Jonathan R. Polimeni, Ph.D. - US grants

DESCRIPTION (provided by applicant): This project will support the training and career development of a junior faculty member, with prior training in computational neuroscience and electrical engineering, transitioning into the fields of magnetic resonance imaging (MRI) and functional neuroimaging. This training will take place at the A. A. Martinos Center for Biomedical Imaging at the Massachusetts General Hospital, under the mentorship of Prof. L. L. Wald, within the Ultrahigh-field Imaging and Imaging Physics Group. The candidate will conduct a study into quantifying the fundamental biological limits of spatial resolution in functional MRI, and perform precise measurements of the functional architecture of the human visual system using novel methods developed to overcome resolution limits placed by the instrumentation, data acquisition and experimental design, and data analysis. The long-term objective of this project is to enable non-invasive imaging of fine-scale details of the human visual cortex, including the distinctive spatial maps of orientation preference, ocular dominance, and retinotopy, with a spatial resolution sufficient to derive accurate, quantitative measurements of these basic features of the visual system. To quantify the biological limits of spatial resolution, this study will focus on three aims: (i) to develop a methodology for quantifying spatial resolution and accuracy in fMRI;(ii) to measure spatial accuracy across multiple experimental designs and identify which provides the highest achievable resolution;and (iii) to exploit this knowledge to measure and quantify the topographic and columnar structures in primary visual cortex, and thus draw informed conclusions about their organization based on the known measurement accuracy. Although estimates of spatial resolution have been made in the past, new advances in both acquisition and analysis technology, and new insights into experimental design, require that these estimates be re-assessed to determine what is now feasible. Importantly, emerging methods at our disposals enable resolving activity within individual cortical laminae. Not only does laminar fMRI open possibilities for testing new hypotheses about the nervous system and neurovascular coupling, but the proposed methods may yield a practical technique for increasing spatial resolution-due to the tighter biological point-spread expected in central vascular layers distal to large pial veins, targeted sampling of these layers will enable higher achievable spatial resolution. The candidate will receive training in ultrahigh-field imaging methods, accelerated parallel imaging techniques, design and construction of radiofrequency coil detectors, accurate computational analysis of fMRI data, and the anatomy and physiology of the human brain and its vascular system. The tools developed for this study can assist in several applications such as identifying pathological tissue in patients with visual deficits or amblyopia, measuring the impact of localized hyperemia in patients with occipital cerebral amyloid angiopathy, designing cortical prostheses, and will enable future studies into the fine organization of the nervous system. PUBLIC HEALTH RELEVANCE: PROJECT NARRATIVE The spatial accuracy of functional MRI is limited by the biology of blood delivery. We will impose spatial patterns of activity along the cortex to measure the spatial accuracy in individual cortical layers, and use these patterns to test methods for further improving accuracy.

? DESCRIPTION (provided by applicant): The objective of this project is to test the joint hypotheses that sampling the functional MRI (fMRI) signals up to an order of magnitude more rapidly can help extract information related to neuronal signaling; and that the hemodynamic signals that form the basis of fMRI, rather than being sluggish as is commonly believed, respond rapidly and precisely to neuronal activity. Rapid sampling is commonly advocated to enable physiological noise removal-because these systemic noise sources can then be adequately sampled and so are not aliased in the raw fMRI signal-however our goal is to demonstrate that the fMRI signal at short time scales also contains fluctuations that are directly driven by neuronal activation. While the blood-oxygen-level-dependent (BOLD) response is well known to peak 6 s following the onset of neuronal activity, the initial vascular response begins in less tha 1 s. Here we challenge the notion that the BOLD response is slow. We will capitalize on our recent development of Simultaneous Multi-Slice (SMS) imaging for fMRI, which provides temporal sampling that is 12× faster than that of conventional techniques. With the SMS method, the fMRI measurement possesses the temporal resolution to detect brain activation over the entire brain with sub-second precision. Previous work has demonstrated that fMRI time series data acquired with high sampling rates can be used to parcellate global brain networks into smaller nodes, and therefore increase detection power in resting- state functional connectivity studies. Here we propose to extend this key benefit to other common fMRI experimental paradigms. Our preliminary data suggests that, by acquiring fMRI data on a finer time scale using a conventional task-driven block-design paradigm, dramatic increases in detection sensitivity up to factors of 2-3 are achievable. In these cases, faster sampling yields increased sensitivity. This boost will enable new classes of experiments, as well as single-subject analyses and potentially individualized diagnosis. Recent invasive animal neurovascular coupling studies and human fMRI studies have shown that the early stages of the BOLD response are precisely controlled by local vascular responses, and the BOLD response spreads spatially with time. High spatio-temporal resolution fMRI acquisitions can therefore enable higher accuracy by sampling the early phases of the BOLD response. In these cases, faster sampling yields increased specificity. Finally, we will test whether rapid fMRI can help (i) extrac information from continuous, temporally- encoded stimulus designs and (ii) resolve neuronal activations occurring closely in time. For the latter, we will implement a novel calibration procedure designed to remove regional variations in vascular delay from the measured delays in the BOLD response to accurately estimate the neuronal activation onset. Here, faster sampling yields additional information about neuronal function and activation latencies of the brain. Our aim is to demonstrate the benefits of rapid fMRI in these domains and to develop acquisition and analysis frameworks for the inevitable widespread use of this transformative new approach to fMRI.

PROJECT SUMMARY/ABSTRACT All fMRI signals have a vascular origin, and this has been believed to be a major limitation to precise spatiotemporal localization of neuronal activation when using hemodynamic functional contrast such as BOLD. However, significant recent discoveries made using powerful ultrahigh-resolution optical imaging techniques have challenged this belief. Unfortunately these measures require invasive procedures and therefore cannot be performed in humans. Our aim is to transfer knowledge gained from these invasive studies into interpreting human fMRI data in order to help fMRI reach its full potential. In this proposal we plan to combine detailed maps of human macro- and meso-scale vasculature measured with high-resolution MRI with maps of the micro-scale vasculature measured in human brain specimens with CLARITY-assisted microimaging. We will then link this anatomical information with dynamic models built from 2-photon microscopy performed in rodents where the changes in vessel diameter, blood flow and oxygenation can be measured directly in each vessel type across all stages of the vascular hierarchy. We hypothesize that newly introduced models of hemo- and vaso-dynamics built from 2-photon microscopy, linked with a detailed micro- and macroscopically mapped human microvascular anatomy, can be exploited to improve the spatial and temporal specificity of human fMRI. To supply human vasculature reconstructions to our models, we propose a two-scale approach. We first advance 7 Tesla MR Angiography (MRA) techniques to image the pial vascular network as well as intracortical vessels and vascular layers of the cerebral cortex to achieve a mesoscopic model. To form the micron-scale model of vasculature at the capillary level, we will use the CLARITY technique to image the full vascular tree (from arterioles through capillaries to venules) in human primary visual cortex. To predict vasodynamic changes driven by neuronal activation, we will adapt a model derived from dynamic in vivo 2-photon microscopy of vessel diameters in rodents to human microvascular anatomy. To adapt this to human microvasculature requires a careful multi-stage transferal. First we will measure bulk changes in microvessel diameter, a.k.a. cerebral blood volume (CBV), across multiple levels of the vascular hierarchy and confirm that the model can predict the CBV-fMRI signal. The CBV-fMRI signal is used because it is a vasodynamic signal directly reflecting vessel diameter changes occurring alongside local neuronal activity (rather than the subsequent hemodynamic changes). After performing this validation we will build a dynamic model of the microvascular tree in human cortex based on our vascular reconstruction, and again measure CBV-fMRI changes across multiple levels of the vascular hierarchy. We will finally test the ability of this model to improve the neuronal specificity of fMRI by imaging the functional architecture in human visual cortex. This model will also enable the formulation and testing of hypotheses about the discriminability of fMRI responses elicited from nearby neuronal populations, and guide development of future advanced acquisition technologies.

Project Abstract: Functional magnetic resonance imaging (fMRI) studies have produced much of our current knowledge about the functional organization of the human brain. fMRI studies make inferences about neural activity by measuring changes in local blood oxygenation, typically assuming a canonical hemodynamic response function (HRF) that evolves over several seconds. Due to these assumed slow hemodynamics, while fMRI has been a powerful tool for localizing brain activity, it has mostly been limited to mapping activity patterns at low temporal resolution. However, recent studies have demonstrated that the hemodynamics can react much more quickly than previously thought, and that fast fMRI can be used to directly image responses to oscillatory neural activity, suggesting new possibilities for applying fMRI to study fast brain activity. Many high-level cognitive processes, such as attention, language, and perceptual awareness, are associated with 0.1?1 Hz dynamics that cannot currently be precisely localized with noninvasive methods, and these could potentially be studied directly using fast fMRI approaches. An essential step towards using fast fMRI for neuroscience investigations is to identify the physiological properties of these fast BOLD responses: what mechanisms generate fast hemodynamic responses, and how fast and local can these responses be? This project aims to identify the neurovascular dynamics that enable fast fMRI responses, and advance the ultimate spatiotemporal resolution that can be achieved when using fast and ultra high field fMRI. We hypothesize that prolonged and rapidly varying neural activity causes a slow plateau in baseline vessel dilation, which allows fast fluctuations in blood flow to the stimulated brain region, and thus leads to fast and spatially localized responses. This `sustained dilation' hypothesis makes several specific predictions that we will test through advanced methods for human brain imaging of cerebral blood flow, cerebral blood volume, and blood oxygen level dependent (BOLD) fMRI, to identify the vascular dynamics underlying fast responses. We will then advance our acquisition and analysis approaches to explore the boundaries of the possible spatiotemporal resolution of fast BOLD fMRI. Using ultra high field imaging we will test whether oscillatory stimuli elicit more spatially precise hemodynamic responses, across the cortical surface and laminar depth, providing a new experimental paradigm that leads to more neuronally specific fMRI responses. We will also test whether this spatial information can be exploited to detect even faster (up to 1 Hz) responses by selectively analyzing voxels in the middle cortical depths. These aims will identify the vascular dynamics underlying rapid fMRI responses, and will test the limits of the possible spatiotemporal resolution of neural activity at faster timescales. The ultimate goal is to provide a scientific and technical foundation enabling future fMRI studies to directly map 0.1-1 Hz neural dynamics throughout the human brain.

PROJECT SUMMARY / ABSTRACT Functional MRI (fMRI) is the most widely-used tool to noninvasively measure brain function and has produced much of our current knowledge about the functional organization of the human brain. All fMRI methods, however, measure neuronal activity indirectly by tracking the associated local changes in blood flow and oxygenation. While this is often viewed as a limitation of fMRI, recent optical imaging studies in animal models have shown that, surprisingly, the smallest blood vessels in the brain respond rapidly to local neuronal activity, and are thus tightly coupled to neurons, suggesting that fMRI could provide a far more veridical picture of neuronal activity than previously believed?if one can measure fMRI signals such as BOLD exclusively from the microvasculature. In the previous funding cycle, we successfully tested our hypothesis that the neuronal specificity of fMRI can be improved by restricting analyses to the earliest phases of the standard gradient-echo BOLD response, thought to occur in the microvasculature, before the responses spread to larger blood vessels and become less spatially localized. The ability to reliably extract the earliest phases of the BOLD response was achieved through the fast temporal sampling made possible through the fMRI acquisition technologies we developed. Our findings were consistent with our hypothesis?the fastest component of the BOLD response provides the highest microvascular specificity. Here we test the converse hypothesis: that BOLD signals from the microvasculature are fastest and exhibit the highest temporal specificity, while signals from the macrovasculature are temporally delayed and smeared. To test this we will develop technologies for spin-echo BOLD, which exclusively measures from the microvasculature, with fast temporal sampling. In this cycle our central hypothesis is that spin-echo BOLD with exclusive sensitivity to the microvasculature, will yield higher temporal specificity than gradient-echo BOLD, which contains slower signals from the macrovasculature. The challenge is that spin-echo acquisitions in theory provide T2 weighting, endowing spin-echo BOLD with microvascular specificity, however in practice it is difficult to achieve pure T2 weighting. Thus, our goals are to develop and validate fMRI technologies for robust pure T2-weighted BOLD, and to test whether pure T2-weighted BOLD provides higher temporal specificity. These goals can only be achieved by combining several novel MRI technologies we have recently introduced. The core technology is Echo-Planar Time-Resolved Imaging (EPTI), an extension to Echo-Planar Imaging (EPI), which can provide pure T2-weighted BOLD?concurrently with T2*-weighted BOLD, enabling direct comparisons. We will combine this with our new methods for increasing temporal sampling efficiency through and motion-robustness, and maximizing signal when using fast sampling rates. Finally, all experiments will be performed at 7 Tesla, where BOLD exhibits stronger microvascular weighting and higher sensitivity compared to standard field strengths, using parallel transmit RF pulse designs to reduce power deposition and improve the spatial uniformity of fMRI sensitivity.

PROJECT ABSTRACT/SUMMARY The cerebellum is the second largest structure in the human brain. Until recently, the cerebellum was considered solely a motor structure. Discoveries from neuroanatomy, study of patients with cerebellar lesions, and human neuroimaging have all converged to suggest that major zones of the cerebellum participate in advanced forms of cognition. Relevant to mental health, cerebellar dysfunction has been implicated in psychiatric illness and preliminary reports suggest noninvasive stimulation of the cerebellum may benefit symptoms in schizophrenia. However our detailed understanding of cerebellar organization is far behind that of the cerebral cortex, leading to debates about the spatial organization of cerebellar zones linked to human thought and emotion, and even debate about the degree to which the functional organization of the cerebellum is consistent from one person to the next. The goal of the present work is to provide a detailed understanding of cerebellar organization with particular focus on zones implicated in higher-order cognition that include regions accessible to neuromodulation. (1) First, advanced human high-field MRI methods will be used to map networks across the cerebellum fully within individuals preserving the anatomical details that would otherwise be lost with lower resolution approaches or by averaging findings across individuals. To achieve this level of precision each individual will be imaged repeatedly. (2) Second, to establish that the spatially separate regions of the cerebellum are functionally distinct, the same individuals will be administered challenging tasks that probe language, social, and memorial functions to rigorously establish separation between cerebellar zones that may be as little as a few millimeters apart. (3) Enabled by the precision maps of cerebellar organization, open debates will be resolved that include questions about how many times high-order cognitive zones repeat across the cerebellum and whether small, difficult to map, isolated zones of function contribute to the uniqueness of each person?s brain. (4) Critical to the long-term objective of this work to benefit patient care, the precision maps of each individual?s cerebellum will be used to model the possible effects of non-invasive stimulation. In doing so, a path from precision mapping of the cerebellum to neuromodulation will be provided openly as well as high-resolution maps and raw data that can be utilized by the community to further improve available methods for neuromodulation. Most broadly, the present work seeks to better understand the detailed organization of the human cerebellum to serve as a foundation for understanding and further developing novel interventions in the battle against mental illness.

Today, the most widespread tool for measuring whole-brain activity noninvasively is functional magnetic resonance imaging (fMRI). Although fMRI tracks neural activity indirectly through measuring the associated changes in blood flow, volume and oxygenation, recent evidence has suggested that these active hemodynamic changes in the brain are far more precisely coordinated than previously believed, perhaps at the fine spatial scale of the basic modules of functional architecture: cerebral cortical columns and layers. If true, this could enable new studies of brain computation and circuitry as several cortical layers are the well- known inputs and outputs along canonical feedforward and feedback pathways of brain communication. The main challenge faced by this emerging field of ?laminar fMRI? is how to interpret the complex hemodynamic signals to infer the underlying patterns of neural activity. Motivated by this, our overall goal is to improve our ability to measure neural activity from distinct cortical layers with human fMRI through detailed biophysical modeling of the underlying hemodynamic response. We will develop a new computational framework to simulate the fMRI signals using realistic microvascular networks and dynamics of associated blood flow, volume, and oxygenation changes that accompany neural activity. This framework has been validated using optical microscopy measures of the microvascular anatomy and dynamics from small animal models, and here we extend it for the first time to the human cortex. We will combine ultra-high-resolution in vivo vascular anatomical imaging data collected at 9.4 Tesla with our validated algorithm for synthesizing realistic microvascular networks to generate human vascular models specific to individual volunteers, and use these to simulate fMRI responses to motor tasks designed to activate specific cortical layers. We will then simulate responses of several forms of fMRI contrast?that are each sensitive to different aspects of the complex hemodynamic response?and compare our predictions to high-resolution fMRI measurements. Finally, to gain insight into whether fMRI can be used correctly to infer neural activity within cortical layers, we will quantify the discriminability of laminar fMRI by simulating various patterns of neural activity across layers and then comparing the computed fMRI activation profiles. This will tell us which neural activity patterns can be distinguished from one another, and which cannot, to help quantify the ability of laminar fMRI to decipher human brain circuitry. We address a fundamental gap in our knowledge regarding the limits of human fMRI: whether fMRI can accurately report on activation within distinct cortical layers. Our approach will allow us to quantify how fMRI sees the neural activity through the ?filter? of the vascular response, and provide insight into the origins of newly-available fMRI contrasts. This will aid in the interpretability of fMRI for both neuroscience research as well as for translational/clinical research by helping to remove unwanted effects of the vasculature?to translate the observed fMRI patterns into neural activity patterns to better understand brain function in health and disease. RELEVANCE (See instructions): Functional magnetic resonance imaging (fMRI) is the most widespread tool for measuring activity across the entire brain noninvasively and has produced much of our knowledge of the functional organization in the human brain, however fMRI does not measure neuron firing?it detects brain activity by measuring changes in blood flow in the brain that delivers oxygen to the neurons. Here we seek to develop an analysis framework that will allow us to more accurately infer which groups of neurons are firing based on human fMRI data by using computational modeling of blood flow and blood oxygenation changes through networks of the smallest blood vessels in the brain. Today fMRI is indispensable for experimental human neuroscience; by improving the neural specificity of this technique, fMRI can become a more reliable tool for measuring brain function in health and disease, expanding its utility in basic neuroscience and translational/clinical research including investigations into neurological and psychiatric disease, and may also provide deeper mechanistic understanding into small vessel disease and other vascular contributions to cognitive impairment and dementia. PHS 398 (Rev. 03/2020 Approved Through 02/28/2023) OMB No. 0925-0001 Page 2 Form Page 2 Program Director/Principal Investigator (Last, First, Middle): Polimeni, Jonathan Rizzo