Unified Probabilistic Framework For Predicting And Detecting Seizure Onsets In The Brain And Multitherapeutic Device
First Claim
1. A method for periodic learning to improve and maintain the performance of a device implanted in an individual subject to seizures to provide treatment therapies, comprising the steps of:
- marking a time of unequivocal electrographic onset (UEO) in each recorded seizure over a fixed period of time;
creating a plurality of sets of learning data based on the UEOs by clipping a plurality of intracranial electroencephalographic (IEEG) epochs immediately preceding seizures and labeling the clipped epochs as preseizure raw data;
clipping and labeling randomly chosen, non-overlapping IEEG data as nonpreseizure raw data;
generating a time series of each feature in a feature library from the preseizure and nonpreseizure raw data;
searching for an optimal feature vector in the feature library to minimize a classifier-based performance metric;
synthesizing a posterior probability estimator for the optimal feature vector; and
coupling an optimal therapy activation threshold to the probability estimator.
2 Assignments
0 Petitions
Accused Products
Abstract
A method and an apparatus for predicting and detecting epileptic seizure onsets within a unified multiresolution probabilistic framework, enabling a portion of the device to automatically deliver a progression of multiple therapies, ranging from benign to aggressive as the probabilities of seizure warrant. Based on novel computational intelligence algorithms, a realistic posterior probability function P(ST|x) representing the probability of one or more seizures starting within the next T minutes, given observations x derived from IEEG or other signals, is periodically synthesized for a plurality of prediction time horizons. When coupled with optimally determined thresholds for alarm or therapy activation, probabilities defined in this manner provide anticipatory time-localization of events in a synergistic logarithmic-like array of time resolutions, thus effectively circumventing the performance vs. prediction-horizon tradeoff of single-resolution systems. The longer and shorter prediction time scales are made to correspond to benign and aggressive therapies respectively. The imminence of seizure events serves to modulate the dosage and other parameters of treatment during open-loop or feedback control of seizures once activation is triggered. Fast seizure onset detection is unified within the framework as a degenerate form of prediction at the shortest, or even negative, time horizon. The device is required to learn in order to find the probabilistic prediction and control strategies that will increase the patient'"'"'s quality of life over time. A quality-of-life index (QOLI) is used as an overall guide in the optimization of patient-specific signal features, the multitherapy activation decision logic, and to document if patients are actually improving.
-
Citations
30 Claims
-
1. A method for periodic learning to improve and maintain the performance of a device implanted in an individual subject to seizures to provide treatment therapies, comprising the steps of:
-
marking a time of unequivocal electrographic onset (UEO) in each recorded seizure over a fixed period of time;
creating a plurality of sets of learning data based on the UEOs by clipping a plurality of intracranial electroencephalographic (IEEG) epochs immediately preceding seizures and labeling the clipped epochs as preseizure raw data;
clipping and labeling randomly chosen, non-overlapping IEEG data as nonpreseizure raw data;
generating a time series of each feature in a feature library from the preseizure and nonpreseizure raw data;
searching for an optimal feature vector in the feature library to minimize a classifier-based performance metric;
synthesizing a posterior probability estimator for the optimal feature vector; and
coupling an optimal therapy activation threshold to the probability estimator. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16)
-
-
17. A system for periodic learning to improve and maintain the performance of a device implanted in an individual subject to seizures in providing treatment therapies comprising:
-
a signal acquisition component to condition and digitize a plurality of intracranial electroencephalographic (IEEG) signals received from a transducer implanted in the individual;
a storage medium for storing the plurality of digitized IEEG signals as IEEG data, and for storing a library of feature vectors generated from the IEEG data;
a processor for executing a learning and training component including;
a module for marking a time of unequivocal electrographic onset (UEO) in each recorded seizure over a fixed period of time;
a module for creating learning sets of data based on the UEOs by clipping all the IEEG epochs immediately preceding seizures and labeling the clipped epochs as preseizure raw data;
a module for clipping and labeling randomly chosen, non-overlapping IEEG data as nonpreseizure raw data;
a module for generating a time series of each feature in a feature library from the preseizure and nonpreseizure raw data;
a module for searching for an optimal feature vector in the feature library to minimize a classifier-based performance metric;
a module for synthesizing a posterior probability estimator for the optimal feature vector; and
a module for coupling an optimal therapy activation threshold to the probability estimator. - View Dependent Claims (18, 19, 20, 21, 22, 23, 24, 25)
-
-
26. A computer readable medium containing instructions for causing a computer system to improve and maintain the performance of a device implanted in an individual subject to seizures to provide treatment therapies, the computer readable medium comprising:
-
program instructions that mark a time of unequivocal electrographic onset (UEO) in each recorded seizure over a fixed period of time;
program instructions that create a plurality of sets of learning data based on the UEOs by clipping a plurality of intracranial electroencephalographic (IEEG) epochs immediately preceding seizures and labeling the clipped epochs as preseizure raw data;
program instructions that clip and label randomly chosen, non-overlapping IEEG data as nonpreseizure raw data;
program instructions that generate a time series of each feature in a feature library from the preseizure and nonpreseizure raw data;
program instructions that search for an optimal feature vector in the feature library to minimize a classifier-based performance metric;
program instructions that synthesize a posterior probability estimator for the optimal feature vector; and
program instructions that couple an optimal therapy activation threshold to the probability estimator. - View Dependent Claims (27, 28, 29, 30)
-
Specification