Magnetoencephalography source imaging for neurological functionality characterizations
Inventors
Huang, Ming-Xiong • Lee, Roland R.
Assignees
Office of General Counsel of VA • University of California San Diego UCSD
Publication Number
US-10433742-B2
Publication Date
2019-10-08
Expiration Date
2034-08-05
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
Methods, systems, and devices are disclosed for implementing magnetoencephalography (MEG) source imaging. In one aspect, a method includes determining a covariance matrix based on sensor signal data in the time domain or frequency domain, the sensor signal data representing magnetic-field signals emitted by a brain of a subject and detected by MEG sensors in a sensor array surrounding the brain, defining a source grid containing source locations within the brain that generate magnetic signals, the source locations having a particular resolution, in which a number of source locations is greater than a number of sensors in the sensor array, and generating a source value of signal power for each location in the source grid by fitting the selected sensor covariance matrix, in which the covariance matrix is time-independent based on time or frequency information of the sensor signal data.
Core Innovation
The invention provides methods, systems, and devices for implementing magnetoencephalography (MEG) source imaging to detect loci of neuronal injury and abnormal neuronal networks that conventional neuroimaging techniques cannot visualize. The approach includes determining a covariance matrix based on sensor signal data representing magnetic-field signals emitted by a brain and detected by MEG sensors, defining a source grid within the brain with a resolution where the number of source locations exceeds the number of sensors, and generating a source value of signal power for each source location by fitting the covariance matrix, where the covariance matrix is time-independent based on time or frequency information of the sensor data.
The problem being addressed stems from limitations in existing neuroimaging techniques such as X-ray, CT, MRI, and diffusion tensor imaging (DTI), which mainly detect blood products, calcification, and edema but have low sensitivity to axonal injuries and abnormal functional connectivity. These standard techniques have low diagnostic rates for neuronal disorders, for example, less than 10% positive findings in mild traumatic brain injury (TBI) patients. There is a need for more sensitive imaging modalities capable of detecting neuronal injuries and abnormal functional connectivity with high spatial and temporal resolution.
The disclosed technology includes a fast MEG source imaging technique based on an L1-minimum-norm solution named Fast-Vector-based Spatial-Temporal Analysis (Fast-VESTAL). This approach acquires time-independent signal-related spatial modes from the MEG sensor waveform signals, obtains spatial source images of the brain based on these modes, and determines source time-courses with millisecond temporal resolution using an inverse operator constructed from the spatial images. The Fast-VESTAL technique is complemented by an objective pre-whitening method to remove correlated noise such as brain noise, enhancing detection sensitivity and enabling high-resolution imaging even under poor signal-to-noise conditions. The technology allows localization of a large number of focal and distributed neuronal sources, accurate source time-course reconstruction, low signal leakage, and voxel-based whole brain imaging for functional connectivity analyses.
Claims Coverage
The claims include three independent claims covering a method, a system, and a computer program product for MEG source imaging with key inventive features related to covariance matrix-based imaging, source grid definition, and signal processing techniques.
Covariance matrix-based high-resolution MEG source imaging method
Determining a time-independent covariance matrix from sensor signal data representing brain magnetic-field signals detected by MEG sensors; defining a source grid within the brain containing more source locations than sensors, at a particular resolution; and generating source signal power values for each location by fitting the covariance matrix to the sensor data.
MEG source imaging system with spatial mode processing
A system including MEG sensors acquiring magnetic field signals and a processing unit that determines time-independent signal-related spatial modes from MEG sensor waveform signals, obtains spatial source images based on these modes, and determines source time-courses of the spatial images using an inverse operator. The system can objectively remove correlated noise and includes means to remove bias toward grid nodes and coordinate axes.
Computer program product for MEG source imaging
A tangible non-transitory storage medium with instructions for determining a time-independent covariance matrix from MEG sensor signals, defining a source grid with more source locations than sensors, and generating source power values by fitting the covariance matrix, analogous to the described method, thus enabling software-based implementation of the MEG source imaging approach.
The claims encompass a comprehensive MEG source imaging solution that leverages a time-independent covariance matrix for high-resolution imaging with source grids exceeding sensor counts, coupled with signal processing innovations for noise removal, bias correction, and source time-course reconstruction, implementable as method steps, system components, and computer program instructions.
Stated Advantages
The disclosed Fast-VESTAL technique can localize and resolve a large number of focal and distributed neuronal sources with any degree of temporal correlations.
It can obtain accurate source time-courses and functional connectivity under poor signal-to-noise ratio conditions, including negative decibel ranges.
Fast-VESTAL operates with substantially low signal leakage to brain areas without sources, improving source image quality.
The technique facilitates imaging registration and group analyses by providing voxel-based whole brain imaging of MEG signals.
Implementation of an objective pre-whitening method enhances removal of correlated brain noise for improved imaging sensitivity.
The method achieves low computational costs compared to previous techniques, enabling faster processing without compromising resolution or accuracy.
Documented Applications
Detecting loci of neuronal injury and abnormal neuronal networks in neurological and psychiatric disorders such as mild traumatic brain injury (TBI), post-traumatic stress disorder (PTSD), Alzheimer's Disease/dementia (AD), multiple sclerosis (MS), autism, and schizophrenia.
High-resolution MEG source imaging of resting-state brain activity for different neural frequency bands to characterize normal brain function and develop normative databases.
Analyzing human somatosensory responses, e.g., median-nerve evoked responses, to localize correlated brain sources and reconstruct source time-courses consistent with known electrophysiology.
Non-invasive, in vivo biomarker data acquisition from healthy and diseased brain tissues to enhance diagnosis and functional characterization in clinical neuroscience.
Providing comprehensive whole-brain MEG source amplitude images for functional connectivity analyses and clinical research involving brain functionality assessment.
Interested in licensing this patent?