Methods for prediction and early detection of neurological events
Inventors
Truccolo, Wilson • Hochberg, Leigh R. • Donoghue, John P. • Cash, Sydney S.
Assignees
General Hospital Corp • Brown University • US Department of Veterans Affairs
Publication Number
US-10448877-B2
Publication Date
2019-10-22
Expiration Date
2031-12-05
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
Several methods for prediction and detection of neurological events are proposed based on spatiotemporal patterns in recorded neural signals. The methods are illustrated with examples from neural data recorded from human and non-human primates.
Core Innovation
The invention provides several methods for prediction and early detection of neurological events based on analyzing spatiotemporal patterns in recorded neural signals, specifically focusing on single-neuron action potentials (spiking) and neural ensemble activity. The methods involve recording continuous electrical signals from single neurons or other brain cells, measuring their spiking activity, characterizing this activity as neural point process sample paths, and estimating probability distributions of these sample paths over specified time intervals. By comparing current spiking activity to previously observed activity distributions, the methods predict or detect the occurrence of neurological events, such as epileptic seizures.
The methods further include estimating conditional intensity function models for spike trains of recorded neurons or cells, calculating probabilities of neuron spiking based on these models, and deriving receiver operating characteristic (ROC) curves to measure relative predictive power for neurological event prediction. The approach extends to constructing graphical connectivity models from estimated ensemble temporal filters and analyzing spectral coherence between spike trains and local field potentials to aid in event prediction or detection.
The problem being addressed arises from the difficulty in predicting, detecting, and understanding neurological events such as epilepsy seizures, which remain significant medical concerns. Traditional EEG-based methods provide averaged signals that obscure the detailed, heterogeneous activity of single neurons during seizures. Existing human studies have been limited in number and scope, often examining only few neurons or brain regions. There is a need for improved methods that can capture collective dynamics and heterogeneity in neuronal spiking, which may lead to better prediction, diagnosis, and treatment of neurological disorders.
Claims Coverage
The patent includes multiple independent claims focusing on methods for predicting, detecting, and treating neurological events, particularly epilepsy, based on neural signal analysis. The claims cover inventive features involving sample path distributions, conditional intensity function models, graphical connectivity models, and spike-field spectral coherence.
Predicting and detecting neurological events using neural point process sample paths
The method records electrical signals from single neurons or cells, measures spiking activity represented as sample paths, estimates probability distributions of these sample paths over specified past time intervals, and determines whether current activity falls outside confidence intervals of these distributions to predict or detect neurological events. It includes weighting neurons based on importance for prediction and using overlapping or non-overlapping sample paths.
Calculating relative predictive power for event prediction from conditional intensity function models
The method fits conditional intensity function models to spike trains of recorded neurons or cells over specific time intervals, computes spiking probabilities, derives receiver operating characteristic curves, calculates relative predictive power (RPP), and predicts or detects neurological events by comparing RPP values to confidence intervals from prior data, including weighting neuron contributions and using cross-validation.
Predicting neurological events using graphical models derived from conditional intensity functions
The method derives a graphical neuronal network connectivity model from estimated ensemble temporal filters for recorded neurons or cells during a current time interval, extracts a parameter such as graph density from the graphical model, compares this parameter to a probability distribution from prior time windows, and predicts or detects neurological events based on this comparison.
Predicting neurological events based on spike-field spectral coherence analysis
The method measures spiking activity and local field potentials, estimates pairwise spike-field spectral coherence at given frequencies across recorded neuron and field potential pairs, computes probability distributions of coherence during current and prior time intervals, and predicts or detects neurological events by statistically comparing these distributions, including use of tests such as the two-sample Kolmogorov-Smirnov test.
The inventive features collectively provide comprehensive methods for prediction, detection, and treatment of neurological events by analyzing neural spiking activities through statistical sample paths, conditional intensity models, graphical connectivity, and spike-field coherence, employing confidence interval-based comparisons and weighting schemes.
Stated Advantages
The methods enable early warning to patients and medical personnel by predicting neurological events such as epileptic seizures.
They offer improved prediction and detection accuracy by analyzing single neuron spiking activity and ensemble collective dynamics rather than relying solely on macroscopic EEG signals.
The invention supports diagnosis, prognosis, and treatment guidance of neurological disorders based on detailed neuronal activity patterns.
Documented Applications
Prediction and detection of epileptic seizures.
Diagnosis and prognosis of epilepsy.
Detection of neuronal activity changes indicating disordered, diseased, or injured states such as epilepsy, encephalopathy, neural oligemia, or ischemia.
Monitoring or prognosis of spreading cortical dysfunction following traumatic brain injury.
Detection of incipient ischemia in cerebral vasospasm following subarachnoid hemorrhage.
Monitoring the depth of pharmacologically-induced anesthesia, sedation, or brain activity suppression.
Determining resolution of status epilepticus and emergence from pharmacology-induced burst-suppression behavior.
Severity assessment of metabolic encephalopathy in critical medical illness, including liver failure.
Interested in licensing this patent?