Surgery – Diagnostic testing – Detecting brain electric signal
Reexamination Certificate
1999-06-10
2003-04-15
Lateef, Marvin M. (Department: 3737)
Surgery
Diagnostic testing
Detecting brain electric signal
C600S300000, C600S545000
Reexamination Certificate
active
06549804
ABSTRACT:
MICROFICHE APPENDIX
A microfiche appendix as Appendix 1 containing a source code of a computer program useful in accordance with the present invention is appended hereto as 1 sheet of microfiche containing 20 frames.
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to the field of neuroscience for analyzing signals representative of a subject's brain activity including signals indicative or predictive of epileptic seizures. More particularly, the invention concerns the automated analysis of brain activity signals to detect an activity state and transitions between states, and to detect precursors predictive of a change in the subject's activity state to a different state.
The invention is based on ideas and research in the fields of mathematics, neurology, statistics and engineering which enable the real-time analysis of biologic signals such as the electro-encephalogram (EEG) or electro-corticogram (ECoG), by the simultaneous execution of multiple methods. In the preferred embodiment, these signals are rapidly, accurately, and automatically analyzed in order to:
1) Detect and signal the occurrence of an epileptic seizure in real time (or contemporaneously with the arrival of the signal at the processor/device),
2) To predict behavioral changes typically associated with seizures,
3) To predict seizures by detecting precursors to the onset of the electrographic or clinical components of a seizure,
4) To detect and further analyze epileptiform discharges (spikes), and
5) To download the detection or prediction outputs to devices for warning, or therapeutic interventions or the storage of data.
2. Description of the Prior Art
Humans and animals have several normal states of behavior such as wakefulness and sleep, as well as multiple sub-states such as attentive wakefulness and REM sleep. Abnormal states include reversible states such as seizures and progressive states such as dementia.
Epilepsy, a disabling disease, affects 1-2% of the American and industrialized world's population, and up to 10% of people in under-developed countries. Electroencephalography is the single most important ancillary test in the investigation of this disease. EEG's are recorded continuously for hours to days in an increasing number of cases with unclear diagnosis or poor response to adequate medical treatment. The amount of EEG data for analysis is extremely large (e.g., 64 channels of data at 240 Hz gives 1.3 billion data points/24 hr or 2.6 Gigabytes/day) and consists of complex waveforms with infinite variations.
Visual analysis of these signals remains (exclusive of this invention) the “gold standard” but it is impracticable for continuous EEG interpretation as this is the most time-consuming part of any electrodiagnostic test and requires special training and skills which make this procedure expensive and thus of limited access and use. Valuable EEG data is often discarded unexamined. The length of recording is unnecessarily prolonged in a specially equipped hospital suite until patients have several seizures. If the patient is unaware of the seizures, a common occurrence, then a nurse or relative must observe and document the occurrence of these changes. As seizures are brief and previously considered unpredictable, the need for continuous observation becomes imperative, adding to the cost in a non-effective manner.
Present methods of seizure detection are not only expensive, but rely on poorly discriminating methods, increasing the review time and nursing assistance because of the large number of false positives, and increasing the length of hospitalization through the false negatives. Furthermore, these methods often “detect” the seizure well after its onset or even its end, when prevention or abatement of the seizure is not possible or irrelevant.
The inability to process data in real time has thwarted the scientific and clinical development of the fields of epilepsy and electroencephalography. Cardiology has developed into a clinical science largely based on the power of electrocardiography to analyze the heart's electrical activity in a rapid and accurate manner. This has resulted in pacemakers, implanted defibrilators, and other devices which have saved thousands of individuals from premature death. The comparison between cardiology/EKG and epilepsy/EEG must take into account the fact that the electrical brain signals are far more complex than those originating from the heart. This explains in large part the developmental lag between these two disciplines.
Electrical brain signals, because of their spatial and temporal characteristics such as non-stationarity, have resisted accurate real-time automatic manipulation. The prior art methods presently used to characterize these states are severely limited. For example, the prior art consists of a long history of failed attempts to identify changes in EEG during certain behavioral states or tasks and to discern epi-phenomenology from phenomenology, a distinction that would help answer questions of fundamental importance. Other limitations include the inability to determine whether spikes are a static marker of epilepsy, or whether they are dynamically related to seizure generation.
Present methods of automatic EEG analysis have many major limitations which render them virtually useless for widespread, safe and effective clinical applications. These limitations include:
1) Lack of speed. The time it takes most methods to analyze the input signals and produce an output which detects or predicts a state change is too great for use in warning, intervention, or prevention of epileptic seizures and other abnormal brain states.
2) Lack of accuracy. Prior art methods have a large number of false positives (incorrectly identifying non-seizure activity as a seizure) and false negatives (failure to identify a true seizure), increasing the technical and financial burden.
3) Lack of adaptability to subject or seizure type; no compromise between speed vs. accuracy.
4) Lack of portability and implantability.
5) High cost.
Accurate and reproducible prediction of behavioral or biologic signal changes associated with seizures has not been possible as these events occur unpredictably. Our methods and devices enable seizure prediction by providing a worthwhile prediction time that makes warning, abortion/abatement, and prevention of seizures possible. The new treatment modalities that can be based on this method will lead to a significant reduction in seizure frequency and, consequently, to a reduction in the occurrence of injuries and fatalities, allowing persons with epilepsy to become productive and lead normal lives.
The prior art in automated seizure and spike detection consists of variations of two primary methods: “rule-based” analysis and, more recently, analysis by artificial neural networks. The most popular is a “rule-based” method which has been under development since the late 1970's, primarily by Dr. Jean Gotman. In the Gotman method, the signal is initially replaced by a piecewise linear approximation which connects maxima and minima.
In the Gotman method, there is a list of rules which are then applied to throw out some of the smaller line segments in an attempt to remove fast activity that is superimposed on an underlying wave of interest. The larger line segments which remain are called “half waves.” Gotman's algorithm then compares properties of the half waves such as averages of amplitude, duration, rhythmicity, and sharpness in moving ⅓ sec. windows to those of past and future data segments. As currently implemented, the method uses a total of 30 seconds of past data and 8-10 seconds of future data in these comparisons. A set of rules and thresholds are given to determine when these comparisons of past, present, and future properties yield a detection of a spike or seizure.
These rule-based methods have a number of limitations, including a large false positive rate, and usually a long delay to detect even abrupt changes (often 10 or more seconds).
Another method for
Dorfmeister Josef
Frei Mark
Lerner David
Osorio Ivan
Ralston John
Lateef Marvin M.
Mantis Mercader Eleni
Schoonover Donald R.
University of Kansas
LandOfFree
System for the prediction, rapid detection, warning,... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System for the prediction, rapid detection, warning,..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System for the prediction, rapid detection, warning,... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3030595