Merged magnetic resonance and light sheet microscopy of the whole mouse brain.

We have developed workflows to align 3D magnetic resonance histology (MRH) of the mouse brain with light sheet microscopy (LSM) and 3D delineations of the same specimen. We start with MRH of the brain in the skull with gradient echo and diffusion tensor imaging (DTI) at 15 µm isotropic resolution which is ~ 1,000 times higher than that of most preclinical MRI. Connectomes are generated with superresolution tract density images of ~5 µm. Brains are cleared, stained for selected proteins, and imaged by LSM at 1.8 µm/pixel. LSM data are registered into the reference MRH space with labels derived from the ABA common coordinate framework. The result is a high-dimensional integrated volume with registration (HiDiver) with alignment precision better than 50 µm. Throughput is sufficiently high that HiDiver is being used in quantitative studies of the impact of gene variants and aging on mouse brain cytoarchitecture and connectomics.

Strategies for Implementing Machine Learning Algorithms in the Clinical Practice of Radiology.

Despite recent advancements in machine learning (ML) applications in health care, there have been few benefits and improvements to clinical medicine in the hospital setting. To facilitate clinical adaptation of methods in ML, this review proposes a standardized framework for the step-by-step implementation of artificial intelligence into the clinical practice of radiology that focuses on three key components: problem identification, stakeholder alignment, and pipeline integration. A review of the recent literature and empirical evidence in radiologic imaging applications justifies this approach and offers a discussion on structuring implementation efforts to help other hospital practices leverage ML to improve patient care. Clinical trial registration no. 04242667 © RSNA, 2024 Supplemental material is available for this article.

Spatiotemporal profile of atrophy in the first year following moderate-severe traumatic brain injury.

Traumatic brain injury (TBI) triggers progressive neurodegeneration resulting in brain atrophy that continues months-to-years following injury. However, a comprehensive characterization of the spatial and temporal evolution of TBI-related brain atrophy remains incomplete. Utilizing a sensitive and unbiased morphometry analysis pipeline optimized for detecting longitudinal changes, we analyzed a sample consisting of 37 individuals with moderate-severe TBI who had primarily high-velocity and high-impact injury mechanisms. They were scanned up to three times during the first year after injury (3 months, 6 months, and 12 months post-injury) and compared with 33 demographically matched controls who were scanned once. Individuals with TBI already showed cortical thinning in frontal and temporal regions and reduced volume in the bilateral thalami at 3 months post-injury. Longitudinally, only a subset of cortical regions in the parietal and occipital lobes showed continued atrophy from 3 to 12 months post-injury. Additionally, cortical white matter volume and nearly all deep gray matter structures exhibited progressive atrophy over this period. Finally, we found that disproportionate atrophy of cortex along sulci relative to gyri, an emerging morphometric marker of chronic TBI, was present as early as 3 month post-injury. In parallel, neurocognitive functioning largely recovered during this period despite this pervasive atrophy. Our findings demonstrate msTBI results in characteristic progressive neurodegeneration patterns that are divergent across regions and scale with the severity of injury. Future clinical research using atrophy during the first year of TBI as a biomarker of neurodegeneration should consider the spatiotemporal profile of atrophy described in this study.

Aberrant impulse control circuitry in obesity.

The ventromedial prefrontal cortex (vmPFC) to nucleus accumbens (NAc) circuit has been implicated in impulsive reward-seeking. This disinhibition has been implicated in obesity and often manifests as binge eating, which is associated with worse treatment outcomes and comorbidities. It remains unclear whether the vmPFC-NAc circuit is perturbed in impulsive eaters with obesity. Initially, we analyzed publicly available, high-resolution, normative imaging data to localize where vmPFC structural connections converged within the NAc. These structural connections were found to converge ventromedially in the presumed NAc shell subregion. We then analyzed multimodal clinical and imaging data to test the a priori hypothesis that the vmPFC-NAc shell circuit is linked to obesity in a sample of female participants that regularly engaged in impulsive eating (i.e., binge eating). Functionally, vmPFC-NAc shell resting-state connectivity was inversely related to body mass index (BMI) and decreased in the obese state. Structurally, vmPFC-NAc shell structural connectivity and vmPFC thickness were inversely correlated with BMI; obese binge-prone participants exhibited decreased vmPFC-NAc structural connectivity and vmPFC thickness. Finally, to examine a causal link to binge eating, we directly probed this circuit in one binge-prone obese female using NAc deep brain stimulation in a first-in-human trial. Direct stimulation of the NAc shell subregion guided by local behaviorally relevant electrophysiology was associated with a decrease in number of weekly episodes of uncontrolled eating and decreased BMI. This study unraveled vmPFC-NAc shell circuit aberrations in obesity that can be modulated to restore control over eating behavior in obesity.

Rates of longitudinal change in 18 F-flortaucipir PET vary by brain region, cognitive impairment, and age in atypical Alzheimer's disease.

INTRODUCTION: Longitudinal positron emission tomography (PET) studies of tau accumulation in Alzheimer's disease (AD) have noted reduced increases or frank decreases in tau signal. We investigated how such reductions related to analytical confounds and disease progression markers in atypical AD. METHODS: We assessed regional and interindividual variation in longitudinal change on 18 F-flortaucipir PET imaging in 24 amyloid beta (Aß)+ patients with atypical, early-onset amnestic or non-amnestic AD plus 62 Aß- and 132 Aß+ Alzheimer's Disease Neuroimaging Initiative (ADNI) participants. RESULTS: In atypical AD, 18 F-flortaucipir uptake slowed or declined over time in areas with high baseline signal and older, more impaired individuals. ADNI participants had reduced longitudinal change in early Braak stage regions relative to late-stage areas. DISCUSSION: Results suggested radioligand uptake plateaus or declines in advanced neurodegeneration. Further research should investigate whether results generalize to other radioligands and whether they relate to changes of the radioligand binding site structure or accessibility.

Diminishing Efficacy of Prone Positioning With Late Application in Evolving Lung Injury.

OBJECTIVES: It is not known how lung injury progression during mechanical ventilation modifies pulmonary responses to prone positioning. We compared the effects of prone positioning on regional lung aeration in late versus early stages of lung injury. DESIGN: Prospective, longitudinal imaging study. SETTING: Research imaging facility at The University of Pennsylvania (Philadelphia, PA) and Medical and Surgical ICUs at Massachusetts General Hospital (Boston, MA). SUBJECTS: Anesthetized swine and patients with acute respiratory distress syndrome (acute respiratory distress syndrome). INTERVENTIONS: Lung injury was induced by bronchial hydrochloric acid (3.5 mL/kg) in 10 ventilated Yorkshire pigs and worsened by supine nonprotective ventilation for 24 hours. Whole-lung CT was performed 2 hours after hydrochloric acid (Day 1) in both prone and supine positions and repeated at 24 hours (Day 2). Prone and supine images were registered (superimposed) in pairs to measure the effects of positioning on the aeration of each tissue unit. Two patients with early acute respiratory distress syndrome were compared with two patients with late acute respiratory distress syndrome, using electrical impedance tomography to measure the effects of body position on regional lung mechanics. MEASUREMENTS AND MAIN RESULTS: Gas exchange and respiratory mechanics worsened over 24 hours, indicating lung injury progression. On Day 1, prone positioning reinflated 18.9% ± 5.2% of lung mass in the posterior lung regions. On Day 2, position-associated dorsal reinflation was reduced to 7.3% ± 1.5% (p < 0.05 vs Day 1). Prone positioning decreased aeration in the anterior lungs on both days. Although prone positioning improved posterior lung compliance in the early acute respiratory distress syndrome patients, it had no effect in late acute respiratory distress syndrome subjects. CONCLUSIONS: The effects of prone positioning on lung aeration may depend on the stage of lung injury and duration of prior ventilation; this may limit the clinical efficacy of this treatment if applied late.

Image- versus histogram-based considerations in semantic segmentation of pulmonary hyperpolarized gas images.

PURPOSE: To characterize the differences between histogram-based and image-based algorithms for segmentation of hyperpolarized gas lung images. METHODS: Four previously published histogram-based segmentation algorithms (ie, linear binning, hierarchical k-means, fuzzy spatial c-means, and a Gaussian mixture model with a Markov random field prior) and an image-based convolutional neural network were used to segment 2 simulated data sets derived from a public (n = 29 subjects) and a retrospective collection (n = 51 subjects) of hyperpolarized 129Xe gas lung images transformed by common MRI artifacts (noise and nonlinear intensity distortion). The resulting ventilation-based segmentations were used to assess algorithmic performance and characterize optimization domain differences in terms of measurement bias and precision. RESULTS: Although facilitating computational processing and providing discriminating clinically relevant measures of interest, histogram-based segmentation methods discard important contextual spatial information and are consequently less robust in terms of measurement precision in the presence of common MRI artifacts relative to the image-based convolutional neural network. CONCLUSIONS: Direct optimization within the image domain using convolutional neural networks leverages spatial information, which mitigates problematic issues associated with histogram-based approaches and suggests a preferred future research direction. Further, the entire processing and evaluation framework, including the newly reported deep learning functionality, is available as open source through the well-known Advanced Normalization Tools ecosystem.

Brain MRI Deep Learning and Bayesian Inference System Augments Radiology Resident Performance.

Automated quantitative and probabilistic medical image analysis has the potential to improve the accuracy and efficiency of the radiology workflow. We sought to determine whether AI systems for brain MRI diagnosis could be used as a clinical decision support tool to augment radiologist performance. We utilized previously developed AI systems that combine convolutional neural networks and expert-derived Bayesian networks to distinguish among 50 diagnostic entities on multimodal brain MRIs. We tested whether these systems could augment radiologist performance through an interactive clinical decision support tool known as Adaptive Radiology Interpretation and Education System (ARIES) in 194 test cases. Four radiology residents and three academic neuroradiologists viewed half of the cases unassisted and half with the results of the AI system displayed on ARIES. Diagnostic accuracy of radiologists for top diagnosis (TDx) and top three differential diagnosis (T3DDx) was compared with and without ARIES. Radiology resident performance was significantly better with ARIES for both TDx (55% vs 30%; P < .001) and T3DDx (79% vs 52%; P = 0.002), with the largest improvement for rare diseases (39% increase for T3DDx; P < 0.001). There was no significant difference between attending performance with and without ARIES for TDx (72% vs 69%; P = 0.48) or T3DDx (86% vs 89%; P = 0.39). These findings suggest that a hybrid deep learning and Bayesian inference clinical decision support system has the potential to augment diagnostic accuracy of non-specialists to approach the level of subspecialists for a large array of diseases on brain MRI.

Artificial Intelligence System Approaching Neuroradiologist-level Differential Diagnosis Accuracy at Brain MRI.

Background Although artificial intelligence (AI) shows promise across many aspects of radiology, the use of AI to create differential diagnoses for rare and common diseases at brain MRI has not been demonstrated. Purpose To evaluate an AI system for generation of differential diagnoses at brain MRI compared with radiologists. Materials and Methods This retrospective study tested performance of an AI system for probabilistic diagnosis in patients with 19 common and rare diagnoses at brain MRI acquired between January 2008 and January 2018. The AI system combines data-driven and domain-expertise methodologies, including deep learning and Bayesian networks. First, lesions were detected by using deep learning. Then, 18 quantitative imaging features were extracted by using atlas-based coregistration and segmentation. Third, these image features were combined with five clinical features by using Bayesian inference to develop probability-ranked differential diagnoses. Quantitative feature extraction algorithms and conditional probabilities were fine-tuned on a training set of 86 patients (mean age, 49 years ± 16 [standard deviation]; 53 women). Accuracy was compared with radiology residents, general radiologists, neuroradiology fellows, and academic neuroradiologists by using accuracy of top one, top two, and top three differential diagnoses in 92 independent test set patients (mean age, 47 years ± 18; 52 women). Results For accuracy of top three differential diagnoses, the AI system (91% correct) performed similarly to academic neuroradiologists (86% correct; P = .20), and better than radiology residents (56%; P < .001), general radiologists (57%; P < .001), and neuroradiology fellows (77%; P = .003). The performance of the AI system was not affected by disease prevalence (93% accuracy for common vs 85% for rare diseases; P = .26). Radiologists were more accurate at diagnosing common versus rare diagnoses (78% vs 47% across all radiologists; P < .001). Conclusion An artificial intelligence system for brain MRI approached overall top one, top two, and top three differential diagnoses accuracy of neuroradiologists and exceeded that of less-specialized radiologists. © RSNA, 2020 Online supplemental material is available for this article. See also the editorial by Zaharchuk in this issue.

Positional Therapy and Regional Pulmonary Ventilation.

BACKGROUND: Prone ventilation redistributes lung inflation along the gravitational axis; however, localized, nongravitational effects of body position are less well characterized. The authors hypothesize that positional inflation improvements follow both gravitational and nongravitational distributions. This study is a nonoverlapping reanalysis of previously published large animal data. METHODS: Five intubated, mechanically ventilated pigs were imaged before and after lung injury by tracheal injection of hydrochloric acid (2 ml/kg). Computed tomography scans were performed at 5 and 10 cm H2O positive end-expiratory pressure (PEEP) in both prone and supine positions. All paired prone-supine images were digitally aligned to each other. Each unit of lung tissue was assigned to three clusters (K-means) according to positional changes of its density and dimensions. The regional cluster distribution was analyzed. Units of tissue displaying lung recruitment were mapped. RESULTS: We characterized three tissue clusters on computed tomography: deflation (increased tissue density and contraction), limited response (stable density and volume), and reinflation (decreased density and expansion). The respective clusters occupied (mean ± SD including all studied conditions) 29.3 ± 12.9%, 47.6 ± 11.4%, and 23.1 ± 8.3% of total lung mass, with similar distributions before and after lung injury. Reinflation was slightly greater at higher PEEP after injury. Larger proportions of the reinflation cluster were contained in the dorsal versus ventral (86.4 ± 8.5% vs. 13.6 ± 8.5%, P < 0.001) and in the caudal versus cranial (63.4 ± 11.2% vs. 36.6 ± 11.2%, P < 0.001) regions of the lung. After injury, prone positioning recruited 64.5 ± 36.7 g of tissue (11.4 ± 6.7% of total lung mass) at lower PEEP, and 49.9 ± 12.9 g (8.9 ± 2.8% of total mass) at higher PEEP; more than 59.0% of this recruitment was caudal. CONCLUSIONS: During mechanical ventilation, lung reinflation and recruitment by the prone positioning were primarily localized in the dorso-caudal lung. The local effects of positioning in this lung region may determine its clinical efficacy.

Longitudinal progression of grey matter atrophy in non-amnestic Alzheimer's disease.

Recent models of Alzheimer's disease progression propose that disease may be transmitted between brain areas either via local diffusion or long-distance transport via white matter fibre pathways. However, it is unclear whether such models are applicable in non-amnestic Alzheimer's disease, which is associated with domain-specific cognitive deficits and relatively spared episodic memory. To date, the anatomical progression of disease in non-amnestic patients remains understudied. We used longitudinal imaging to differentiate earlier atrophy and later disease spread in three non-amnestic variants, including logopenic-variant primary progressive aphasia (n = 25), posterior cortical atrophy (n = 20), and frontal-variant Alzheimer's disease (n = 12), as well as 17 amnestic Alzheimer's disease patients. Patients were compared to 37 matched controls. All patients had autopsy (n = 7) or CSF (n = 67) evidence of Alzheimer's disease pathology. We first assessed atrophy in suspected sites of disease origin, adjusting for age, sex, and severity of cognitive impairment; we then performed exploratory whole-brain analysis to investigate longitudinal disease spread both within and outside these regions. Additionally, we asked whether each phenotype exhibited more rapid change in its associated disease foci than other phenotypes. Finally, we investigated whether atrophy was related to structural brain connectivity. Each non-amnestic phenotype displayed unique patterns of initial atrophy and subsequent neocortical change that correlated with cognitive decline. Longitudinal atrophy included areas both proximal to and distant from sites of initial atrophy, suggesting heterogeneous mechanisms of disease spread. Moreover, regional rates of neocortical change differed by phenotype. Logopenic-variant patients exhibited greater initial atrophy and more rapid longitudinal change in left lateral temporal areas than other groups. Frontal-variant patients had pronounced atrophy in left insula and middle frontal gyrus, combined with more rapid atrophy of left insula than other non-amnestic patients. In the medial temporal lobes, non-amnestic patients had less atrophy at their initial scan than amnestic patients, but longitudinal rate of change did not differ between patient groups. Medial temporal sparing in non-amnestic Alzheimer's disease may thus be due in part to later onset of medial temporal degeneration than in amnestic patients rather than different rates of atrophy over time. Finally, the magnitude of longitudinal atrophy was predicted by structural connectivity, measured in terms of node degree; this result provides indirect support for the role of long-distance fibre pathways in the spread of neurodegenerative disease. 10.1093/brain/awz091_video1 awz091media1 6041544065001.

Unstable Inflation Causing Injury. Insight from Prone Position and Paired Computed Tomography Scans.

RATIONALE: It remains unclear how prone positioning improves survival in acute respiratory distress syndrome. Using serial computed tomography (CT), we previously reported that "unstable" inflation (i.e., partial aeration with large tidal density swings, indicating increased local strain) is associated with injury progression. OBJECTIVES: We prospectively tested whether prone position contains the early propagation of experimental lung injury by stabilizing inflation. METHODS: Injury was induced by tracheal hydrochloric acid in rats; after randomization to supine or prone position, injurious ventilation was commenced using high tidal volume and low positive end-expiratory pressure. Paired end-inspiratory (EI) and end-expiratory (EE) CT scans were acquired at baseline and hourly up to 3 hours. Each sequential pair (EI, EE) of CT images was superimposed in parametric response maps to analyze inflation. Unstable inflation was then measured in each voxel in both dependent and nondependent lung. In addition, five pigs were imaged (EI and EE) prone versus supine, before and (1 hour) after hydrochloric acid aspiration. MEASUREMENTS AND MAIN RESULTS: In rats, prone position limited lung injury propagation and increased survival (11/12 vs. 7/12 supine; P = 0.01). EI-EE densities, respiratory mechanics, and blood gases deteriorated more in supine versus prone rats. At baseline, more voxels with unstable inflation occurred in dependent versus nondependent regions when supine (41 ± 6% vs. 18 ± 7%; P < 0.01) but not when prone. In supine pigs, unstable inflation predominated in dorsal regions and was attenuated by prone positioning. CONCLUSIONS: Prone position limits the radiologic progression of early lung injury. Minimizing unstable inflation in this setting may alleviate the burden of acute respiratory distress syndrome.

Human Disc Nucleotomy Alters Annulus Fibrosus Mechanics at Both Reference and Compressed Loads.

Nucleotomy is a common surgical procedure and is also performed in ex vivo mechanical testing to model decreased nucleus pulposus (NP) pressurization that occurs with degeneration. Here, we implement novel and noninvasive methods using magnetic resonance imaging (MRI) to study internal 3D annulus fibrosus (AF) deformations after partial nucleotomy and during axial compression by evaluating changes in internal AF deformation at reference loads (50 N) and physiological compressive loads (∼10% strain). One particular advantage of this methodology is that the full 3D disc deformation state, inclusive of both in-plane and out-of-plane deformations, can be quantified through the use of a high-resolution volumetric MR scan sequence and advanced image registration. Intact grade II L3-L4 cadaveric human discs before and after nucleotomy were subjected to identical mechanical testing and imaging protocols. Internal disc deformation fields were calculated by registering MR images captured in each loading state (reference and compressed) and each condition (intact and nucleotomy). Comparisons were drawn between the resulting three deformation states (intact at compressed load, nucleotomy at reference load, nucleotomy at compressed load) with regard to the magnitude of internal strain and direction of internal displacements. Under compressed load, internal AF axial strains averaged -18.5% when intact and -22.5% after nucleotomy. Deformation orientations were significantly altered by nucleotomy and load magnitude. For example, deformations of intact discs oriented in-plane, whereas deformations after nucleotomy oriented axially. For intact discs, in-plane components of displacements under compressive loads oriented radially outward and circumferentially. After nucleotomy, in-plane displacements were oriented radially inward under reference load and were not significantly different from the intact state at compressed loads. Re-establishment of outward displacements after nucleotomy indicates increased axial loading restores the characteristics of internal pressurization. Results may have implications for the recurrence of pain, design of novel therapeutics, or progression of disc degeneration.

Tidal changes on CT and progression of ARDS.

BACKGROUND: Uncertain prediction of outcome in acute respiratory distress syndrome (ARDS) impedes individual patient management and clinical trial design. OBJECTIVES: To develop a radiological metric of injurious inflation derived from matched inspiratory and expiratory CT scans, calibrate it in a model of experimental lung injury, and test it in patients with ARDS. METHODS: 73 anaesthetised rats (acid aspiration model) were ventilated (protective or non-protective) for up to 4 hours to generate a spectrum of lung injury. CT was performed (inspiratory and expiratory) at baseline each hour, paired inspiratory and expiratory images were superimposed and voxels tracked in sequential scans. In nine patients with ARDS, paired inspiratory and expiratory CT scans from the first intensive care unit week were analysed. RESULTS: In experimental studies, regions of lung with unstable inflation (ie, partial or reversible airspace filling reflecting local strain) were the areas in which subsequent progression of injury was greatest in terms of progressive infiltrates (R=0.77) and impaired compliance (R=0.67, p<0.01). In patients with ARDS, a threshold fraction of tissue with unstable inflation was apparent: >28% in all patients who died and ≤28% in all who survived, whereas segregation of survivors versus non-survivors was not possible based on oxygenation or lung mechanics. CONCLUSIONS: A single set of superimposed inspiratory-expiratory CT scans may predict progression of lung injury and outcome in ARDS; if these preliminary results are validated, this could facilitate clinical trial recruitment and individualised care.

Visualizing the Propagation of Acute Lung Injury.

BACKGROUND: Mechanical ventilation worsens acute respiratory distress syndrome, but this secondary "ventilator-associated" injury is variable and difficult to predict. The authors aimed to visualize the propagation of such ventilator-induced injury, in the presence (and absence) of a primary underlying lung injury, and to determine the predictors of propagation. METHODS: Anesthetized rats (n = 20) received acid aspiration (hydrochloric acid) followed by ventilation with moderate tidal volume (V(T)). In animals surviving ventilation for at least 2 h, propagation of injury was quantified by using serial computed tomography. Baseline lung status was assessed by oxygenation, lung weight, and lung strain (V(T)/expiratory lung volume). Separate groups of rats without hydrochloric acid aspiration were ventilated with large (n = 10) or moderate (n = 6) V(T). RESULTS: In 15 rats surviving longer than 2 h, computed tomography opacities spread outward from the initial site of injury. Propagation was associated with higher baseline strain (propagation vs. no propagation [mean ± SD]: 1.52 ± 0.13 vs. 1.16 ± 0.20, P < 0.01) but similar oxygenation and lung weight. Propagation did not occur where baseline strain was less than 1.29. In healthy animals, large V(T) caused injury that was propagated inward from the lung periphery; in the absence of preexisting injury, propagation did not occur where strain was less than 2.0. CONCLUSIONS: Compared with healthy lungs, underlying injury causes propagation to occur at a lower strain threshold and it originates at the site of injury; this suggests that tissue around the primary lesion is more sensitive. Understanding how injury is propagated may ultimately facilitate a more individualized monitoring or management.

Eigenanatomy: sparse dimensionality reduction for multi-modal medical image analysis.

Rigorous statistical analysis of multimodal imaging datasets is challenging. Mass-univariate methods for extracting correlations between image voxels and outcome measurements are not ideal for multimodal datasets, as they do not account for interactions between the different modalities. The extremely high dimensionality of medical images necessitates dimensionality reduction, such as principal component analysis (PCA) or independent component analysis (ICA). These dimensionality reduction techniques, however, consist of contributions from every region in the brain and are therefore difficult to interpret. Recent advances in sparse dimensionality reduction have enabled construction of a set of image regions that explain the variance of the images while still maintaining anatomical interpretability. The projections of the original data on the sparse eigenvectors, however, are highly collinear and therefore difficult to incorporate into multi-modal image analysis pipelines. We propose here a method for clustering sparse eigenvectors and selecting a subset of the eigenvectors to make interpretable predictions from a multi-modal dataset. Evaluation on a publicly available dataset shows that the proposed method outperforms PCA and ICA-based regressions while still maintaining anatomical meaning. To facilitate reproducibility, the complete dataset used and all source code is publicly available.

Decomposing cerebral blood flow MRI into functional and structural components: a non-local approach based on prediction.

We present RIPMMARC (Rotation Invariant Patch-based Multi-Modality Analysis aRChitecture), a flexible and widely applicable method for extracting information unique to a given modality from a multi-modal data set. We use RIPMMARC to improve the interpretation of arterial spin labeling (ASL) perfusion images by removing the component of perfusion that is predicted by the underlying anatomy. Using patch-based, rotation invariant descriptors derived from the anatomical image, we learn a predictive relationship between local neuroanatomical structure and the corresponding perfusion image. This relation allows us to produce an image of perfusion that would be predicted given only the underlying anatomy and a residual image that represents perfusion information that cannot be predicted by anatomical features. Our learned structural features are significantly better at predicting brain perfusion than tissue probability maps, which are the input to standard partial volume correction techniques. Studies in test-retest data show that both the anatomically predicted and residual perfusion signals are highly replicable for a given subject. In a pediatric population, both the raw perfusion and structurally predicted images are tightly linked to age throughout adolescence throughout the brain. Interestingly, the residual perfusion also shows a strong correlation with age in selected regions including the hippocampi (corr = 0.38, p-value <10(-6)), precuneus (corr = -0.44, p < 10(-5)), and combined default mode network regions (corr = -0.45, p < 10(-8)) that is independent of global anatomy-perfusion trends. This finding suggests that there is a regionally heterogeneous pattern of functional specialization that is distinct from that of cortical structural development.

Relating brain anatomy and cognitive ability using a multivariate multimodal framework.

Linking structural neuroimaging data from multiple modalities to cognitive performance is an important challenge for cognitive neuroscience. In this study we examined the relationship between verbal fluency performance and neuroanatomy in 54 patients with frontotemporal degeneration (FTD) and 15 age-matched controls, all of whom had T1- and diffusion-weighted imaging. Our goal was to incorporate measures of both gray matter (voxel-based cortical thickness) and white matter (fractional anisotropy) into a single statistical model that relates to behavioral performance. We first used eigenanatomy to define data-driven regions of interest (DD-ROIs) for both gray matter and white matter. Eigenanatomy is a multivariate dimensionality reduction approach that identifies spatially smooth, unsigned principal components that explain the maximal amount of variance across subjects. We then used a statistical model selection procedure to see which of these DD-ROIs best modeled performance on verbal fluency tasks hypothesized to rely on distinct components of a large-scale neural network that support language: category fluency requires a semantic-guided search and is hypothesized to rely primarily on temporal cortices that support lexical-semantic representations; letter-guided fluency requires a strategic mental search and is hypothesized to require executive resources to support a more demanding search process, which depends on prefrontal cortex in addition to temporal network components that support lexical representations. We observed that both types of verbal fluency performance are best described by a network that includes a combination of gray matter and white matter. For category fluency, the identified regions included bilateral temporal cortex and a white matter region including left inferior longitudinal fasciculus and frontal-occipital fasciculus. For letter fluency, a left temporal lobe region was also selected, and also regions of frontal cortex. These results are consistent with our hypothesized neuroanatomical models of language processing and its breakdown in FTD. We conclude that clustering the data with eigenanatomy before performing linear regression is a promising tool for multimodal data analysis.

Histology-derived volumetric annotation of the human hippocampal subfields in postmortem MRI.

Recently, there has been a growing effort to analyze the morphometry of hippocampal subfields using both in vivo and postmortem magnetic resonance imaging (MRI). However, given that boundaries between subregions of the hippocampal formation (HF) are conventionally defined on the basis of microscopic features that often lack discernible signature in MRI, subfield delineation in MRI literature has largely relied on heuristic geometric rules, the validity of which with respect to the underlying anatomy is largely unknown. The development and evaluation of such rules are challenged by the limited availability of data linking MRI appearance to microscopic hippocampal anatomy, particularly in three dimensions (3D). The present paper, for the first time, demonstrates the feasibility of labeling hippocampal subfields in a high resolution volumetric MRI dataset based directly on microscopic features extracted from histology. It uses a combination of computational techniques and manual post-processing to map subfield boundaries from a stack of histology images (obtained with 200µm spacing and 5µm slice thickness; stained using the Kluver-Barrera method) onto a postmortem 9.4Tesla MRI scan of the intact, whole hippocampal formation acquired with 160µm isotropic resolution. The histology reconstruction procedure consists of sequential application of a graph-theoretic slice stacking algorithm that mitigates the effects of distorted slices, followed by iterative affine and diffeomorphic co-registration to postmortem MRI scans of approximately 1cm-thick tissue sub-blocks acquired with 200µm isotropic resolution. These 1cm blocks are subsequently co-registered to the MRI of the whole HF. Reconstruction accuracy is evaluated as the average displacement error between boundaries manually delineated in both the histology and MRI following the sequential stages of reconstruction. The methods presented and evaluated in this single-subject study can potentially be applied to multiple hippocampal tissue samples in order to construct a histologically informed MRI atlas of the hippocampal formation.

Subject-specific functional parcellation via prior based eigenanatomy.

We present a new framework for prior-constrained sparse decomposition of matrices derived from the neuroimaging data and apply this method to functional network analysis of a clinically relevant population. Matrix decomposition methods are powerful dimensionality reduction tools that have found widespread use in neuroimaging. However, the unconstrained nature of these totally data-driven techniques makes it difficult to interpret the results in a domain where network-specific hypotheses may exist. We propose a novel approach, Prior Based Eigenanatomy (p-Eigen), which seeks to identify a data-driven matrix decomposition but at the same time constrains the individual components by spatial anatomical priors (probabilistic ROIs). We formulate our novel solution in terms of prior-constrained ℓ1 penalized (sparse) principal component analysis. p-Eigen starts with a common functional parcellation for all the subjects and refines it with subject-specific information. This enables modeling of the inter-subject variability in the functional parcel boundaries and allows us to construct subject-specific networks with reduced sensitivity to ROI placement. We show that while still maintaining correspondence across subjects, p-Eigen extracts biologically-relevant and patient-specific functional parcels that facilitate hypothesis-driven network analysis. We construct default mode network (DMN) connectivity graphs using p-Eigen refined ROIs and use them in a classification paradigm. Our results show that the functional connectivity graphs derived from p-Eigen significantly aid classification of mild cognitive impairment (MCI) as well as the prediction of scores in a Delayed Recall memory task when compared to graph metrics derived from 1) standard registration-based seed ROI definitions, 2) totally data-driven ROIs, 3) a model based on standard demographics plus hippocampal volume as covariates, and 4) Ward Clustering based data-driven ROIs. In summary, p-Eigen incarnates a new class of prior-constrained dimensionality reduction tools that may improve our understanding of the relationship between MCI and functional connectivity.