneous activity rates across all cortical regions, which revealed that resting brain is indeed active with heterogeneous neuronal activity rates averaging around 1.2 Hz, comprising around 70% of the glucose oxidation of the cortex. Additionally, synaptic density dominates spatial patterns of energetics, suggesting that the cortical energetics rely heavily on the distribution of synaptic connections. Recent evidence from functional imaging studies suggests that some cortical areas act as hubs (i.e., interconnecting distinct and functionally active regions). An inverse allometric relationship was observed between hub metabolic rates versus hub volumes. Hubs with smaller volumes have higher synapse density, metabolic rate, and activity rates compared to nonhubs. The open-source BrainEnergyAtlas provides a granular framework for exploring revealing design principles in energy-constrained human cortical circuits across multiple spatial scales.

Spatiotemporal features of neurovascular (un)coupling with stimulus-induced activity and hypercapnia challenge in cerebral cortex and olfactory bulb.

Carbon dioxide (CO2) is traditionally considered as metabolic waste, yet its regulation is critical for brain function. It is well accepted that hypercapnia initiates vasodilation, but its effect on neuronal activity is less clear. Distinguishing how stimulus- and CO2-induced vasodilatory responses are (dis)associated with neuronal activity has profound clinical and experimental relevance. We used an optical method in mice to simultaneously image fluorescent calcium (Ca2+) transients from neurons and reflectometric hemodynamic signals during brief sensory stimuli (i.e., hindpaw, odor) and CO2 exposure (i.e., 5%). Stimuli-induced neuronal and hemodynamic responses swiftly increased within locally activated regions exhibiting robust neurovascular coupling. However, hypercapnia produced slower global vasodilation which was temporally uncoupled to neuronal deactivation. With trends consistent across cerebral cortex and olfactory bulb as well as data from GCaMP6f/jRGECO1a mice (i.e., green/red Ca2+ fluorescence), these results unequivocally reveal that stimuli and CO2 generate comparable vasodilatory responses but contrasting neuronal responses. In summary, observations of stimuli-induced regional neurovascular coupling and CO2-induced global neurovascular uncoupling call for careful appraisal when using CO2 in gas mixtures to affect vascular tone and/or neuronal excitability, because CO2 is both a potent vasomodulator and a neuromodulator.

Fluorescently-tagged magnetic protein nanoparticles for high-resolution optical and ultra-high field magnetic resonance dual-modal cerebral angiography.

Extremely small paramagnetic iron oxide nanoparticles (FeMNPs) (<5 nm) can enhance positive magnetic resonance imaging (MRI) contrast by shortening the longitudinal relaxation time of water (T1), but these nanoparticles experience rapid renal clearance. Here, magnetic protein nanoparticles (MPNPs) are synthesized from protein-conjugated citric acid coated FeMNPs (c-FeMNPs) without loss of the T1 MRI properties and tagged with fluorescent dye (f-MPNPs) for optical cerebrovascular imaging. The c-FeMNPs shows average size 3.8 ± 0.7 nm with T1 relaxivity (r1) of 1.86 mM-1 s-1 and transverse/longitudinal relaxivity ratio (r2/r1) of 2.53 at 11.7 T. The f-MPNPs show a higher r1 value of 2.18 mM-1 s-1 and r2/r1 ratio of 2.88 at 11.7 T, which generates excellent positive MRI contrast. In vivo cerebral angiography with f-MPNPs enables detailed microvascular contrast enhancement for differentiation of major blood vessels of murine brain, which corresponds well with whole brain three-dimensional time-of-flight MRI angiograms (17 min imaging time with 60 ms repetition time and 40 µm isotropic voxels). The real-time fluorescence angiography enables unambiguous detection of brain capillaries with diameter < 40 µm. Biodistribution examination revealed that f-MPNPs were safely cleared by the organs like the liver, spleen, and kidneys within a day after injection. Blood biochemical assays demonstrated no risk of iron overload in both rats and mice. With hybrid neuroimaging technologies (e.g., MRI-optical) on the rise, f-MPNPs built on this platform can generate exciting neuroscience applications.

Imaging Hallmarks of the Tumor Microenvironment in Glioblastoma Progression.

Glioblastoma progression involves multifaceted changes in vascularity, cellularity, and metabolism. Capturing such complexities of the tumor niche, from the tumor core to the periphery, by magnetic resonance imaging (MRI) and spectroscopic imaging (MRSI) methods has translational impact. In human-derived glioblastoma models (U87, U251) we made simultaneous and longitudinal measurements of tumor perfusion (Fp), permeability (Ktrans), and volume fractions of extracellular (ve) and blood (vp) spaces from dynamic contrast enhanced (DCE) MRI, cellularity from apparent diffusion coefficient (ADC) MRI, and extracellular pH (pHe) from an MRSI method called Biosensor Imaging of Redundant Deviation in Shifts (BIRDS). Spatiotemporal patterns of these parameters during tumorigenesis were unique for each tumor. While U87 tumors grew faster, Fp, Ktrans, and vp increased with tumor growth in both tumors but these trends were more pronounced for U251 tumors. Perfused regions between tumor periphery and core with U87 tumors exhibited higher Fp, but Ktrans of U251 tumors remained lowest at the tumor margin, suggesting primitive vascularization. Tumor growth was uncorrelated with ve, ADC, and pHe. U87 tumors showed correlated regions of reduced ve and lower ADC (higher cellularity), suggesting ongoing proliferation. U251 tumors revealed that the tumor core had higher ve and elevated ADC (lower cellularity), suggesting necrosis development. The entire tumor was uniformly acidic (pHe 6.1-6.8) early and throughout progression, but U251 tumors were more acidic, suggesting lower aerobic glycolysis in U87 tumors. Characterizing these cancer hallmarks with DCE-MRI, ADC-MRI, and BIRDS-MRSI will be useful for exploring tumorigenesis as well as timely therapies targeted to specific vascular and metabolic aspects of the tumor microenvironment.

Doppler OCT clutter rejection using variance minimization and offset extrapolation.

Doppler optical coherence tomography (OCT) is widely used for high-resolution mapping of flow velocities and is based on analysis of temporal changes in the phase of an OCT signal (i.e., how fast the OCT signal rotates in the complex plane). Determination of the rate of phase change or rotation speed critically depends on the center of rotation. Here, we demonstrate the bias in high-pass filtering, the current widely used method to determine the center of rotation, and propose two advanced methods for Doppler OCT clutter rejection. The bias in the high-pass filtering method becomes increasingly significant with lower velocities or larger signal noise. Two novel methods based on variance minimization and offset extrapolation can potentially reduce this bias and thereby improve the accuracy of Doppler OCT measurements of flow velocities, even for low-velocity and/or high-noise signals. The two novel methods and the current standard method (high-pass filtering) have been tested in combination with several currently used velocity measurement algorithms: Kasai, autocorrelation function fitting, and maximum likelihood estimation. The newly proposed methods are shown to improve the accuracy in both the center of rotation and resultant velocity by up to 60 percentage points and reduce the flow conservation error by 30% when applied to in vivo cerebral blood flow imaging of the rodent brain cortex.