Systems and methods for detection of neurophysiological signal oscillations
Inventors
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
Systems and methods for detecting oscillations in neural signals are disclosed that provide for high precision and specificity in detecting neural oscillations in time and frequency domains. The disclosed systems and methods identify oscillations according to criteria including 1/f noise, number of cycles, and auto-correlation. The disclosed method detects periodic signals and filters out spurious oscillations associated with harmonic frequencies.
Core Innovation
A computer-implemented method identifies oscillations within electromagnetic activity of a brain of a patient by receiving a plurality of EM measurements indicative of EM activity and corresponding times, and transforming the plurality of EM measurements into a three-dimensional time-frequency space. The method removes aperiodic activity by performing a first filtering in the three-dimensional time-frequency space, where the first filtering removes EM measurements containing pink noise to generate a flattened three-dimensional time-frequency space by providing thresholds within the time-frequency space to identify periodic signals for each frequency point.
The method constructs a first plurality of bounding boxes that enclose a contiguous group of the EM measurements to capture onset and offset times of candidate oscillations within the flattened three-dimensional time-frequency space, where the contiguous group is detected where spectral power is above the thresholds. A second filtering rejects bounding boxes enclosing less than two candidate oscillation cycles to generate a second plurality of bounding boxes, with candidate oscillations detected based on detecting harmonic peaks in a corresponding bounding box.
An autocorrelation is performed within each bounding box of the second plurality of bounding boxes to identify a corresponding frequency by analyzing intervals of peaks and troughs. A third filtering retains only bounding boxes of the second plurality of bounding boxes containing candidate oscillations whose corresponding frequencies equal a center frequency of the candidate oscillations, where the center frequency is calculated as a center of frequencies of a plurality of EM signals in the corresponding bounding box. The method displays the candidate oscillations to a user and provides, based on the candidate oscillations, at least one of targeted neurofeedback to the patient, phase-locked neuromodulation to the patient, or therapeutic treatment to the patient.
Claims Coverage
The document includes three independent claims: a computer-implemented method, a system with processor-executable configuration, and non-transitory computer-readable media. Across the independent claims, the inventive coverage centers on a three-dimensional time-frequency transformation, pink-noise removal via thresholding to produce a flattened representation, bounding-box-based oscillation candidate selection with cycle and harmonic constraints, and autocorrelation-based frequency identification followed by center-frequency consistency filtering, leading to communication and provision of neurofeedback, phase-locked neuromodulation, or therapy.
Three-dimensional time-frequency transformation of EM measurements
Receiving a plurality of EM measurements indicative of EM activity of a brain of a patient and corresponding times; transforming the plurality of EM measurements into a three-dimensional time-frequency space.
Pink-noise removal to generate flattened time-frequency space
Removing aperiodic activity to perform a first filtering on the plurality of EM measurements in the three-dimensional time-frequency space, wherein the first filtering removes EM measurements containing pink noise to generate a flattened three-dimensional time-frequency space by providing thresholds within the time-frequency space to identify periodic signals for each frequency point.
Bounding-box construction to capture onset and offset of candidate oscillations
Constructing a first plurality of bounding boxes, each bounding box of the first plurality of bounding boxes enclosing a contiguous group of the EM measurements to capture onset and offset times of candidate oscillations within the flattened three-dimensional time-frequency space, wherein the contiguous group of the EM measurements are detected where spectral power is above the thresholds within the time-frequency space.
Cycle and harmonic-based bounding-box filtering with autocorrelation frequency matching
Performing a second filtering on the first plurality of bounding boxes to reject bounding boxes enclosing less than two candidate oscillation cycles to generate a second plurality of bounding boxes, wherein candidate oscillations are detected based on detecting harmonic peaks in a corresponding bounding box; performing an autocorrelation on the EM measurements within each bounding box of the second plurality of bounding boxes to identify a corresponding frequency by analyzing intervals of peaks and troughs; performing a third filtering on the second plurality of bounding boxes to retain only bounding boxes containing candidate oscillations with corresponding frequencies equal to a center frequency of the candidate oscillations to generate a third plurality of bounding boxes.
User communication and patient provision based on identified oscillations
Displaying the candidate oscillations in the third plurality of bounding boxes to a user; and providing, based on the candidate oscillations, at least one of targeted neurofeedback to the patient, phase-locked neuromodulation to the patient, or therapeutic treatment to the patient.
The independent claims collectively require transforming time-stamped brain EM measurements into a three-dimensional time-frequency space, removing pink-noise aperiodic activity using thresholds to obtain a flattened time-frequency space, detecting candidate oscillations as contiguous high-spectral-power regions via bounding boxes, rejecting candidates with fewer than two cycles using harmonic peaks, estimating frequency within each bounding box using autocorrelation, and retaining only candidates whose autocorrelation-derived frequency matches a center frequency. The resulting candidate oscillations are displayed and used to provide targeted neurofeedback, phase-locked neuromodulation, or therapeutic treatment.
Stated Advantages
Improved specificity vs prior methods.
Documented Applications
Sleep stage monitoring.
Depth of anesthesia monitoring.
Mental fatigue monitoring.
Stroke and Parkinson monitoring.
Brain-computer interface (BCI).
Alpha neurofeedback.
Neurofeedback and phase-locked electrical stimulation/therapeutic treatment.
Interested in licensing this patent?