Correlating intentional and unintentional changes in brain states with brain signals

Abstract

To provide methods of correlating a brain signal to intentional and unintentional changes in a brain state.SOLUTION: Disclosed herein are methods of analysis to extract and assess brain data collected from subject animals, including humans, to detect intentional and unintentional brain activity and other unexpected signals. These signals are correlated with higher cognitive brain functions or unintended, potentially adverse events, such as a stroke or seizure, and with translation of those signals into defined trigger events or tasks.SELECTED DRAWING: Figure 6

Description

[Technical field]

[0001] cross reference This application claims priority to U.S. Provisional Patent Application No. 61 / 590,235, filed Jan. 24, 2012, and U.S. Patent Application No. 13 / 749,619, filed Jan. 24, 2013, the contents of which are incorporated herein in their entireties.

[0002] FIELD OF THEINVENTION The present invention relates to analytical methods for extracting and evaluating brain data collected from animal subjects, including humans, to detect intentional and unintentional and other unexpected brain signals that are associated with higher brain cognitive functions or unintentional potential adverse events, such as stroke or seizure, and to convert these signals into defined trigger events or tasks. More specifically, the present invention relates to obtaining physiological data from EEG, EMG, EOG, MEG, ECoG, iEEG, fMRI, LFP, or other signals obtained from peripheral channels that are modulated by or modulate the subject's brain activity. [Background technology]

[0003] 2. Background of the Invention An electroencephalogram (EEG) is a tool used to measure electrical activity produced by the brain. Functional brain activity is collected by electrodes placed on the scalp. EEG has traditionally provided important information about the functioning of a patient's brain. Scalp EEG is thought to measure the set of currents present at postsynapses in the extracellular space, resulting from the flow of ions from and to dendrites that are connected by neurotransmitters. Thus, EEG and similar modalities are primarily used in neurology as a diagnostic tool for epilepsy, although the technique can be used in the study of other pathological conditions, including sleep disorders.

[0004] Recent advances in EEG and other signal detection have enabled real-time automated detection of sleep and wake states through normalization and other manipulation of brain activity data. Furthermore, such applications and methods can be used to automatically access disease states and the effects of drug treatments. Related technology has enabled real-time access to such data using single channel detectors. This in turn has provided the opportunity to further probe sleep and wake states, including clarifying the distinction between REM and deep sleep. To aid in the efficient collection of such data, head-and-harness systems have been developed that use single channel wireless data transmission. See, e.g., International Application No. 2006 / 018120; International Application No. 2009 / 064632; International Application No. 2010 / 054346; U.S. Patent Application No. 8,073,574; and Low, Philip Steven (2007). "A new way to look at sleep: separation and convergence". Published Thesis, University of California San Diego Electronic Theses and Dissertations (Identified: b6635681). To date, this technology has been applied primarily to sleep-related diagnostic applications and the effects of pathological conditions and drug treatments.

[0005] Development continues in the area of ​​exoskeletons and related prosthetics that hold the promise of enabling paraplegic patients to walk again and perform other tasks that they cannot currently perform. Furthermore, such devices may be useful for able-bodied individuals such as soldiers in the field, first responders, and construction workers. Companies such as Esko Bionics, Parker Hannifan, and Argo Medical Technologies are continually advancing such technology. See, for example, U.S. Pat. No. 8,096,965, International Publication No. 2010101595, and U.S. Patent Application No. 11 / 600,291, filed Nov. 15, 2006. Additionally, other devices are utilized to enable individuals severely impaired by ALS, MS, and the like to communicate using voice synthesizers and the like. Such devices are typically activated by movement of the buccinator muscles, eye movement using an eye tracker, and the like.

[0006] Thus, there is a need for non-invasive methods to detect intentional and unintentional communication from subjects, including individuals with disabilities, in order to assess and potentially respond to or prepare for the physiological implications of these changes in brain state. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] International Application No. 2006 / 018120 [Patent Document 2] International Application No. 2009 / 064632 [Patent Document 3] International Application No. 2010 / 054346 [Patent Document 4] U.S. Patent Application No. 8,073,574 [Patent Document 5] U.S. Patent No. 8,096,965 [Patent Document 6] International Application Publication No. 2010101595 [Patent Document 7] U.S. Patent Application Serial No. 11 / 600,291 [Non-patent literature]

[0008] [Non-Patent Document 1] Low, Philip Steven (2007). "A new way to look at sleep: separation and convergence". Published Thesis, University of California San Diego Electronic Theses and Dissertations (Identified: b6635681) Summary of the Invention

[0009] The present invention provides methods for non-invasive detection of intentional and unintentional communications from healthy and diseased subjects (including individuals impaired by neurological diseases such as ALS, MS, etc.) in the form of physiological data, for example, by EEG, EMG, EOG, MEG, ECoG, iEEG, fMRI, LFP, etc., and for correlating these intentional and unintentional signals with changes in brain states, including higher cognitive functions. There is also a need to utilize such intentional communications, for example, to simulate speech or to power prosthetic devices. There is also a need to access unintentional signals from subjects with pathological conditions, such as epilepsy (or physical illnesses that cause changes in brain activity), for correlation with pathological conditions and optionally for use in triggering alarms and / or intervening to alter, suppress, or prepare for unintentional events.

[0010] In a preferred method of the present invention, intentional brain signals from a subject are detected by attaching at least a single sensor to the subject, acquiring data indicative of brain activity, analyzing the data indicative of brain activity, and correlating the analyzed data with intentional higher-level cognitive functions from the subject. Preferably, the data is acquired non-invasively by applying at least a single sensor to the subject, more preferably by applying at least a single dry sensor or at least a single wet sensor. Furthermore, the data is preferably received from at least a single channel of EEG, EMG, EOG, MEG, ECoG, iEEG, fMRI, LFP, or a peripheral channel modulated by the subject's intention. In an alternative embodiment, the data is received through a multi-channel detector. Furthermore, the data is preferably communicated and received wirelessly.

[0011] In another preferred embodiment, the data is analyzed by normalizing a spectrogram of the data (including the normalized spectrogram) at least once in time across frequency, and normalizing a spectrogram of the same data (including the normalized spectrogram) at least once in frequency across time, where both normalizations may be performed in either order and may be repeated. In a further preferred embodiment, the data is analyzed by calculating a spectrogram of the data, normalizing the spectrogram, performing an independent or principal component analysis of the normalized spectrogram, and identifying clusters. Furthermore, the analyzing step may also include performing a temporal fragmentation analysis, a preferred frequency analysis, a repeated (preferably two or more) preferred frequency analysis, and / or a spectral fragmentation analysis.

[0012] In particularly preferred embodiments, the methods of the present invention further comprise transforming the analyzed data to perform tasks associated with higher cognitive functions including, but not limited to, intent, speech, recall, thinking, imagination, and planning (including but not limited to movement). Importantly, tasks resulting from transforming the analyzed data include simulating speech on a display, simulating speech with a speech synthesizer, or animating an artificial prosthesis or exoskeleton, or the like.

[0013] In yet another preferred embodiment of the method of the present invention, the brain signal from the subject is associated with at least one unintentional event by attaching at least one sensor to the subject, acquiring data indicative of electroencephalographic activity, analyzing the data indicative of brain activity, and associating the analyzed data with at least one unintentional event. In a further embodiment, after associating the data with the unintentional event, an alarm is activated. Alternatively, after associating the data with the unintentional event, a response may be initiated to improve the action of the unintentional event, which may include modifying, suppressing, or preparing for the unintentional event (and may be activating an alarm). This response may be particularly effective in the case of unintentional events, high fragmentation events, changes in fragmentation of events, startle, tremor, convulsion, injury, or pathological conditions, including but not limited to epileptic attacks, migraines, strokes, heart attacks, or infarctions.

[0014] In another preferred embodiment, the method of the present invention detects an intentional signal from a subject by attaching at least one detector capable of detecting the intentional signal to the subject, acquiring data indicative of the detected activity using EEG, EMG, EOG, MEG, ECG, ECoG, iEEG, LFP, fMRI, or peripheral channels modulated by the intentional signal from the subject, analyzing the data indicative of the detected activity, and correlating the analyzed data with an intentional higher-order cognitive function from the subject. It is contemplated that any method, system, or information described herein may be implemented with respect to any other method, system, or information described herein.

[0015] Unless otherwise defined, all terms herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs.Methods and materials are described herein for use in the present invention, and other suitable methods and materials known in the art can also be used.The materials and methods, as well as the examples, are merely illustrative and are not intended to be limiting.All publications, patent applications, patents, and other references mentioned herein are incorporated by reference in their entirety.In case of conflict, the present specification, including definitions, will take precedence.

[0016] These and other embodiments of the present invention will be better appreciated and understood when considered in conjunction with the following description and the accompanying drawings. It should be understood, however, that the following description, while indicating various embodiments of the present invention and many specific details thereof, is by way of illustration and not limitation. Many substitutions, modifications, additions, and / or alterations may be made within the scope of the present invention without departing from the spirit thereof, and the invention includes all such substitutions, modifications, additions, and / or alterations. [The present invention 1001] 1. A method for detecting intentional brain signals from a subject, comprising: (a) attaching at least one sensor to the object; (b) obtaining data indicative of brain activity; (c) analyzing the data indicative of brain activity; and (d) correlating the analyzed data with an intentional higher-level cognitive function from the subject. [The present invention 1002] The method of claim 1001, wherein the acquired data is received non-invasively by applying the at least one sensor to the subject. [The present invention 1003] The method of claim 1002, wherein said at least one sensor is selected from the group consisting of a single dry sensor or a single wet sensor. [The present invention 1004] The method of the present invention 1002, wherein the acquired data is received from at least a single channel of EEG, EMG, EOG, MEG, ECoG, iEEG, fMRI, LFP, or a peripheral channel modulated by the subject's intent. [The present invention 1005] The method of the present invention 1004, wherein the acquisition data is received from at least a single channel of an EEG. [The present invention 1006] The method of the present invention 1005, wherein the acquired data is received from a multi-channel detector including at least a single channel of EEG, EMG, EOG, MEG, ECoG, iEEG, fMRI, LFP, or a peripheral channel modulated by the subject's intent. [The present invention 1007] The method of the present invention 1002, wherein the acquired data is received wirelessly. [The present invention 1008] The analyzing step comprises: (a) normalizing a spectrogram of the data at least once in time across frequency, the spectrogram including a normalized spectrogram; and (b) normalizing spectrograms of the same data at least once in frequency over time, including the normalized spectrograms. The method of the present invention 1001, comprising: [The present invention 1009] The analyzing step comprises: (a) calculating said spectrogram of said data; (b) normalizing the spectrogram one or more times; and (c) performing principal component analysis and / or independent component analysis of the normalized spectrogram. The method of the present invention 1008, comprising: [The present invention 1010] 1001. The method of claim 1001, wherein said analyzing step comprises performing a temporal fragmentation analysis. [The present invention 1011] 1001. A method according to claim 1001, wherein said analyzing step comprises performing a preferred frequency analysis. [The present invention 1012] 1012. A method according to claim 1011, wherein the analysis of preferred frequencies is performed on a spectrogram that has been normalized at least twice. [The present invention 1013] 1001. The method of claim 1001, wherein the analyzing step includes performing a spectral fragmentation analysis. [The present invention 1014] The method of claim 1001, further comprising, after the correlating step, transforming the analyzed data to perform a task associated with a higher cognitive function. [The present invention 1015] The method of claim 1001, wherein said higher cognitive function is selected from the group consisting of intention, speech, recollection, planned movement, thinking, and imagination. [The present invention 1016] 1012. The method of claim 1011, wherein the task is selected from the group consisting of simulating speech on a display, simulating speech with a speech synthesizer, and moving an artificial prosthesis. [The present invention 1017] 1. A method for detecting a brain signal associated with at least one unintentional event from a subject, comprising: (a) attaching at least one sensor to the object; (b) obtaining data indicative of electroencephalographic activity; (c) analyzing the data indicative of brain activity; and (d) correlating the analyzed data with at least one unintended event. [The present invention 1018] The method of claim 1017 further comprising the step of activating an alarm after said associating step. [The present invention 1019] The method of claim 1017, further comprising, after said associating step, a step of responding to said unintended event to ameliorate its effect. [The present invention 1020] The method of claim 1019, wherein said response to said unintended event comprises intervening to alter, inhibit, or prepare for said unintended event. [The present invention 1021] The method of the present invention 1017, wherein the unintended event is selected from the group consisting of a high fragmentation event, a change in fragmentation of an event, startle, tremor, convulsion, injury, or a pathological condition including, but not limited to, an epileptic seizure, migraine, stroke, heart attack, or infarction. [The present invention 1022] The method of the present invention 1017, wherein the response to the unintentional event enhances or replaces a polygraph test. [The present invention 1023] 1017. The method of claim 1017, wherein said analyzing step comprises performing a temporal fragmentation analysis. [The present invention 1024] 1017. A method according to claim 1017, wherein said analyzing step comprises performing a preferred frequency analysis. [The present invention 1025] A method according to claim 1024, wherein the analysis of preferred frequencies is performed on a spectrogram that has been normalized at least twice. [The present invention 1026] 1017. A method according to claim 1017, wherein said analyzing step comprises performing a spectral fragmentation analysis. [The present invention 1027] The analyzing step comprises: (a) normalizing a spectrogram of the data at least once in time across frequency, the spectrogram including a normalized spectrogram; and (b) normalizing spectrograms of the same data at least once in frequency over time, including the normalized spectrograms. The method of the present invention 1017, comprising: [The present invention 1028] The analyzing step comprises: (a) calculating said spectrogram of said data; (b) normalizing the spectrogram one or more times; and (c) performing principal component analysis and / or independent component analysis of the normalized spectrogram. The method of the present invention 1027, comprising: [The present invention 1029] 1. A method for detecting an intentional signal from a subject, comprising: (a) attaching to the subject at least one detector capable of detecting an intentional signal; (b) acquiring data indicative of the detected activity using EEG, EMG, EOG, MEG, ECG, ECoG, iEEG, LFP, fMRI, or peripheral channels modulated by the subject's intent; (c) analyzing the data indicative of the detected activity; and (d) correlating the analyzed data with an intentional higher-level cognitive function from the subject. [Brief description of the drawings]

[0017] So that the invention may be clearly understood and readily practiced, the invention is described in conjunction with the following drawings, in which like reference characters indicate the same or similar elements, and in which:

[0018] The patent or application file contains at least one drawing executed in color. Copies of this patent or patent application publication containing color drawing(s) will be provided by the Office upon request and payment of the necessary fee.

[0019] [Figure 1-1] 1 is a series of graphs showing examples of brain signal analysis, with the left column showing a conventional analysis and the right column showing an analysis using the method of the present invention for closing the eyes (A), clenching the left hand (B), clenching the right hand (C), clenching the left foot (D), and clenching the right foot (E). The last two graphs (F) show the conventional analysis in the left column and the analysis using the method of the present invention in the right column, in a subject in a resting state. [Figure 1-2] This is a graph showing a continuation of Figure 1-1. [Diagram 2] A-E are data for the same task shown in Fig. 1A-E (without the resting state in Fig. 1F) analyzed by the method of the present invention. The green line indicates the timing of the verbal cue to start the task for 2 or 4 seconds, and the red line indicates the time of the verbal cue to stop the task and relax for 10 seconds. [Diagram 3] A-E are data for the same task shown in Fig. 1A-E (without the resting state in Fig. 1F) analyzed using temporal fragmentation using the methods of the present invention. The green lines indicate the timing of the verbal cue to start the task for 2 or 4 seconds, and the red lines indicate the time of the verbal cue to stop the task and relax for 10 seconds. [Figure 4] A-E are data for the same tasks shown in Fig. 1A-E analyzed with one or more normalizations (without the resting state in Fig. 1F). Specifically, the graphs plot summed high frequency power in the gamma and very high gamma (hgamma) ranges. The green lines indicate the timing of the verbal cue to start the task for 2 or 4 seconds, and the red lines indicate the time of the verbal cue to stop the task and relax for 10 seconds. [Diagram 5]A-E are data for the same tasks shown in Fig. 1A-E analyzed using one or more normalizations (no resting state in Fig. 1F). Specifically, the graphs plot sum of alpha frequency power against sum of gamma frequency power for Fig. 5A-D, whereas delta frequency was used for Fig. 5F (eyes closed). The green lines indicate the timing of the verbal cue to start the task for 2 or 4 seconds, and the red lines indicate the time of the verbal cue to stop the task and relax for 10 seconds. [Figure 6] 4 is a flow chart illustrating the application of the method of the present invention to an intentional signal. [Figure 7] 1 is a flow chart illustrating the application of the method of the present invention to unintended signals, for example in epilepsy. [Figure 8] Data are analyzed using one or more normalizations for gamma frequency power (A), alpha frequency power (B), the sum of alpha and gamma frequency power (C), and the ratio of the sum of alpha and very high gamma frequency power, while alternating between visualizing kicking a football and looking at the bedroom. The green line indicates the timing of the verbal cue to start and the red line indicates the time of the verbal cue to stop for a 10 second interval. [Figure 9] FIG. 9 is a screenshot of a computer interface for using the method of the present invention. [Figure 10] Two-sample Kolmogorov-Smirnov (KS) tests of cross-axially repeated 10-fold normalized preferred frequency spectrograms from each task listed in Figure 1A-E performed by an immobile ALS subject show that multiple intentional events can be discriminated. Each spectrogram comes from a distribution distinct from that of any other task (though to a lesser extent for this particular test in this subject: left hand grasping and right foot grasping). DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0020] Detailed Description It will be understood that the drawings and description of the invention have been simplified for purposes of clarity to show elements relevant to a clear understanding of the invention, excluding other elements that are well known. A detailed description is provided herein below with reference to the accompanying drawings.

[0021] A detailed description of one or more embodiments of the present invention is provided below along with the accompanying drawings illustrating the principles of the present invention. Although the present invention will be described in connection with such embodiments, the present invention is not limited to any embodiment. The scope of the present invention is limited only by the claims, and the present invention encompasses numerous alternatives, modifications, and equivalents. Numerous specific details are set forth in the following specification to provide a thorough understanding of the present invention. These details are provided for illustrative purposes, and the present invention may be practiced according to the claims without some or all of these specific details. For purposes of clarity, technical matters known in the technical field related to the present invention have not been described in detail so as not to unnecessarily obscure the present invention.

[0022] In this application, the term "subject" refers to both animals and humans. Referring now to FIG. 6, a flow chart 100 is disclosed illustrating a preferred embodiment of the method. Specifically, the subject imagines and generates an intentional brain signal that is received by at least one sensor 101 according to the present invention. Preferably, the sensor includes a single wet electrode or a single dry electrode. The input signal is relayed to a computing device, such as a computer, where it is analyzed (102) to determine if a defined unintentional event has occurred (103). Preferably, this is indicative of brain activity from the subject that is associated with higher cognitive functions. The computer then generates a signal (104) that is relayed to one or more receivers 105 in a peripheral device. Examples of peripheral devices may include speech synthesizers, prosthetic devices including exoskeletons, and the like. As shown in FIG. 6, these methods are preferably performed in real time and provide a continuous response and feedback look between the subject's intentional signals and the conversion of these signals into commands associated with the peripheral device controlled by the subject's intentions. In a particularly preferred embodiment, data signals from the subject are received from at least a single channel of EEG, EMG, EOG, MEG, ECoG, iEEG, fMRI, LFP, or a peripheral channel controlled by the subject's intent. In a preferred alternative, data is received through a multi-channel detector that wirelessly interfaces with a computing device and a peripheral device.

[0023] Referring now to FIG. 7, another flow chart 106 is disclosed illustrating an alternative preferred embodiment of the method. Specifically, a subject is monitored for the presence of an unintended event (e.g., stroke or seizure) via brain signals received by at least one sensor 107 according to the present invention. Preferably, the sensor includes a single wet electrode or a single dry electrode. The input signal is relayed to a computing device, such as a computer, where it is analyzed (108) to determine whether a defined unintended event has occurred (109). There are many examples of unintended potentially adverse events, including but not limited to startles, tremors, convulsions, injuries, or pathological conditions, including but not limited to epileptic seizures, migraines, strokes, heart attacks, or infarctions.

[0024] The computer then generates a signal that is relayed to one or more receivers of the peripheral device (110). In this example, the peripheral device is a stimulator 111 used to improve the effects of dangerous seizures in epilepsy patients by stimulating a brain region 112 of the subject. As will be appreciated by those skilled in the art, the peripheral device 111 may also inhibit or not alter (or affect) the subject 112, for example, activating an alarm to alert the subject or a caregiver. In a particularly preferred embodiment, the data signal from the subject is received from at least a single channel of EEG, EMG, EOG, MEG, ECoG, iEEG, fMRI, LFP, or a peripheral channel that is modulated by the unintentional occurrence of the monitored unintentional event. In a preferred alternative, the data is received through a multi-channel detector that wirelessly connects with the computing device and the peripheral device.

[0025] In another preferred embodiment of the method shown in both Figures 6 and 7, the data is analyzed by normalizing a spectrogram of the data (including the normalized spectrogram) at least once in time across frequency, and normalizing a spectrogram of the same data (including the normalized spectrogram) at least once in frequency across time, where both normalizations may be performed in either order or repeated. In a further preferred embodiment, the data is analyzed by calculating a spectrogram of the data, normalizing the spectrogram, performing an independent or principal component analysis of the normalized spectrogram, and identifying clusters. Furthermore, the analysis may also include performing a temporal fragmentation analysis, a preferred frequency analysis, a repeated (preferably more than once) preferred frequency analysis, and / or a spectral fragmentation analysis.

[0026] In a preferred embodiment, it is disclosed to detect brain signals (e.g., high frequency signals detected from the brain) and to correlate the detected signals with higher cognitive functions. Examples of higher cognitive functions include, among others, directed tasks that involve intention, speech, recall, thinking, imagination, and planning (including but not limited to movement), as well as imagination and other cognitive processing and functions.

[0027] In some embodiments, the subject's brain signals are converted into speech. For example, the subject imagines one or more language elements, including but not limited to letters, numbers, words, or symbols, and brain signals associated with the language element or elements are detected to determine the language element or elements. In another example, a speech-impaired subject is taught to associate an image, symbol, word, letter, or number with an imagined action. The subject imagines an imagined action, and brain signals associated with the imagined action are detected to determine the word, letter, or number associated with the imagined action. In one embodiment, the subject's imagined action controls a cursor on a display that selects the image, symbol, word, letter, or number that the subject intends to use. In some embodiments, the determined image, symbol, word, letter, or number is used with one or more other determined images, symbols, words, letters, or numbers to form a grammatically correct speech. In some embodiments, the determined image, symbol, word, letter, or number is communicated and / or displayed using a speech synthesizer. In some embodiments, the brain signal and / or output is assigned a non-verbal value, including but not limited to a tone, a series of tones, a micro-pitch, a color, an image, an electrical stimulus, a one-dimensional or multi-dimensional graphic. In some embodiments, the brain signal and / or output is assigned to a brain signal. In some embodiments, the brain signal follows and / or precedes and / or occurs concurrently with one or more intrinsic and / or exogenous events and / or conditions (including but not limited to pathological and / or abnormal events and / or conditions).

[0028] In some embodiments, the subject's brain signals are converted into movements of the prosthesis, for example, detected brain signals associated with imagined movements are used to at least partially control the prosthesis.

[0029] In some embodiments, one or more physiological recordings, including but not limited to EEG, EMG, EOG, MEG, ECoG, iEEG, fMRI, LFP, or peripheral channels modulated by the subject's intention, or reading unintentional signals related to unintentional events, are used to detect one or more brain signals. For example, a single channel iBrain® EEG recording is made on a high-functioning 70-year-old ALS patient attempting to move one of his limbs after a verbal cue (left and right hand and foot). The EEG signal is analyzed with algorithms including the SPEARS algorithm to make the brain signal detectable. A simultaneous video recording may be acquired. During the movement attempt, the subject's brain activity shows broad spectrum pulses ranging from gamma and very high gamma ranges. Such pulses are present even in the absence of actual movement and are absent when the subject is not attempting a movement. Activity in the alpha range is detected when the subject closes his eyes. Such high-bandwidth biomarkers open the possibility of linking intentional movements to a library of words and translating them into speech, providing ALS patients with a communication tool that utilizes brain signals.

[0030] In some embodiments, clear and broad spectrum patterns of activity across many frequencies can be detected for real or imagined movements compared to different patterns for resting states. These patterns coincide with the timing of the real or imagined movements of the subject. Traditional spectral analysis does not reveal such patterns. Analysis of physiological data can be used to detect brain signals that are timed and associated with the real or imagined movements of the subject. In some embodiments, these signals are transmitted and analyzed in real time to provide additional freedom for brain-based communication.

[0031] The methods described herein are disclosed in detail in International Application No. 2006 / 018120; International Application No. 2009 / 064632; International Application No. 2010 / 054346; U.S. Patent Application No. 8,073,574; and Low, Philip Steven (2007). "A new way to look at sleep: separation and convergence". Published Thesis, University of California San Diego Electronic Theses and Dissertations (Identified: b6635681), the disclosures of which are incorporated herein by reference in their entireties.

[0032] The present invention utilizes systems and methods for acquiring and classifying EEG data in both animals and humans. The acquired EEG signals are low power frequency signals and follow a 1 / f distribution, such that the power in the signal is inversely related, e.g., inversely proportional, to frequency.

[0033] EEG signals are often examined in successive periods of time called epochs. An epoch can be segmented into different sections using a scanning window that defines the time series units of the different sections. The scanning window can move through a sliding window, where the sections of the sliding window have overlapping time series sequences. Alternatively, an epoch can span the entire time series, for example.

[0034] In a preferred embodiment of the present invention, a single channel EEG was sufficient to obtain data indicative of intentional (or unintentional or otherwise unexpected) brain activity.

[0035] Typically, the source data acquired by the methods of the present invention are adjusted to increase the dynamic range of power in at least one low-power frequency range of the frequency spectrum of the source data compared to a second frequency range of higher power. Many adjustment techniques as described herein may be used, including normalization and frequency weighting.

[0036] In one embodiment, the electroencephalography source data is normalized to increase data with lower power in higher frequency ranges compared to data with higher power in lower frequency ranges, or more generally, to normalize the power in different parts of the signal.

[0037] After the source data is adjusted, various other processing may be performed. For example, a visualization of the adjusted source data may be presented. Additionally, low power frequency information may be extracted from the adjusted source data. For example, low power frequency information may be extracted from the adjusted electroencephalography source data. Also, high power frequency information may be extracted from the adjusted source data.

[0038] The method described in any of this or other examples may be a computer-implemented method performed via computer-executable instructions in one or more computer-readable media. Any action shown may be performed by software embedded in a signal processing system or any other signal data analysis system. For example, the present invention may be implemented in a variety of ways, including as a process, an apparatus; a system; an arrangement; a computer program product embodied in a computer-readable storage medium; and / or a processor (e.g., a processor configured to execute instructions stored in and / or provided by a memory connected to the processor). In this regard, FIG. 9 is a screenshot of a computer interface utilizing the method and output of the present invention. These embodiments or any other form the present invention may take may be referred to herein as techniques. In general, the order of steps of a disclosed process may be changed within the scope of the present invention. Unless otherwise stated, components such as a processor or memory described as configured to perform a task may be implemented as general components temporarily configured to perform the task at a given time, or as specific components manufactured to perform the task. As used herein, the term "processor" refers to one or more devices, circuits, and / or processing cores configured to process data, such as computer program instructions.

[0039] Another embodiment uses multiple normalizations to further increase the dynamic range. Normalization can be done by normalizing frequency over time or by normalizing time over frequency.

[0040] For example, electroencephalography data of at least one low-power frequency range may be received. Artifacts in the data may be removed from the source data. Artifact data may be removed from the source data, for example, manually or automatically via filtering (e.g., DC filtering) or data smoothing techniques. The source data may also be preprocessed by component analysis (e.g., principal component analysis or independent component analysis). The source data may be segmented into one or more epochs, where each epoch is a portion of data from a series of data. For example, the source data may be segmented into multiple time segments via various separation techniques. Scanning and sliding windows may be used to separate the source data in time series units. The one or more epochs are normalized in time for the difference in power in the one or more epochs. For example, the power in each epoch at one or more frequencies may be normalized in time to determine an appropriate frequency window for extracting information. Such normalization may reveal low-power, statistically significant power shifts at one or more frequencies (e.g., delta, gamma, alpha, etc.). Any frequency range may be identified and utilized for the analysis. After an appropriate frequency window is established, information is calculated for each of the one or more epochs. Such information may include low frequency power (e.g., delta power), high frequency power (e.g., gamma power), standard deviation, maximum amplitude (e.g., maximum absolute value of peaks), etc. Further calculations may be performed on the information calculated for each of the one or more epochs to generate information as gamma power / delta power, time derivative of delta, time derivative of gamma power / delta power, etc. The time derivative may be calculated across the preceding and subsequent epochs. After the information is calculated, it may then be normalized across the one or more epochs. Various data normalization techniques may be performed, including z-scoring and other similar techniques.

[0041] The results of adjusting the source data to account for differences in power across a spectrum of frequencies over time can be shown as one or more epochs of data, for example, frequency-weighted epochs can be shown as adjusted source data.

[0042] The electroencephalography data of a subject is acquired and input, and the data is segmented into one or more epochs. In practice, the epochs are of similar (e.g., the same) length. The length of the epochs can be adjusted via a configurable parameter. The one or more epochs are then input, and the frequency data in the one or more epochs is normalized across time, thereby frequency-weighting the electroencephalography data of the one or more epochs. The frequency-weighted one or more epochs are then input to a classifier, which classifies the data into a state of intention and a state of relaxation or non-intention.

[0043] For example, electroencephalography (EEG) data of a subject is received. For example, electroencephalography data may be received in which the dynamic range of power in at least one low-power first frequency range in a frequency spectrum is lower compared to a second frequency range in the frequency spectrum. The electroencephalography data of the subject is segmented into one or more epochs. For example, the EEG data may be segmented into one or more epochs through various separation techniques. Scanning windows and sliding windows may be used to separate the EEG data into one or more epochs. Also, the source data may be actively filtered with DC during, before, or after segmentation. Also, the source data is pre-processed by component analysis (e.g., principal component analysis or independent component analysis). In the EEG data for the entire night, higher frequencies (e.g., gamma) show lower power than lower frequencies (e.g., alpha, delta, theta, etc.) in the EEG data for the entire night. The frequency power of one or more epochs is weighted in the time direction. For example, the power of each epoch at one or more frequencies may be normalized in time to determine an appropriate frequency window for extracting information. Such normalization may reveal a statistically significant shift of low power at one or more frequencies (e.g., alpha, delta, gamma, etc.). Furthermore, each epoch may be represented by the frequency with the highest relative power over time to determine an appropriate frequency window for extracting information. Alternatively, component analysis (e.g., principal component analysis (PCA) or independent component analysis (ICA)) may be used after normalization to further determine an appropriate frequency window for extracting information. Any frequency range may be revealed and used for analysis.

[0044] After an appropriate frequency window is established (e.g., after frequency weighting), information may be calculated for each of the one or more epochs. Such information may include low frequency power (e.g., alpha power), high frequency power (e.g., gamma power), standard deviation, maximum amplitude (e.g., maximum absolute value of peaks), etc. Further calculations may be performed on the calculated information for each of the one or more epochs to generate information such as gamma power / alpha power, time derivative of delta, time derivative of gamma power / alpha power, etc. The time derivative may be calculated across the preceding and subsequent epochs. After calculating the information, the information may then be normalized across the one or more epochs. Various data normalization techniques may be performed, including z-scoring, etc., where the higher frequency data is more clearly visible.

[0045] The intended state of the subject is classified based on one or more frequency-weighted epochs. For example, the one or more frequency-weighted epochs can be clustered by any of a variety of clustering techniques, including K-means clustering. The clustering can be based on information calculated from the epochs (e.g., alpha power, gamma power, standard deviation, maximum amplitude (gamma / alpha), time derivative of delta, time derivative of gamma / alpha, etc.). A causal analysis (e.g., PCA or ICA) can be used to determine the parameter space (e.g., type of information used) in the clustering.

[0046] Following clustering, the epochs may be assigned a designation of an intended state. The epochs designated as intended states may then be presented as indications of the intended and relaxed (unintended) states of the subject during the time period represented by the epoch. The classification may also incorporate manually determined intended states (e.g., manually determined "intentional activity" vs. "relaxed" states). Additionally, artifact information may be utilized in the classification.

[0047] Artifact data may also be used in intent state classification. For example, artifacts may be used to analyze whether an epoch that was originally assigned an intent state designation may be reassigned a new intent state designation due to adjacent artifact data. Thus, for example, artifact data may be utilized in data smoothing techniques.

[0048] Any of a variety of data smoothing techniques can be used during the assignment of intent states. For example, numbers (e.g., 0 and 1) can be used to represent the designated intent states. The brain state designation numbers of adjacent epochs can then be averaged to determine if one of the epochs has been erroneously assigned an intent state designation. Thus, if a group of epochs are assigned intent state designations that show rapid fluctuations in brain states, smoothing techniques can be applied to improve the accuracy of the assignment.

[0049] The previous embodiment has shown how normalization, for example using z-scoring, can be used to allow the analysis of more information from the brain activity signal. The analysis performed so far normalized the power information in the frequency direction. The normalization preferably used z-scoring, but any other type of data normalization can be used. The normalization used is preferably unit-less, such as z-scoring. As is well known in the art, z-scoring can be used to normalize a distribution without changing the shape of the envelope of the distribution. The z-scores are essentially changed to units of standard deviation. The normalized units of each z-score reflect the amount of power in the signal relative to the mean of the signal. The scores are converted to a form of mean deviation by subtracting the mean from each score. The scores are then normalized relative to the standard deviation. All z-scored normalized units have a standard deviation equal to unity.

[0050] Although the above describes normalization using Z-scores, it should be understood that other normalizations can be performed, including T-scores, etc. Multiple normalizations can also be used. Normalization can be performed by normalizing frequency in the time direction or normalizing time in the frequency direction.

[0051] The above embodiments describe normalizing the power at each frequency within a certain range. The range can be from 0 to 100hz, or 128hz, or 500hz. The frequency range is only limited by the sampling rate. With a typical sampling rate of 30KHz, up to 15KHz can be analyzed.

[0052] According to this embodiment, an additional normalization is performed to normalize the power in time for each frequency. This results in frequency- and time-normalized information, which is used to create a normalized spectrogram. This embodiment may obtain additional information from EEG data, which describes automatically detecting different periods of intention and relaxation from the analyzed data. According to an important feature, a single channel of EEG activity (acquired from a single location on the human skull) is used for the analysis. As mentioned above, the acquired data may be one channel of EEG information from a human or other subject. The acquired EEG data may be collected, for example, using a 256Hz sampling rate, or may be sampled at a higher rate. The data is divided into epochs, for example 30-second epochs, and characterized according to frequency.

[0053] A first frequency normalization is performed. The power information is normalized using a z-scoring technique at each frequency bin. In this embodiment, the bins may range from 1 to 100 Hz, with 30 bins per Hertz. The normalization is performed in the time direction. This creates a normalized spectrogram or NS, where each frequency band from the signal has approximately the same weighting. In this embodiment, each 30 second epoch is represented by a "preferred frequency", which is the frequency with the largest z-score within that epoch.

[0054] This produces a specific frequency space, called the preferred frequency space. An analysis of how these patterns are formed and the characteristics of the patterns can be done. Thus, different brain states can be defined according to a discriminant function that looks for activity in some regions and inactivity in other regions. The function can evaluate the brain state according to which frequencies in the region have activity and which do not.

[0055] More generally, however, any form of dynamic spectral scoring can be performed on the corrected data. The discriminant function may require a feature value, or may simply require that a certain amount of activity is present or absent in each of a number of frequency ranges. The discriminant function may simply be matching the envelope of the frequency response. The discriminant function may also look at spectral and temporal fragmentation.

[0056] A second normalization is performed in the frequency direction. The second normalization produces a doubly normalized spectrogram. This produces a new frequency space in which bands are even more pronounced. The values ​​of the doubly normalized spectrogram can be used to produce a filter that maximally separates the values ​​in the space.

[0057] A clustering technique performed on doubly normalized frequencies. For example, the clustering technique may be the K-means technique as described in the previous embodiment. Each cluster may represent an intent state.

[0058] The clusters are in fact multi-dimensional clusters and can themselves be graphed to find additional information. The number of dimensions can depend on the number of clustering variables. This shows how the doubly normalized spectrogram allows for even more measurement properties.

[0059] It is also possible to measure the average spread in the power normalized in the frequency direction, which indicates the fragmentation of the spectrum. Alternatively, the fragmentation value can be based on the temporal fragmentation for the different states, which can also be used as part of the discrimination function.

[0060] These two functions are evaluated on doubly normalized spectra and rely on a uniform increase in gain at all frequencies, which in a singly normalized spectrum would cause motion artifacts and result in anomalously elevated fragmentation values. These fragmentation values ​​can be used as part of a discriminant function. Importantly, as noted above, this discriminant function is typically not evident from any previous analytical techniques, including traditional manual techniques.

[0061] The calculations can be characterized by segmentation or can use overlapping or sliding windows to increase time alignment. This enables many techniques that were not possible before. By performing characterization on the fly, this allows for dynamic spectral scoring to distinguish between states of relaxation and states of intention using only the EEG signature.

[0062] The exemplary methods of data analysis described above were combined with standard non-invasive EEG methods for humans. The result is the ability to non-invasively extract attenuated rhythms in animals, automatically analyze brain activity from single channel EEG, and fully classify the brain state parameters of the animals. EXAMPLES

[0063] A single channel iBrain® EEG recording was performed in a well-functioning 70-year-old ALS patient attempting to move one of his limbs following a verbal cue (left and right hand and foot). The raw EEG signal was analyzed with the SPEARS algorithm, in part, to make high frequency / low spectral power signals detectable. A video recording was acquired at the same time. During the movement attempt, the subject's brain activity showed clear, broad spectral pulses that spanned the gamma and very high gamma ranges. Such pulses were present even in the absence of actual movement and were absent when the subject was not attempting a movement. As expected, activity in the alpha range was detected when the subject closed his eyes. Using such a high-bandwidth biomarker, based on intentional movements against a library of words, allows the signal to be translated into speech, thus providing the ALS patient with a communication tool that relies more on the brain than the body.

[0064] Specifically, application of the method of the present invention reveals patterns of higher frequencies that correspond to the timing of the subject's actual, imagined, or intended movements. In an exemplary application, a frequency spectrum is generated from the time series data and normalized to reveal these higher frequencies that are not revealed by standard methods. This application to brain EEG data is illustrated in FIG. 1. In this example, a highly functional 70-year-old subject with ALS who is unable to move is asked to close his eyes, to remain still, or to imagine his hands or feet. For each of these tasks, the subject was given a verbal cue to start the task and another cue 4 seconds later to stop. The task was repeated after 6 seconds, for a total of 12 attempts and 120 seconds of time per task. FIG. 1 shows the standard frequency power spectrum (i) and the enhanced frequency power spectrum (ii) of the ALS subject while performing these tasks. Stronger signals appear redder, while weaker signals are reduced in intensity with shades of orange, yellow, and blue. Clear bands of high frequency activity appear at high intensity in the enhanced spectrum, approximating the timing of cues for the start and end of the task. Figure 1A illustrates this for a task of closing the eyes for 4 seconds. Figure 1B illustrates this for a task of imagining or attempting to clench the left hand for 4 seconds and relaxing for 6 seconds. Figure 1C, D, and E illustrate similar tasks, but for imagining or attempting to clench the right hand, left foot, and right foot, respectively. In all cases, the spectrograms generated in this application clearly reveal high frequency spectral content that approximates the timing of the subject's attempts to perform the instructed task. Figure 1F shows the same spectral analysis of the same subject but at rest, displaying high frequency content at quite different timings, which may represent environmental noise, background talk, and / or baseline electrical activity. EXAMPLES

[0065] Application of the method of the present invention in ALS patients described in Example 1 reveals components of the data that approximate the time of the event. Figure 2 shows one such application of the method of the present invention (here by a single channel component analysis in a doubly normalized spectrogram) to reveal unrelated and independent data components. Figures 2A-E are data from the same tasks shown in Figures 1A-E. Each plot shows the resulting extracted independent components after all the described analyses, with peaks that approximate the individual tasks attempted by the subject. The green line shows the timing of the verbal cue to start the task for 4 seconds, and the red line shows the time of the verbal cue to stop the task and relax for 6 seconds. The peaks in each component are roughly aligned with the start of the task being performed (indicated by the peak points in, on, or immediately after the green line). This analysis may also be combined with that in Example 1 to enhance or supplement the detection and timing of events. EXAMPLES

[0066] The application of the method of the present invention to single channel brain EEG data in the ALS patient described in Example 1 to assess data stability and to reveal changes in data stability indicative of intentional activity is provided. Specifically, application of the method of the present invention to generate temporal fragmentation reveals changes in stability that approximate the timing and duration of intentional activity. Figure 3, A to E, shows temporal fragmentation of the same task as shown in Figure 1, A to E. In each plot, the shifts of points from negative to positive represent decreases in stability. These shifts approximate both the verbal cue to start the task (green line), where stability begins to decrease and the shift to positive begins, and the verbal cue to end the task (red line), where data begins to stabilize and the shift to positive ends. Zero line crossings (purple dots) indicate increases in unstable periods (during or immediately after the task) and decreases in the count of stable periods (relaxation periods before or after each task). These shifts and line crossings may also be combined with the method of Example 1 to enhance or reinforce event detection and timing. EXAMPLES

[0067] Detection of high frequency events and timing of events relative to low frequency events in an ALS patient of Example 1. Generation of a spectrogram, followed by one or more normalizations across one or both data axes, followed by feature sharpening using the doubly normalized spectrogram, followed by extraction of standard known frequencies, frequency ranges, their sums, their ratios, and / or other frequency relationships reveals the timing, intervals, and / or duration of events. Figures 4A-E show the tasks in Figures 1A-B, plotting the sum of high frequency power in the gamma and very high gamma (hgamma) range of frequencies (all greater than 30 Hz). In the brain EEG, increased power at these frequencies is associated with heightened concentration, such as occurs when imagining moving a limb. The peak in the sum of high frequency power approximates the timing of the task, indicated by the green (task start) and red (task end) columns of lines. Figure 5A-D show the tasks in Figure 1 for grasping the right hand, left hand, right foot, and left foot, respectively, plotting the ratio of the sum of alpha frequency (8-13 Hz) power to the sum of gamma frequency power (30-50 Hz). Increased power in alpha frequency in the brain EEG is associated with relaxation of mental effort, such as after a period of attempting to grasp the limb. Thus, alpha and gamma frequency power are inversely related for these tasks, with peaks in the ratio of alpha to gamma appearing between tasks (between the red line indicating stopping and the green line indicating starting again). Figure 5E shows an analysis of delta frequency (<5 Hz) during the eyes-closing task. Delta frequency power is associated with changes in eyes open or closed, approximating the timing of subjects closing (green line) and opening (red line) their eyes. These and other frequency analyses may be combined with each other and with the methods in Examples 1, 2, and 3 to enhance or augment event detection and timing, and to characterize events with known associated brain states indicated by the analyzed frequencies. EXAMPLES

[0068] Simultaneous detection of multiple events occurring at the same time for a second ALS patient. Generation of a spectrogram, followed by one or more normalizations across one or both data axes, followed by feature sharpening, followed by extraction of standard known frequencies, frequency ranges, their sums, their ratios, and / or other frequency relationships reveals multiple events in one analysis. Figure 8 shows data from a task designed to induce both increased focus (increased gamma frequency power) and increased shifts in thinking (changes in alpha frequency power) indicative of decreased and increased relaxation. A non-mobile ALS patient subject was instructed to alternate between two 10-second imaginings (kicking a football: green line, and looking at the bedroom: red line) five times. Figure 8A plots normalized and enhanced gamma frequency power, showing a steep increase at the beginning of the alternating sequence (first green line, 13 seconds) and a subsequent decrease at the end of all sequences (last red line, 103 seconds). FIG. 8B plots normalized and enhanced alpha frequency power, showing the change in relaxation state (peaks between colored lines) between the two different imaginations. FIG. 8C plots normalized and enhanced alpha and gamma power, showing the simultaneous detection of both the plateau in gamma frequency power during the imagination sequence and the peak in alpha as imaginations change, as two unique signals detected in one analysis. FIG. 8D plots normalized and enhanced alpha to very high gamma (hgamma) ratios, with a steep drop in ratio at the beginning of the sequence, individual peaks during each imagination, and a rise in ratio after the sequence. EXAMPLES

[0069] Use of iteratively normalized spectrograms to distinguish and characterize multiple types of intentional events. Analysis involving application of the SPEARS algorithm followed by a two-sample Kolmogorov-Smirnov (KS) test for sampling at the same distribution between any two spectrograms reveals distinguishable imagined motor actions. This application allows multiple degrees of freedom based on at least one event type being distinguishable from the others. The p-values ​​of the KS test for the cross-axially 10-fold normalized spectrograms from each task listed in Figure 1A-E performed by the immobile ALS subject indicate that each spectrogram comes from a distribution different from that at rest (P<0.01) (Table 1) and from that of most any other task (P<0.05) (Figure 10), making it an effective application for characterizing the same or multiple events at detection, although to a lesser extent for the left hand and right foot in this particular trial.

[0070] Table 1. p-values ​​of the KS test for the imaginary task versus the resting state TIFF0007667835000001.tif73128

[0071] Throughout this application, various publications, patents, and / or patent applications are referenced in order to more fully describe the state of the art to which this invention pertains. The disclosures of these publications, patents, and / or patent applications are hereby incorporated by reference in their entireties and with respect to subject matter specifically referenced in the same or preceding sentence to the same extent as if each individual publication, patent, and / or patent application was specifically and individually indicated to be incorporated by reference.

[0072] Although only a few embodiments have been disclosed in detail above, other embodiments are possible and the inventors intend these to be encompassed within this specification. This specification describes specific examples for achieving a more general goal that may be achieved in another way. This disclosure is intended to be exemplary, and the claims are intended to cover any modifications or alternatives that would be predictable to one of ordinary skill in the art. For example, other applications are possible, and other forms of discriminant functions and characterizations are possible. Although the above has broadly described characterizing frequencies in terms of their "preferred frequencies," it should be understood that more precise characterization of the information may be possible. Also, while the above only refers to determining intent states from EEG data, and only a few different types of intent state determinations, it should be understood that other applications are possible.

[0073] While the principles of the present invention have been shown and described in exemplary embodiments, it will be apparent to those skilled in the art that the described embodiments are exemplary and illustrative embodiments and that changes in arrangement and detail may be made without departing from such principles. Techniques from any embodiment may be incorporated into any other embodiment or embodiments. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the invention being indicated by the following claims.

Claims

1. receiving data indicative of a subject's brain activity, at least a portion of the data being associated with intentional brain signals imagined or generated by the subject; generating a doubly normalized spectrogram from the received data indicative of brain activity by calculating a spectrogram of the data, normalizing the spectrogram of the data by time at least once by frequency, and normalizing the spectrogram of the data by frequency at least once by time; forming a plurality of clusters from the doubly normalized spectrogram, at least one cluster of the plurality of clusters representing brain activity indicative of an intentional higher-order cognitive function and at least one other cluster of the plurality of clusters representing brain activity indicative of a relaxed state; and translating the data associated with the at least one cluster of the plurality of clusters representative of brain activity indicative of the intended higher-order cognitive function into commands associated with a peripheral device to cause the peripheral device to perform a task associated with the intended higher-order cognitive function. The method includes:

2. 10. The method of claim 1, wherein the data indicative of the subject's brain activity is collected via a system comprising at least one sensor attached to the subject.

3. The method of claim 2 , wherein the at least one sensor includes at least one of a single wet electrode or a single dry electrode.

4. The method of claim 2 , wherein the data is collected via a system comprising the at least one sensor non-invasively attached to the subject.

5. 2. The method of claim 1, wherein the data is received from at least a single channel of EEG, EMG, EOG, MEG, ECoG, iEEG, fMRI, LFP or a peripheral channel modulated by the subject's intentional brain signals.

6. The method of claim 1 , wherein the data is received from at least a single channel of EEG.

7. The task, Controlling the cursor; Controlling the prosthesis; and Generating synthetic speech The method of claim 1, wherein the compound is selected from the group consisting of

8. The method of claim 1 , wherein the plurality of clusters are formed by applying a K-means clustering technique to the doubly normalized spectrogram.

9. The method of claim 1 , further comprising segmenting the received data into a number of epochs.

10. The method of claim 9 , wherein segmenting the received data into the plurality of epochs comprises segmenting the data into a plurality of time segments.

11. The method of claim 10 , wherein the data is segmented into a plurality of time segments via at least one of a scanning window or a sliding window.

12. The method of claim 11 , wherein at least one cluster is formed for each of the multiple epochs.

13. The step of: assigning an intentional state to each of the epochs, the intentional state comprising one of the higher-order cognitive function and the relaxed state; evaluating one of the epochs against adjacent epochs of the intent state of the one of the epochs; and modifying the intent state of one epoch when an abrupt change in brain state between adjacent epochs is identified.

13. The method of claim 12, further comprising:

14. one or more data processors; and A non-transitory computer readable storage medium containing instructions that, when executed by one or more data processors, cause the one or more data processors to: receiving data indicative of a subject's brain activity, at least a portion of the data being associated with intentional brain signals imagined or generated by the subject; generating a doubly normalized spectrogram from the received data indicative of brain activity by calculating a spectrogram of the data, normalizing the spectrogram of the data by time at least once by frequency, and normalizing the spectrogram of the data by frequency at least once by time; forming a plurality of clusters from the doubly normalized spectrogram, at least one cluster of the plurality of clusters representing brain activity indicative of an intentional higher-order cognitive function and at least one other cluster of the plurality of clusters representing brain activity indicative of a relaxed state; and and translating the data associated with the at least one cluster of the plurality of clusters representing brain activity indicative of the intentional higher-order cognitive function into a command associated with a peripheral device to cause the peripheral device to perform a task associated with the intentional higher-order cognitive function. A non-transitory computer-readable storage medium for causing a set of actions to be performed, the set of actions including: A system including:

15. 15. The system of claim 14, wherein the data indicative of the subject's brain activity is collected via at least a single sensor attached to the subject.

16. 16. The system of claim 15, wherein the at least one sensor includes at least one of a single wet electrode or a single dry electrode.

17. The system of claim 15 , wherein the at least one sensor is non-invasively attached to the subject.

18. 15. The system of claim 14, wherein the data is received from at least a single channel of EEG, EMG, EOG, MEG, ECoG, iEEG, fMRI, LFP or a peripheral channel modulated by the subject's intentional brain signals.

19. The system of claim 14 , wherein the data is received from at least a single channel of EEG.

20. The task, Controlling the cursor; Controlling the prosthesis; and Generating synthetic speech The system of claim 14, wherein the selected

21. 15. The system of claim 14, wherein the plurality of clusters are formed by applying a K-means clustering technique to the doubly normalized spectrogram.

22. 15. The system of claim 14, wherein the instructions, when executed by the one or more data processors, cause the one or more data processors to segment the received data into a number of epochs.

23. 23. The system of claim 22, wherein segmenting the received data into the plurality of epochs comprises segmenting the data into a plurality of time segments.

24. 24. The system of claim 23, wherein the data is segmented into a plurality of time segments via at least one of a scanning window or a sliding window.

25. 25. The system of claim 24, wherein at least one cluster is formed for each of the plurality of epochs.

26. When executed by the one or more data processors, the instructions cause the one or more data processors to: assigning an intended state to each of the epochs, the intended state including one of the intended higher-order cognitive function and the relaxed state; evaluating one of the epochs against adjacent epochs of the intent state of the one of the epochs; and Modifying the intent state of an epoch when an abrupt change in brain state between adjacent epochs is identified. The system of claim 25,

Images