Project Summary Non-invasive imaging of brain anatomy and function is essential for the study of the development and operation of the human brain. It provides clinicians with invaluable information on neurological conditions, both in terms of understanding mechanisms of neurological diseases in general as well as providing guidance for diagnostics and treatment planning of individual patients. Among all functional imaging modalities, magnetoencephalography (MEG) has the best combined spatiotemporal resolution, which makes it an excellent tool for neuroscience and neurology. To exploit the potentially good spatial resolution of MEG, one must solve the inverse problem, i.e., estimate the underlying neural currents from the spatially discretized measurement of the magnetic field. This task, which is non-unique in principle, is accomplished by fitting specific parametrized mathematical models to the acquired multi-channel data and determining a set of parameters that provides the best fit according to a particular optimization criterion. Consequently, these parameters translate to an estimate of the spatial structure of the neural current, which is used in the interpretation of brain function under various tasks and conditions. The spatial precision of MEG can be determined by considering the following question: What is the minimum distance between two nearby spatial concentrations of neural current that can be distinguished as two separate sources instead of one, perhaps extended, source? In principle, this task appears increasingly more difficult as the distance between the sources and the measurement sensors increases. The reason for the difficulty is two-fold: 1) the amplitude of the magnetic field decreases with distance and 2) the spatially complex features of the magnetic field decay with distance faster than the spatially smoother, less informative, features. In conventional inverse modeling, the second type of difficulty may cause distinct sources to become merged as one estimated source even in the hypothetical situation that the sensors have no noise at all. To improve fundamental resolution of MEG, we will utilize our extensive expertise in hierarchical decompositions of magnetic signals by which we can separate signal features corresponding to different levels of spatial complexity, represented as spatial frequencies. In Aim 1, we develop new frequency-dependent hierarchical basis functions applicable to on-scalp measurements as well, optimize the numerical stability of the decomposition of the corresponding frequency components, and develop methodology for frequency-specific inverse modeling that aims at improving spatial resolution with the help of high-frequency components. In Aim 2, we develop methodology for new sensor array design in order to maximize the detectability of a wider frequency spectrum than what is achievable with conventional MEG systems. We exploit the fact that new sensor technologies allow for flexible designs and ...