Sensory perception and behavior require the coordinated activity of large-scale networks of neurons comprising tens to hundreds of thousands of cells acting together. These large-scale functional assemblies operate at a broad range of spatial scales, from coordinating activity among adjacent neurons separated by tens of microns to patterns of activity across brain regions separated by tens of centimeters. Interrogating these large-scale functional networks, therefore, requires a broad range of tools operating across a broad range of spatial resolutions, including electrophysiology, optical imaging, and fMRI. In this talk, I will present our recent findings using optical approaches to examine how millimeter-scale networks involving tens of thousands of neurons emerge during early development. Our data reveal a seemingly universal pattern of initially highly dense self-organizing networks dominated by local recurrent circuitry operating across highly diverse brain regions during early development. These networks then undergo a common coarse-to-fine developmental maturation, becoming increasingly sparse and high dimensional. This common developmental program may serve to build efficient and spatially organized functional representations while allowing sparse and efficient coding in the mature brain.

fMRI at high field strengths (≥ 7T) has enabled unparalleled visualization of functional detail at a laminar or columnar level, bringing fMRI closer to the intrinsic resolution of brain function. Since fMRI measures neuronal activity indirectly via hemodynamics, its ultimate resolution depends on the spatial specificity of hemodynamics to neuronal activity at a detailed spatial scale and on how this specificity evolves over time. A better understanding of neuronal and vascular contributions to the fMRI signal can thus aid in extracting neurophysiological information from fMRI in neuroscience, as well as identifying cortical vessel health markers in the context of neurovascular diseases. This talk will discuss factors that affect the spatial spread and temporal dynamics of hemodynamic responses across cortical depth, with a focus on the laminar vascular organization. It will provide an overview of hemodynamics across the cortical vascular tree based on optical imaging studies, compared to spatiotemporal features of hemodynamics observed across cortical depth in high resolution fMRI. Contributions to the fMRI signal that arise from vascular processes, including macro- and microvascular reactivity and the point-spread-function of BOLD responses, will be addressed. The talk will also give a view of the how the vascular architecture and resulting hemodynamics affect BOLD signals across cortical depth based on simulations using realistic vascular models. Approaches to probe vascular and neurovascular processes across cortical depth will also be discussed.

To be updated...

The introduction of ultra-high field fMRI has opened new opportunities for imaging the human brain at high spatial resolution, enabling testing of neuroscientific theories of brain functioning at the mesoscopic level. In this talk, I will focus on a widely referenced framework, predictive coding (PC), that proposes that the brain continuously generates predictions about incoming sensory information and updates an internal model of the world around us when expectations are violated. Within this framework, different functions are attributed to distinct cortical layers, making high-resolution fMRI a valuable tool for testing hypotheses of PC. I will focus on the auditory cortex as a testbed for examining the implementation of PC. To probe prediction- and error signals, we collected submillimeter data using gradient echo BOLD while presenting stimuli designed to elicit mismatch signals through the violations of expectation. While gradient echo BOLD provides high sensitivity for detecting responses, it is also subject to confounding influences from draining veins, which can obscure the neural origin of measured responses. To address this limitation, we employed a biophysical model that allows the estimation of underlying neural dynamics while taking vascular contributions of draining veins into account. That is, this modeling framework allows us to untangle neural activity from vascular signals. By disentangling these components, we aim to clarify how auditory cortical layers encode expectation violations and model updating signals, offering insights into the neural implementation of PC in the human auditory brain.

The striatum plays a central role in motor planning, habit formation, and sequence learning. Previous studies have shown that striatal activity correlates with motor learning across repeated sessions and different learning phases. However, these studies typically rely on group-level analyses to overcome the inherently low signal-to-noise ratio (SNR) in the basal ganglia (BG), which is largely due to high iron concentrations in these nuclei. The low SNR significantly limits the ability to characterize striatal activity with high spatiotemporal precision. Recent advances in ultra-high-field 7T MRI have enabled the investigation of brain–behavior relationships at the individual level, an approach known as precision mapping. In this study, we apply precision mapping to investigate correlations between striatal subregional activity—specifically within the caudate and putamen—and motor performance during a finger-tapping task involving simple and complex sequences. To achieve the necessary temporal SNR, we evaluated multi-echo EPI (ME-EPI), which has shown benefits for imaging BG, and fast-TR (~300 ms) acquisitions, which improve statistical sensitivity. To further enhance signal quality, we applied NORDIC (NOise Reduction with DIstribution Corrected principal component analysis) denoising. This methodology enabled us, for the first time, to observe significant correlations between striatal subregion activity and motor performance at the individual subject level.

To be updated...

Aging and neurodegeneration affect human sensorimotor cortex structure and function. Yet, the mechanisms that underlie circuit stability and vulnerability as well as disease spread are to a large extent unclear. Layer-specific multi-modal MRI is a valuable tool to develop in-vivo microstructural models of human primary motor cortex (MI) and primary somatosensory cortex (SI) that can be probed with respect to their vulnerability to aging and neurodegeneration, and their sensitivity to plasticity and intervention. I will present recent insights on layer-specific changes in the architecture of somatotopic maps in MI and SI that occur in the course of healthy aging and neurodegeneration, as well as with missing sensory input. These insights for the first time allow us to develop layer-specific in-vivo models of cortical aging, plasticity, and pathology that will help us to target sensorimotor networks more specifically with respect to potential loss-of-functions, or in the course of plasticity and learning. I will argue that 3D models of the human cortex are needed to capture the full range of cortical functioning and dysfunctioning that characterize the system in health and disease.

To be updated...

Human neuroscience is experiencing a growing interest in the acquisition and openly sharing of large-scale functional magnetic resonance imaging (fMRI) datasets. Initial large-scale fMRI data sets have focused on either ‘wide’ sampling: acquiring a few hours of brain data from many participants (n ≥ 100; e.g., the Human Connectome Project (Van Essen et al., 2013)), or ‘deep’ sampling: acquiring many hours of brain data from a few participants (n ≤ 20; e.g., the Midnight Scan Club (Gordon et al. 2017)). By collecting many hours of data from a small group of participants, these deep datasets have enabled detailed investigation of brain structure and function. In this talk, I will highlight an emerging deep sampling approach in fMRI, which we term 'intensive’ fMRI, that aims to extensively sample cognitive phenomena within a small group of individuals to support within-subject computational modeling at the voxel level. I will discuss the key characteristics of intensive fMRI: to create datasets with well-designed experiments that enable a rich hypothesis space, that maximize both data quantity and quality, and that serve as a valuable community resource. Informed by efforts creating the Natural Scenes Dataset (Allen et al., 2022) and the upcoming Visual Cognition Dataset, I will address practical considerations and challenges of intensive fMRI, including optimizing trial and experimental design and screening and selecting participants to maximize data quality. When done well, intensive fMRI datasets enable better models of human cognition and bridge multiple neuroscience communities.

Human functional neuroimaging has traditionally sought to understand the organization of the brain using group-averaging approaches, in which many individuals are scanned and their functional responses are averaged across people at each point in the brain. However, individual brains are variable in their organization, and group-averaging blurs across that variability. Here, I will present an alternative approach termed Precision Functional Mapping, in which large amounts of neuroimaging data are collected from each individual in order to precisely map the functional organization of each person’s brain. I will discuss how this approach greatly improves the reliability of our functional neuroimaging measures. I will demonstrate how it enables us to not only better understand how individual brains differ from each other, but also to identify novel features of brain organization that cannot be clearly observed using group averaged data. I will further discuss how Precision Functional Mapping enable new understandings of neurological and neuropsychiatric disorders, and how it holds the key to developing patient-centric neuromodulation treatments.

Imaging at ultra-high field strengths, such as 7 T, is challenged by RF wave effects and the absence of an integrated body coil, which limits workflow efficiency. Currently, clinical imaging at 7 T is restricted to the head and knee, where dedicated hardware solutions exist. For other body regions, a specialized transmit array is required—often designed as a compromise between transmission and reception in a single transmit/receive array.

A dedicated, all-purpose transmit array, used in conjunction with optimized receive-only arrays, would significantly improve the clinical feasibility of 7 T imaging. Such receive arrays can be made lighter, thinner, and more SNR-efficient, enhancing both patient comfort and image quality.

In this talk, I will present ongoing work at DKFZ on a 32-channel transmit system featuring an integrated transmit array and dedicated receive coils. The audience will gain an overview of the technologies involved and the practical implementation of the transmit chain. This includes a 32-channel add-on transmit system with high-power, high-duty-cycle amplifiers, IQ modulators, and a dedicated safety supervision architecture. Integration into the MRI system’s existing architecture will also be discussed.

The presentation will include numerous imaging examples and comparisons with other transmit array configurations. It will be demonstrated that this technology—already successfully used in clinical settings with single-channel transmit coils and local receive arrays—can bring similar benefits to 7 T imaging when a single-channel body coil is replaced by a suitable transmit array.”

Siemens developed a high power Gradient for UHF – should we call it “the beast”? At least that was my suggestion…. Well, Siemens marketing preferred to call it “Impulse”. Currents in such a gradient coil run with maximum 1200A along the wires and they are doing this in a 7T field, creating Lorentz forces up to ~ 8kN. In total there is more than 400m of gradient wire and all that can be pulsed with slewrate 900T/m/s. Quite a beast… Giving everyone that tries to tame it a hard time: Such forces try to tear the body of the gradient coil to pieces. And if you succeed in avoiding it, that gradient coil will turn into one of the most powerful shakers ever invented. Which will raise some new challenges like extreme acoustic noise levels radiated from its surface. Or the big challenge how to attach parts to it: Gradient cables, mounting structures, shim trays – all threatened by the beast that tries to shake them to “death” or melt them by friction heating.

But it is not only the mechanics that raise some challenges: The fast switching gradient field will stimulate human nerves and turn any conductive part in the vicinity into an efficient eddy current driven heater – even if these parts are intended to be cooling devices…And be aware that it does not need the full gradient field to put some MR components under steam: Even the tiny portion of field leaking towards the magnet is a quite infamous predator looking for  UHF magnet prey…Taming UHF gradients is a big and multidimensional challenge, with destruction waiting just around each corner, no matter what the name of such a gradient coil will be.

To be updated...

To be updated...

To be updated...

The next-generation (NexGen) 7 Tesla MR imaging scanner reaches ultra-high resolution by implementing several advances in hardware, including a head-only asymmetric gradient coil (200 mT m−1, 900 T m−1s−1) with an additional third layer of windings, a 128-channel receiver system with 64- and 96-channel receiver coil arrays to boost signal in the cerebral cortex while reducing g-factor noise to enable higher accelerations, and a 16-channel transmit system reduced power deposition and improved image uniformity. EPI, 3D GRASE and 3D Spiral sequences take advantage of the higher slew rate and Gmax to reach mesoscale resolution.

The NexGen 7 T scanner achieved an important milestone by extending fMRI studies to the mesoscale below 0.1 μl voxel volume, including the use of CBV-weighted VASO imaging at 0.39 mm isotropic resolution to differentiate layer-specific activation in V1 of the human visual cortex that could not be resolved in humans before. Information processing with network analysis tools to study interactions across many areas of the entire brain is now possible, where fMRI responses can be localized to the fundamental modules of brain computation across cortical depth, allowing investigation of brain circuitry with a firmer basis for modeling of neurocircuitry using fMRI connectomics.

In diffusion imaging, the combination of higher SNR at 7 T with achievable higher amplitude gradients enables both higher spatial, temporal and angular resolutions, creating a unique high performance platform for additional neuroscientific microstructural explorations. Gradient coils specifically designed for diffusion imaging at 3T can reach higher Gmax, however, the Impulse gradient is operational at 7 T with the advantage of higher signal at the higher field strength as well as with higher slew rate, allowing for higher resolution with less in-plane acceleration. This enables high SNR diffusion protocols with impressive spatial-angular resolution tradeoffs: 0.9 mm, 1.25 mm, and 1.6 mm isotropic resolutions with b-values 3000, 10,000, and 40,000 s/mm2, respectively. Utilization of such diffusion imaging  is aimed towards  improved cortical mesoscale tractography/connectomics as well as more accurate and innovative cytoarchitecture and microstructural biomarkers.

To be updated...

Large bore, high field magnets are used in searches for a postulated type of dark matter called the axion. The axion model was originally proposed to solve the long-standing puzzle of the vanishing neutron electric dipole moment, but was soon afterwards realized to also provide a compelling explanation for the charge-neutral dark matter whose presence in the universe can be inferred from various astronomical and cosmological observations. Many ongoing experiments attempt to directly detect background dark matter fields gravitationally clumped in our galaxy, and hence permeating our laboratories. These utilize cryogenic, magnetically-infused targets to transduce invisible axion waves into visible photon waves via a 3-wave mixing interaction known as the topological magneto-electric effect.  The very weak mixing would produce occasional single photon signals which can be detected with modern quantum sensing techniques, using e.g. superconducting qubits.