Realistic models for detection of neuronal currents with magnetic resonance imaging

Abstract

A great challenge facing brain research is to “see” neurons in action at high spatial and temporal resolution in the living human brain. While existing non-invasive techniques, such as electroencephalography (EEG), magnetoencephalography (MEG), and functional magnetic resonance imaging (fMRI), have either poor spatial or temporal resolution, neuronal current MRI (nc-MRI) may hold the potential to revolutionize cognitive neuroscience by imaging neuronal activity at high temporal and spatial resolutions. However, the implementation of nc-MRI using existing instrumentation is yet to be convincingly demonstrated. In this project, I investigated the feasibility of nc-MRI via computer simulations. To allow realistic neuronal current simulations, the laminar cortex model (LCM) was first developed. The LCM incorporates the laminar architecture of the cerebral cortex into a continuum cortex model (previously developed by Wright et al.) to simulate the collective activity of cortical neurons. As validations, the LCM has been used to simulate the local field potentials (LFP) of the primary visual cortex. The LCM produced spontaneous LFPs exhibited frequency-inverse (1/f) power spectrum behaviour. The LCM also captured the fundamental as well as the high order harmonics under intermittent light stimulation. To model neuronal currents, I decomposed the neuronal activity simulated by the LCM into action potentials and postsynaptic potentials. The geometries of dendrites and axons were generated dynamically to account for neuronal morphology diversity. Magnetic fields produced by action potentials and postsynaptic potentials were calculated for the cases of spontaneous and stimulated cortical activity, from which the nc-MRI signal was determined. The MRI signal magnitude change was found to be below currently detectable levels (< 0.1 part-per-million), but signal phase change was potentially detectable (in the order of 0.1 milli-radian). Furthermore, nc-MRI signals were sensitive to temporal and spatial variations in neuronal activity and independent of the intensity of neuronal activation. Synchronous neuronal activity produces large phase changes, up to 1 milli-radian, and the signal phase oscillated with neuronal activity. Based on the computer simulation results, I proposed to image oscillatory neuronal currents using a multi-echo spin echo (MESE) and a synchronised multi-echo gradient recalled echo (MEGRE) sequences. A MESE sequence can accumulate phase changes for multiple neuronal activity oscillation periods through applying radio-frequency (RF) excitation pulses at the times when neuronal magnetic fields change sign. MEGRE sequence could be used to extract neuronal current signal from noisy MRI signals, because neuronal current signal but not blood-oxygen-level dependent (BOLD) effect or noise, varies with neuronal oscillation. Because a MEGRE sequence is capable of acquiring MRI signals at a series of closely-spaced time points, the inherent oscillation of neuronal current signals may potentially be deduced from the temporal profile of the MRI signals. I performed MRI experiments to image neuronal currents in the visual cortex induced by intermittent light stimulation using the proposed sequences. Significant neuronal current signal was absent due to the limited signal-to-noise ratio achieved by the system. I concluded that new MRI hardware and software (sequences and image analysis methods) is required for capturing neuronal currents signal in the brain.