Surgery – Diagnostic testing – Detecting brain electric signal
Reexamination Certificate
2000-10-20
2004-01-13
Robinson, Daniel (Department: 3742)
Surgery
Diagnostic testing
Detecting brain electric signal
Reexamination Certificate
active
06678548
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention is in the field of medical devices to treat neurological disorders of the brain. More specifically, the invention is directed to a method and a partially or fully implanted apparatus for predicting and detecting epileptic seizure onsets within a unified multiresolution probabilistic framework, thereby 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, in order to deter the course of a seizure with only the minimally required intervention and associated side effects.
Second to stroke, epilepsy is the most common neurological disorder of the brain. It is characterized by recurrent seizures that significantly impair the quality of life of an estimated 1 to 2% of the world's population. Drugs are the most common form of treatment, but their efficacy is limited. Up to 30% of patients achieve no control of their seizures with drugs, and another 30% experience severe side effects that make it impossible to lead normal lives.
A personal device capable of warning and/or intervening therapeutically in response to imminent seizures would allow those with epilepsy to, at a minimum remove themselves from danger (e.g., stop driving a car), and in the best case become seizure-free, not even noticing times when they were about to have a seizure. Such a device would operate in a continuous-time closed control loop where therapy is responsive to measurements (this includes a patient's own actions in the loop).
Several prior art closed-loop responsive systems with applicability to improving the quality of life of epileptic individuals are known to have been proposed in the field to date. All prior art systems share the following disadvantages: (1) they only detect visually obvious changes in raw signals thus control of seizures is attempted only after individuals actually begin having each seizure; (2) they implement a deterministic approach which is inadequate in face of the uncertainty and complexity of the problem; (3) they offer no means of gauging confidence in the outputs; (4) they implicitly assume a single (infinite) time resolution which may be adequate for seizure detection but not prediction; (5) they suggest a control scheme which is closed-loop only at the triggering instant dictated by detection (treatment beyond that point is open-loop, and is called triggered open-loop control in the present invention); (6) they do not deliver therapy that is graded from benign to aggressive as the situation warrants; (7) they do not consider side effects; (8) they imply detection schemes that are not guided by optimality criteria; (9) they rely on a single input feature or multiple features of the same nature (e.g., power in frequency bands) or only few uncorrelated features; (10) they use the same features for the whole patient population and do not take advantage of patient-specific features; (11) they do not transfer adequate samples of data for offline analysis; (12) they possess little or no computational intelligence with no learning capabilities to automatically improve and maintain performance over time; (13) they directly threshold separate single features instead of an implicit likelihood ratio function of joint features thereby yielding suboptimal decision rules; and (14) they do not account for the fact that training and/or testing seizure detectors/predictors with wrong prior probabilities of seizures/preseizures (as reflected in raw data archives or clinical trials) induces a variety of distortions that must be corrected.
The present invention is directed to overcome the disadvantages and limitations of all prior art.
SUMMARY OF THE INVENTION
The invention is directed to a method and a partially or fully implanted apparatus for predicting and detecting epileptic seizure onsets within a unified multiresolution probabilistic framework, thereby 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, in order to prevent, abort, or mitigate the intensity, duration, frequency, and spread of seizures with only the minimally required intervention and associated side effects. Based on novel computational intelligence algorithms, a realistic posterior probability function P(S
T
|x) representing the probability of one or more seizures starting within the next T minutes, given observations x derived from intracranial EEG (IEEG) or other signals, is periodically synthesized for a plurality of prediction time horizons (scales T, or resolutions 1/T), e.g., a second, a minute, ten minutes, an hour, etc. When coupled with optimally determined thresholds for alarm or therapy activation, probabilities defined in this fashion 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. For example, whereas it is unrealistic to predict the exact onset time of a seizure as 9 minutes and 58 seconds from now, it is both realistic and useful to predict that the onset will occur anytime within the next 10 minutes, a time during which the seizure can be prevented using a benign form of treatment. The longer and shorter prediction time scales are made to correspond to benign and aggressive therapies respectively. In addition to providing degrees of confidence and fine monitoring of patients' states, probabilities can be advantageously treated as degrees of “imminence” of events. Such degrees in turn serve to modulate the dosage and other parameters of treatment during open-loop or feedback control of preseizures 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.
A distinguishing theme of the present invention is that prediction is achieved for most patients and circumstances well before electrographic onset of seizures, and before any changes in raw physiologic signals that are visually obvious to a human expert. These prediction windows afford sufficient time to discourage seizures starting with mild forms of treatment, and escalating into multitherapeutic regimes only as it becomes necessary. Therefore, a principal objective of the invention is to avert seizures in the brain using only the minimally required interventions and their attendant side effects.
The present invention exploits the synergy of multiple signal features of a different nature. Features are accessed from a rich feature library including instantaneous, historical, spatial, and artificial features. Patient-specific signal features are exploited. Conventional features are custom-searched, and artificial features are custom-made, for each patient and prediction horizon, optimizing prediction performance and computational requirements. The invention exploits the synergy of multiple time resolutions in parallel.
The invention displays probabilities of oncoming seizures, each associated with a prediction horizon/resolution, in order to indicate both the time frame when onsets are expected to occur, and degrees of confidence in the predictions.
The value of the probabilities can be deliberately influenced by using them as controlled variables in a hierarchical seizure controller consisting of multitherapy activation decision logic and triggered open-loop or feedback control laws/actuators.
Multitherapy activation decisions can be based on user-selectable classifier-based optimality criteria (e.g.,
Echauz Javier Ramón
Esteller Rosana
Litt Brian
Vachtsevanos George John
Robinson Daniel
The Trustees of the University of Pennsylvania
Womble Carlyle Sandridge & Rice PLLC
LandOfFree
Unified probabilistic framework for predicting and detecting... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Unified probabilistic framework for predicting and detecting..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Unified probabilistic framework for predicting and detecting... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3266940