Method and apparatus for distinguishing unnatural from natural brain patterns of electroencephalographic (EEG) activity

Abstract

An apparatus and method for distinguishing unnatural from natural brain patterns of electroencephalogram (EEG) activity are provided. The method comprises: mathematically processing a specified section of an input electroencephalogram (EEG) record to extract features thereof; and comparing the extracted features against natural sleep EEG data to detect unnatural brain states.

Description

METHOD AND APPARATUS FOR DISTINGUISHING UNNATURAL FROM NATURAL BRAIN PATTERNS OF ELECTROENCEPHALOGRAPHIC (EEG) ACTIVITYCross-Reference to Related Application

[0001] The subject application claims the benefit of and priority to U.S.Provisional Application Serial No. 63 / 745,585 filed on January 15, 2025, the entire content of which is incorporated herein by reference. The subject application is also related to U.S. Patent Application No. 14 / 426,523 filed on September 12, 2013, now issued under U.S. Patent No. 9,763,589 on September 19, 2017, the relevant portions of which are incorporated herein by reference.Field

[0002] The subject application is directed to a method and apparatus for distinguishing unnatural from natural brain patterns of electroencephalographic (EEG) activity.Background

[0003] The brain exists in a variety of states during one’s lifetime or even within minutes or hours. These brain states can be broadly classified into natural and unnatural brain states. Natural brain states are those encountered during daily living in the absence of pharmacological or other agents that alter brain states (e.g., anesthesia, medications, drug abuse...etc.) or organic brain disorders (e.g., seizures, tumors, encephalopathy... etc.). Unnatural brain states are those produced by brain disease or by pharmacological agents.

[0004] The electroencephalogram (EEG) is commonly used to assess brain states of subjects during natural and unnatural conditions. The EEG is very commonly used to evaluate the different stages of natural sleep in the investigation of sleep disorders. For evaluation of sleep disorders, the EEG is typically evaluated (scored) visually by expert technologists, although several digital scoring systems have recently been introduced. The intent of the use of EEG in the investigation of sleep disorders is to determine how much time subjects spend awake and in different stages of sleep depth, as well as the frequency of brief arousals.

[0005] The EEG is also used for the investigation of unnatural brain states such as depth of anesthesia. For this application, some features of the EEG, and muscle activity, are combined to develop an index that reflects depth of anesthesia by reference to clinical signs and anesthetic dosages that correlate with level ofanesthesia. Examples of such systems are the Bispectral index (BIS) monitor and EEG Entropy monitor. These methods are not widely used probably because of the many types of anesthetics used and the likely different EEG patterns that are produced by different anesthetics. For epileptic seizures, the EEG is used for continuous monitoring of subjects with numerous electrodes (high density EEG) to document the EEG changes that occur during seizures and, in particular, where in the brain the seizures activity started.

[0006] Above-referenced U.S. Patent No. 9,763,589 (the “ORP patent”) discloses a method to distinguish wakefulness from different levels of sleep depth in Natural Sleep using the EEG to generate an Odds Ratio Product (ORP). ORP is an empirically derived EEG-based metric that assess brain activity across a continuum of sleep depths and provides a continuous index of sleep depth. ORP ranges from 0 (very deep sleep) to 2.5 (full wakefulness). Preliminary steps in generating ORP is to perform frequency domain analysis of one or more discrete sections of an EEG test record to determine EEG power at different specified frequency bands relevant to natural sleep. Power in each frequency band is then assigned a rank based on its location within the range observed for this frequency band in a plurality of EEG records obtained previously from subjects with and without sleep disorders during clinical sleep studies. The ranks related to the specified EEG frequency ranges are then concatenated into one number or code, referred to as a BIN number, that reflects the powers in the different frequency ranges relative to each other, and by extension reflects a specific EEG pattern in the EEG section being analyzed. This process results in a large number of possible EEG patterns given by the number of subranges within each frequency range raised to the power n, where n is the number of specified frequency ranges. Thus, dividing the full range of power in each frequency range by 10 (deciles) and measuring four (4) frequency ranges gives 10A4 (10,000) potential EEG patterns.

[0007] In the ORP patent, each of the potential EEG patterns is further assigned a probability of occurring during wake stages or during arousals, by reference to a look-up table derived from a plurality of previously recorded sleep studies and scored by qualified sleep technologists. This probability is then used to determine brain status at any moment in the sleep study being analyzed; the lower the value, the deeper sleep is.

[0008] While ORP has proven extremely useful in determining depth of natural sleep, to-date there has been no objective measure of the effect of sedation on brain activity and how to differentiate it from sleep. It is therefore an object to provide anovel method and apparatus for distinguishing unnatural from natural brain patterns of EEG activity.

[0009] This background serves only to set a scene to allow a person skilled in the art to better appreciate the following brief and detailed descriptions. Therefore, none of the above discussion should necessarily be taken as an acknowledgement that this background discussion is part of the state of the art or is common general knowledge.Brief Description

[0010] It should be appreciated that the following Brief Description is provided to introduce a selection of concepts in a simplified form that is further described below in the Detailed Description. This Brief Description is not intended to limit the scope of the claimed subject matter.

[0011] Since its original introduction in 2015, the ORP has been used to determine depth of natural sleep in tens of thousands of clinical sleep studies on normal subjects and ambulatory patients with sleep disorders. While the ORP has proven to be very useful for this purpose, Applicant has found that the large number of possible EEG patterns revealed during the ORP clinical sleep studies has proven to be useful in distinguishing between natural and unnatural brain states. One remarkable finding that has been made is that in normal sleepers and patients with common sleep disorders (sleep apnea, insomnia, movement disorders), over 50% of all the theoretically possible EEG patterns calculated, and which generated 10,000 potential EEG patterns, are not seen, or are seen very rarely, in natural sleep whether the sleepers / patients suffer from a sleep disorder or not. This has indicated that the normal brain utilizes only a fraction of the possible EEG patterns when sleep occurs naturally. Sleep disorders result in different distributions of the same EEG patterns seen in normal subjects, and this simply reflects different sleep depth or wake time, but the EEG patterns seen are the ones seen in normal subjects, sleeping naturally. A search of ORP clinical sleep study results by the Applicant for unnatural EEG patterns to patients with neurologic (strokes, Parkinson’s disease), and disorders of excessive sleepiness (narcolepsy, idiopathic hypersomnia) was expanded and, again, the unnatural EEG patterns were either not seen or not seen in numbers that exceed those found in subjects without these disorders.

[0012] More recently, sleep studies obtained from critically ill patients in intensive care units (ICUs) have been scored. In the later stages of their illness, when they are prepared for removal of mechanical ventilation (weaning), critically ill patients typically receive mild sedation with a variety of drugs. In this phase of their illness,and under certain sedatives (propofol, dexmedetomidine), their sleep continues to show the usual EEG patterns seen in natural sleep in ambulatory patients (Georgopoulos D, Kondili E, Alexopoulou C, Younes M. Effects of Sedatives on Sleep Architecture Measured With Odds Ratio Product in Critically III Patients. Crit Care Explor. 2021 Aug 10;3(8):e0503. doi: 10.1097 / CCE.0000000000000503. PMID: 34396142). However, in the acute phase of critical illness, patients typically receive heavy sedation to help them tolerate the ICU environment, intubation, and mechanical ventilation, and the discomfort of the critical illness for which they were admitted to the ICU. Recently several EEG studies from critically ill patients (in the ICU) in the acute phase of the patients’ illness were analyzed by the Applicant, where heavy sedation was used. Remarkably, it has been found that many of the unnatural EEG patterns that are never seen, or are very rare, in natural sleep, occur in abundance in these patients. These unnatural EEG patterns differ in type and prevalence under different anesthetics and doses.

[0013] It is postulated that the type of unnatural EEG patterns and their prevalence in a given patient receiving continuous intravenous infusion of sedatives / anesthetics reflect the type of drug used and the level of sedation / anesthesia. While further studies are needed to determine the association of prevailing unnatural EEG patterns, and their frequency, with clinical disorders and sedative / anesthetic type and depth of sedation / anesthesia, these recent observations clearly indicate that certain EEG patterns exist that can distinguish natural from unnatural brain states and that this has important implications to the diagnosis and management of conditions associated with altered brain function.

[0014] Accordingly, in one aspect there is provided a method comprising: mathematically processing a specified section of an input electroencephalogram (EEG) record to extract features thereof; and comparing the extracted features against natural sleep EEG data to detect unnatural brain states.

[0015] In one embodiment, the extracted features are compared against the natural sleep EEG data to identify EEG patterns that never or rarely occur in natural sleep.

[0016] In one embodiment, the natural sleep EEG data is generated from EEG records of subjects during natural sleep and with no known brain disorder and not on any agents / drugs that alter brain state.

[0017] In one embodiment, the extracted features are spectral features of the input EEG record.

[0018] In one embodiment, the method further comprises determining the prevalence and / or reoccurrence of detected unnatural brain states.

[0019] In one embodiment, epochs of the specified section of the input EEG record are mathematically processed to extract features thereof.

[0020] In one embodiment, the epochs are continuous and non-overlapping. In one form, the epochs are 3-second epochs.

[0021] In one embodiment, the specified section of the input EEG record is the entire input EEG record or one or more portions of the input EEG record.

[0022] In one embodiment, the extracted features are spectral features of the input EEG record and wherein the extracted features are analyzed to calculate power spectral density over specified frequency ranges.

[0023] In one embodiment, mathematical processing of the specified section of the input EEG record comprises performing frequency domain analysis on epochs of a specified section of an input EEG record to determine EEG power at specified frequencies; for each epoch, calculating EEG power over the specified frequency ranges; for each epoch, assigning, for each specified frequency range, a rank to the calculated EEG power, each rank being determined based on values of EEG power encountered in epochs of a plurality of reference EEG records; for each epoch, determining a code based on the assigned ranks; and analyzing the codes determined for the epochs against natural sleep EEG data to detect codes representative of unnatural brain states.

[0024] In one embodiment, wherein the analyzing comprises comparing the codes determined for the epochs to frequency of code occurrence data generated from EEG records of subjects during natural sleep and with no known brain disorder and not on any agents / drugs that alter brain state.

[0025] In one embodiment, the frequency of code occurrence data generated from the EEG records of subjects populates a look-up table and includes at least one frequency of occurrence entry for each possible code. In one form, the at least one frequency of occurrence entry at least comprises median and 95thpercentile of frequency of occurrence entries for each possible code.

[0026] In one embodiment, during the performing frequency domain analysis, EEG power is determined incrementally over the specified frequencies. In one form, the specified frequencies are in the range between 0.33 Hz to 60 Hz and EEG power is calculated at 0.33 Hz intervals over the range. In one form, the specified frequency ranges comprise four frequency ranges. The four frequency ranges may comprise a 0.33 Hz to 2.33 Hz frequency range, a 2.67 Hz to 6.33 Hz frequency range, a 7.0 Hz to 14.0 Hz frequency range, and a 14.3 Hz to 35 Hz frequency range.

[0027] In one embodiment, during the assigning, for each epoch, the EEG powers are summed over the specified frequency ranges and wherein the EEGpower sums are compared to EEG power sum data generated from epochs of the plurality of reference EEG records to determine the rank. In one form, the EEG power sum data generated from epochs of the plurality of reference EEG records populates a look-up table that divides the range of EEG power sums for each frequency range into ten ranks.

[0028] In one embodiment, for each epoch, the ranks are concatenated to form the code with each value of the code representing the level of EEG power in a respective specified frequency range.

[0029] In one embodiment, the method further comprises preprocessing the specified section of the input EEG record. In one form, the preprocessing comprises low-pass and high-pass filtering of the input EEG record.

[0030] In one embodiment, the method further comprises using the detected codes representative of unnatural brain states to adjust sedation / anesthesia of a patient or to prompt an investigative procedure. In one form, sedation / anesthesia of the patient is adjusted manually or automatically.

[0031] In one embodiment, the method further comprises displaying and / or outputting the analyzing results.

[0032] According to another aspect there is provided an apparatus for distinguishing unnatural from natural patterns of electroencephalographic (EEG) activity comprising: memory embodying computer executable instructions; and at least one processor configured to communicate with the memory and to execute the instructions to cause the apparatus to carry out the method of any one of preceding paragraphs

[0014] to

[0031] ,

[0033] According to another aspect there is provided a non-transitory computer-readable medium embodying executable program instructions, which when executed by at least one processor, cause an apparatus to carry out the method of any one of preceding paragraphs

[0014] to

[0031] ,

[0034] According to another aspect there is provided a computer program comprising executable instructions, which when executed by at least one processor, cause an apparatus to carry out the method of any one of preceding paragraphs

[0014] to

[0031] ,

[0035] According to another aspect there is provided an apparatus comprising: memory embodying computer executable instructions; and at least one processor configured to communicate with the memory and to execute the instructions to cause said apparatus to: memory embodying computer executable instructions; and at least one processor configured to communicate with said memory and to execute said instructions to cause said apparatus to: mathematically process a specified section ofan input electroencephalogram (EEG) record to extract features thereof; and compare the results against natural sleep EEG data to detect unnatural brain states.

[0036] In one embodiment, the apparatus is caused to compare the extracted features against the natural sleeps EEG data to identify EEG patterns that never or rarely occur in natural sleep.

[0037] In one embodiment, the natural sleep EEG data is generated from EEG records of subjects during natural sleep and with no known brain disorder and not on any agents / drugs that alter brain state.

[0038] In one embodiment, the apparatus is further caused to determine the prevalence and / or reoccurrence of detected unnatural brain states.

[0039] In one embodiment, the extracted features are spectral features of the input EEG record.

[0040] In one embodiment, epochs of the specified section of the input EEG record are mathematically processed to extract features thereof.

[0041] In one embodiment, the epochs are continuous and non-overlapping. In one form, the epochs are 3-second epochs.

[0042] In one embodiment, the specified section of the input EEG record is the entire input EEG record or one or more portions of the input EEG record.

[0043] In one embodiment, the extracted features are spectral features of the input EEG record and wherein the extracted features are analyzed to calculate power spectral density over specified frequency ranges.

[0044] In one embodiment, wherein during mathematically processing the specified section of the input EEG record, the apparatus is caused to: perform frequency domain analysis on epochs of a specified section of an input EEG record to determine EEG power at specified frequencies; for each epoch, calculate EEG power over the specified frequency ranges; for each epoch, assign, for each specified frequency range, a rank to the calculated EEG power, each rank being determined based on values of EEG power encountered in epochs of a plurality of reference EEG records; for each epoch, determine a code based on the assigned ranks; and analyze the codes determined for the epochs against natural sleep EEG data to detect codes representative of unnatural brain states.

[0045] In one embodiment, wherein during the analyzing, the apparatus is caused to compare the codes determined for the epochs to frequency of code occurrence data generated from EEG records of subjects during natural sleep and with no known brain disorder and not on any agents / drugs that alter brain state

[0046] In one embodiment, the frequency of code occurrence data generated from the EEG records of subjects populates a look-up table and includes at least onefrequency of occurrence entry for each possible code. The at least one frequency of occurrence entry may at least comprise median and 95thpercentile of frequency of occurrence entries for each possible code.

[0047] In one embodiment, during the performing frequency domain analysis, the apparatus is caused to determine EEG power incrementally over the specified frequencies. In one form, the specified frequencies are in the range between 0.33 Hz to 60 Hz and EEG power is calculated at 0.33 Hz intervals over the range. In one form, the specified frequency ranges comprise four frequency ranges. In one form, the four frequency ranges comprise a 0.33 Hz to 2.33 Hz frequency range, a 2.67 Hz to 6.33 Hz frequency range, a 7.0 Hz to 14.0 Hz frequency range, and a 14.3 Hz to 35 Hz frequency range.

[0048] In one embodiment, during the assigning, the apparatus is caused to, for each epoch, sum the EEG powers over the specified frequency ranges and compare the EEG power sums to EEG power sum data generated from epochs of the plurality of reference EEG records to determine the rank. In one form, the EEG power sum data generated from epochs of the plurality of reference EEG records populates a look-up table that divides the range of EEG power sums for each frequency range into ten ranks.

[0049] In one embodiment, for each epoch, the apparatus is caused to concatenate the ranks to form the code with each value of the code representing the level of EEG power in a respective specified frequency range.

[0050] In one embodiment, the apparatus is caused to preprocess the specified section of the input EEG record. In one form, during preprocessing, the apparatus is caused to low-pass and high-pass filter the input EEG record.

[0051] In one embodiment, the apparatus further comprises a display device and wherein the apparatus is further caused to display information representative of unnatural brain states on the display device for use in adjusting sedation / anesthesia.

[0052] In one embodiment, the apparatus is caused to provide output to a drug delivery system for use in adjusting sedation / anesthesia.Brief Description of the Drawings

[0053] Embodiments will now be described more fully with reference to the accompanying drawings in which:

[0054] Figure 1 is a block diagram of an apparatus for distinguishing unnatural from natural brain patterns of electroencephalographic (EEG) activity;

[0055] Figure 2 is a flowchart of steps performed for distinguishing unnatural from natural brain patterns of electroencephalographic (EEG) activity; and

[0056] Figure 3 is another flowchart of further steps performed for distinguishing unnatural from natural brain patterns of electroencephalographic (EEG) activity.Detailed Description

[0057] The foregoing summary and the following detailed description of certain examples will be better understood when read in conjunction with the appended drawings. As used herein, an element or feature introduced in the singular and preceded by the word "a" or "an" should not necessarily exclude the plural of the elements or features. Further, references to "one example" or “one embodiment” are not intended to be interpreted as excluding the existence of additional examples or embodiments that also incorporate the described elements or features. Moreover, unless explicitly stated to the contrary, examples or embodiments "comprising" or "having" or “including” an element or feature or a plurality of elements or features having a particular property may include additional elements or features not having that property. Also, it will be appreciated that the terms “comprises,” “has,” and “includes” means “including but not limited to,” and the terms “comprising,” “having,” and “including” have equivalent meanings.

[0058] As used herein, the term “and / or” can include any and all combinations of one or more of the associated listed elements or features.

[0059] Reference herein to “example” means that one or more feature, structure, element, component, characteristic, and / or operational steps described in connection with the example is included in at least one embodiment and / or implementation of the subject matter according to the subject disclosure. Thus, the phrases “an example,” “another example,” and similar language throughout the subject disclosure may not necessarily refer to the same example. Further, the subject matter characterizing any example may, but does not necessarily, include the subject matter characterizing any other example.

[0060] Reference herein to “configured” denotes an actual state of configuration that fundamentally ties the element or feature to the physical characteristics of the element or feature preceding the phrase “configured to.”

[0061] Unless otherwise indicated, the terms “first,” “second,” etc., are used herein merely as labels and are not intended to impose ordinal, positional, or hierarchical requirements on the items to which these terms refer. Moreover, reference to a “second” item does not require or preclude the existence of a lower-numbered item (e.g., a “first” item) and / or a higher-numbered item (e.g., a “third” item).

[0062] In general, a method and apparatus for distinguishing unnatural from natural brain states of electroencephalographic activity are described below. During the method, an input real-time or previously recorded EEG signal or record (hereinafter referred to as “record”) is mathematically processed to extract features thereof that are compared against natural sleep EEG data to detect unnatural brain states i.e. EEG patterns that never or rarely occur during natural sleep. Detected unnatural EEG patterns can be used to carry out a protocol, such as, managing sedation / anesthesia, prompting investigative procedures to determine the existence of a pathology, generating reports outlining the prevalence of unnatural EEG and the times that were detected etc.

[0063] In one embodiment, spectral features are extracted from the input EEG record. In particular, frequency domain analysis is performed on epochs of a specified section of an input EEG record to determine EEG power at specified frequencies. For each epoch, EEG power is calculated over specified frequency ranges. For each specified frequency range of each epoch, a rank is assigned to the calculated EEG power, each rank being determined based on values of EEG power encountered in epochs of a plurality of reference EEG records. For each epoch, a code is determined based on the assigned rank. The codes are analyzed against the natural sleep EEG data to detect codes representative of unnatural brain states. Further specifics of the method and apparatus will now be described.

[0064] Turning now to Figure 1 , an apparatus for distinguishing unnatural from natural brain states of electroencephalographic (EEG) activity is shown and is generally identified by reference numeral 50. In this embodiment, the apparatus 50 is in the form of a programmed computing device such as a personal computer, server, distributed network or other suitable processing device(s). The programmed computing device comprises, for example, a processing unit 52 comprising one or more processors, system memory 54 (volatile and / or non-volatile memory), other non-removable or removable memory 56 (e.g., a hard disk drive, RAM, ROM, EEPROM, CD-ROM, DVD, flash memory, etc.), a display monitor or screen 58 and a system bus 60 coupling the various programmed computing device components to the processing unit 52. The programmed computing device also comprises an input interface 62 configured to receive input EEG records for storage in memory 56 and processing by the processing unit 52. The programmed computing device may also comprise networking capabilities using Ethernet, WiFi, and / or other suitable network format, to enable connection to shared or remote drives, one or more networked computers, or other networked devices. The programmed computing device 50 mayoptionally comprise one or more other input devices such as a mouse, keyboard, trackball etc.

[0065] The memory 56 stores software comprising executable program instructions that are executed by the processing unit 52 to allow received input EEG records to be processed to allow natural and unnatural brain states of EEG activity to be distinguished. In this embodiment, the software is written in C#, although those of skill in the art will appreciate that other suitable program languages may be used to write the executable program instructions. The apparatus 50 can be used to process input EEG records in real time or can be used to process input EEG records post hoc as will be described.

[0066] Figure 2 is a flowchart of the general steps performed by the apparatus 50 during execution of the executable program instructions in order to process an input EEG record. When an input EEG record is to be processed, the input EEG record is mathematically processed to extract features thereof (step 70) that are then compared against natural sleep EEG data to detect unnatural brain states i.e. EEG patterns that never or rarely occur during natural sleep (step 72). The results of the comparison are then used to determine and carry out a protocol (step 74) as will be described.

[0067] During step 70, various mathematical techniques may be employed to mathematically process the input EEG record to extract features thereof. For example, quantitative EEG analysis, Entropy or other suitable technique may be employed to extract features from discrete sections, segments or epochs (hereinafter referred to as “epochs”) of the input EEG record and compare the extracted features against the natural sleep EEG data. In this embodiment, quantitative EEG analysis is employed that performs spectral or frequency domain analysis by applying a fast Fourier transform to the input EEG record allowing calculation of power spectral density for specified frequency ranges as will be further described.

[0068] Turning now to Figure 3, a flowchart showing further steps performed by the apparatus 50 during execution of the executable program instructions in order to process the input EEG record is shown. In this embodiment, the input EEG record is initially pre-processed, if necessary (step 80). The pre-processed input EEG record is then split into consecutive, non-overlapping epochs. In this embodiment, the input EEG record is split into 3-second epochs (step 82). Those of skill in the art will however appreciate that the length of the epochs into which the input EEG record is split may be adjusted.

[0069] Frequency domain analysis is then performed on the epochs (step 84) followed by calculation of total EEG power in different frequency ranges (step 86). The calculated total EEG power in the different frequency ranges is then ranked usingEEG power data generated from previously ranked reference EEG records as will be described. The ranks are then concatenated into a BIN code for each epoch of the input EEG record (step 88). The BIN codes are then analyzed against natural sleep EEG data as will be described to determine whether the EEG patterns in the epochs as represented by the BIN codes represent a natural or unnatural brain state (step 90). The results of the analysis are then used to determine and carry out a protocol (step 74).

[0070] For example, depending on the environment in which the input EEG record is obtained, detection of unnatural EEG patterns can be used to control or manage sedation / anesthesia, identify the need to test for pathologies, generate reports etc.

[0071] During pre-processing at step 80, if the input EEG record has not been previously filtered, a 35Hz low-pass filter and a 0.3 Hz high pass filter are applied to the input EGG record. Also, if the input EEG record was generated using a sampling rate higher than 120 Hz, the input EEG record is downsampled to 120 Hz.

[0072] Following preprocessing at step 80, if necessary, the input EEG record is divided into epochs at step 82 as described above. During frequency domain analysis at step 84, a Fourier Transform is performed on the epochs across a specified section of the input EEG record, which usually includes the entire input EEG record but not necessarily. A montage file is used to instruct the software which EEG channels of the specified section of the input EEG record to use. The software is capable of scoring two (2) EEG channels at a time. Most commonly the central EEG derivations are used (C3 / M2 and C4 / M1) but any EEG channel can be analyzed provided that the spectral analysis performed on the reference EEG records to generate the EEG powers for those reference EEG records is done using the same channels.

[0073] Using the Fourier Transform, the software determines, for each epoch, EEG power at frequencies from 0.33 Hz to 60.0 Hz in 0.33 Hz increments, thereby generating one hundred and eighty (180) EEG power values for each epoch. Next the software sums the EEG power values that occur in each of the following four (4) frequency ranges: 0.33 Hz to 2.33 Hz, 2.67 Hz to 6.33 Hz, 7.0 Hz to 14.0 Hz, and 14.3 Hz to 35.0 Hz at step 86 resulting in four (4) EEG power sums.

[0074] For each epoch, each of the four (4) EEG power sums is then compared to EEG power sum data generated from epochs of reference EEG records in a Frequency vs. Power look-up table (see Table 1 below) to determine a rank for each of the four (4) EEG power sums. As can be seen in Table 1 below, the Frequency vs. Power look-up table divides a full range of EEG power sums tabulated from epochs of reference EEG records for each of the four (4) frequency ranges into ten (10) equal ranks (deciles) numbered 0 to 9.

[0075] The Frequency vs. Power look-up table was constructed as follows. As discussed in above-incorporated U.S. Patent No. 9,763,589 and in Younes et al. Sleep. 2015 Apr 1;38(4):641-54. doi: 10.5665 / sleep.4588, fifty-six (56) polysomnograms recorded in laboratory from ambulatory subjects with normal sleep as well as assorted sleep disorders (sleep apnea, insomnia, periodic limb movements, narcolepsy) were used as the reference EEG records. These recorded polysomnograms provided > 400,000 epochs. Each of these epochs was subjected to the same Fourier Transform and EEG power calculations described above, and for each frequency range, the calculated EEG power sums were used to populate the full range of EEG power sums in Table 1 that are divided into the ten (10) equal ranks.

[0076] For each epoch, the four (4) ranks corresponding to each of the four (4) EEG power sums are then concatenated into a single 4-digit BIN code at step 88. Thus, for example, the BIN code 8250 represents an epoch with a high level (“8”) of EEG power values in the slowest frequency range (0.33 Hz to 2.33 Hz), a relatively low level (“2”) of EEG power values in the next higher frequency range (2.67 Hz to 6.33 Hz), a moderate level (“5”) of EEG power values in the next higher frequency range (7.33 Hz to 14.0 Hz) frequency range, and a very low level (“0”) of EEG power values in the highest frequency range (14.3 Hz to 35.0 Hz). As will be appreciated, up to this point, the steps performed as described above are similar to the initial steps described in the ORP patent.TABLE 1

[0077] Following generation of the BIN code for each of the epochs, the BIN codes are analyzed against natural sleep EEG data to determine whether the EEG pattern in each epoch represented by the BIN code represents a natural or unnatural brain state at step 90.

[0078] In this embodiment, a BIN Code vs. Frequency look-up table stored in the software is used during analysis of the BIN codes against the natural sleep EEG data. The BIN Code vs. Frequency table was constructed by analyzing EEG records obtained using the same recording specifications from 200 community-based subjects with no known brain disorder and not on any agents / drugs that alter brain state, selected randomly from 5800 subjects of the Sleep Heart Health Study available in the public domain. The BIN Code vs. Frequency table in this embodiment includes at least the median and 95thpercentile of frequency of occurrence of each of the 10,000 possible EEG patterns (10A4) in the 200 subjects, normalized per 10,000 analyzed epochs, to adjust for differences in the duration of the different EEG records. A representative section of the BIN Code vs. Frequency look-up table is shown below:

[0079] As can be seen in the above section of the BIN Code vs. Frequency lookup table, some BIN codes in the BIN Code vs. Frequency look-up table are well represented with a 95thpercentile up to nearly 100 per 10,000 epochs (highlighted), while other BIN codes in the BIN Code vs. Frequency look-up table do not appear atall (95thpercentile = 0) or appear very rarely despite analyzing > 3.5 million epochs in this reference population.

[0080] In particular, during the analyzing, the BIN codes generated for the epochs of the input EEG record are compared to the BIN Code vs. Frequency look-up table to determine whether the BIN codes and hence, the EEG patterns in the epochs are representative of natural or unnatural brain states (i.e whether or not the BIN codes are well represented in the BIN Code vs. Frequency look-up table or not). If EEG patterns in the epochs of the input EEG record, that are representative of unnatural brain states, are detected, their prevalence and / or reoccurrence can be recorded and a protocol initiated at step 74, which is dependent on the environment in which the input EEG record is captured.

[0081] With respect to the initiated protocol, a number of options are available. When the input EEG record is obtained in real time from a patient undergoing continuous intravenous sedation / anesthesia, detection of unnatural brain states can be used by the medical professional(s) managing the sedation / anesthesia, to adjust the rate of sedative / anesthetic infusion to minimize the prevalence of unnatural brain states. In this case, the input EEG record from the patient is processed in real time as calculation of the BIN code for each epoch of the input EEG record can be achieved in only a few milliseconds. The calculated BIN codes can then be compared to the BIN Code vs. Frequency look-up table to detect BIN codes representative of unnatural brain states and whether unnatural brain states are occurring at more than a specified prevalence (e.g., 0.1, 0.5, 2.0 per 10,000...etc.) in a preceding specified interval (e.g., 2 minutes, 5 minutes, ...etc.). The BIN codes representative of unnatural brain states and their prevalence / reoccurrence and / or other information representative of unnatural brain states can then be displayed to the medical professional(s) managing the sedation / anesthesia via monitor 56 or other suitable output interface, who can then adjust the rate of sedative / anesthetic infusion effectively in real time to minimize or reduce the prevalence / reoccurrence of unnatural brain states during sedation / anesthesia.

[0082] Alternatively, if the administration of sedatives / anesthetics to the patient is performed by an automated drug delivery system, the BIN codes representative of unnatural brain states and their prevalence / reoccurrence can be used to generate output to the automated drug delivery system so that, in response, the automated drug delivery system can adjust the rate of sedative / anesthetic infusion to minimize or reduce the prevalence / reoccurrence of unnatural brain states duringsedation / anesthesia.

[0083] Alternatively, rather than using the results of unnatural brain state detection in real time, the input EEG record can be processed post hoc and a report generated that contains the prevalence / reoccurrence of unnatural brain states detected and the times they were detected. These results can then be associated with the clinical status that existed at these times. This kind of output is more suited for post hoc analysis to calibrate results (i.e., relation between sedative dose or clinical neurological status and type and prevalence of unnatural brain states).

[0084] To collaborate the above, a study was conducted as described in the nonpatent reference entitled “Sedation-related Electroencephalographic Patterns in Acute Hypoxemic Respiratory Failure” authored by Rodrigues et al., Anesthesiology, November 2025, Volume 143, Pages 1266 to 1278, the content of which is incorporated herein by reference. In the study, the relevance of unnatural EEG patterns that never or rarely appear in sleep studies and their association with sedation at the early phase of acute hypoxemic respiratory failure (AHRF) were explored.

[0085] The study was a prospective cohort study including patients mechanically ventilated for AHRF and Pa02 / fraction of inspired oxygen less than 200 mmHg receiving various sedation-opioid regimens and doses as per clinical indication.Continuous EEG monitoring was performed for study inclusion until extubation, death, or up to seven (7) days. EEG quantified the relative power of each frequency band (slow delta, fast delta plus theta, alpha-sigma, beta) and determined the frequency of unnatural EEG patterns.

[0086] A total of 1 ,832h of EEG recordings were analyzed (mean ± SD, 43 ± 25h / patient) from 23 patients (median [interquartile range, 25 to 75%], 58 [48 to 70] yr; 87% male; Pa02 / fraction of inspired oxygen, 150 [116 to 198 mmHg; intensive care unit mortality, 22%). Unnatural EEG patterns accounted for 42% of the total recording time overall, differed among drug combinations, and exceeded 50% with some sedation-opioid combinations. Brief wake intrusions, a marker of physiologic sleep, were extremely low. Unnatural EEG patterns prevalence was higher with sedationopioid combinations (P < 0.029), high sedation dose (P < 0.035), and deeper clinical sedation score (P < 0.024), and was associated with intensive care unit mortality (P < 0.001). Continuous intravenous sedation results in unnatural EEG patterns that are not present in natural sleep, correlated with dose of sedation, clinical sedation score, and clinical outcomes.

[0087] For input EEG records obtained from patients not undergoing sedation / anesthesia, whether processed in real time or post hoc, detection ofunnatural brain states may be suggestive of the existence of a pathology allowing the results of input EEG record processing to be used to prompt investigative procedures.

[0088] Although the BIN code vs. Frequency look-up table described above includes frequency of occurrence data for median and 95thpercentile, those of skill in the art will appreciate that the look-up table may include frequency of occurrence data for additional or alternative percentiles.

[0089] Although uses for the detection of unnatural brain states are described above, once the BIN codes of the epochs of input EEG records are determined and analyzed to detect those BIN codes representative of natural and unnatural brain states, there are numerous options for expressing the results and this can be individualized to the preference of the user.

[0090] Although use of look-up tables has been described above, those of skill in the art will appreciate that the look-up tables may be replaced with mathematical models that can be used to determine the BIN codes and / or whether the BIN codes are representative of unnatural or natural EEG patterns.

[0091] Although embodiments have been described, those of skill in the art will appreciate that variations and modifications may be made without departing from the scope of the appended claims.

Claims

What is claimed is:

1. A method comprising:mathematically processing a specified section of an input electroencephalogram (EEG) record to extract features thereof; andcomparing the extracted features against natural sleep EEG data to detect unnatural brain states.

2. The method of claim 1 , wherein the extracted features are compared against the natural sleep EEG data to identify EEG patterns that never or rarely occur in natural sleep.

3. The method of claim 2, wherein the natural sleep EEG data is generated from EEG records of subjects during natural sleep and with no known brain disorder and not on any agents / drugs that alter brain state.

4. The method of any one of claims 1 to 3, further comprising determining the prevalence and / or reoccurrence of detected unnatural brain states.

5. The method of any one of claims 1 to 4, wherein the extracted features are spectral features of the input EEG record.

6. The method of any one of claims 1 to 5, wherein epochs of the specified section of the input EEG record are mathematically processed to extract features thereof.

7. The method of claim 6, wherein the epochs are continuous and nonoverlapping.

8. The method of claim 7, wherein the epochs are 3-second epochs.

9. The method of any one of claims 1 to 8, wherein the specified section of the input EEG record is the entire input EEG record or one or more portions of the input EEG record.

10. The method of claim 1 , wherein the extracted features are spectral features of the input EEG record and wherein the extracted features are analyzed to calculate power spectral density over specified frequency ranges.

11. The method of claim 10, wherein mathematically processing the specified section of the input EEG record comprises:performing frequency domain analysis on epochs of the specified section of the input EEG record to determine EEG power at specified frequencies;for each epoch, calculating EEG power over the specified frequency ranges;for each epoch, assigning, for each specified frequency range, a rank to the calculated EEG power, each rank being determined based on values of EEG power encountered in epochs of a plurality of reference EEG records;for each epoch, determining a code based on the assigned ranks; and analyzing the codes determined for the epochs against the natural sleep EEG data to detect codes representative of unnatural brain states.

12. The method of claim 11 , wherein the analyzing comprises comparing the codes determined for the epochs to frequency of code occurrence data generated from EEG records of subjects during natural sleep and with no known brain disorder and not on any agents / drugs that alter brain state.

13. The method of claim 12, wherein the frequency of code occurrence data generated from the EEG records of subjects populates a look-up table and includes at least one frequency of occurrence entry for each possible code.

14. The method of claim 13, wherein the at least one frequency of occurrence entry at least comprises median and 95thpercentile of frequency of occurrence entries for each possible code.

15. The method of any one of claims 11 to 14, wherein during the performing frequency domain analysis, EEG power is determined incrementally over the specified frequencies.

16. The method of claim 15, wherein the specified frequencies are in the range between 0.33 Hz to 60 Hz and EEG power is calculated at 0.33 Hz intervals over the range.

17. The method of claim 15 or 16, wherein the specified frequency ranges comprise four frequency ranges.

18. The method of claim 17, wherein the four frequency ranges comprise a 0.33 Hz to 2.33 Hz frequency range, a 2.67 Hz to 6.33 Hz frequency range, a 7.0 Hz to 14.0 Hz frequency range, and a 14.3 Hz to 35 Hz frequency range.

19. The method of any one of claims 15 to 18, wherein during the assigning, for each epoch, the EEG powers are summed over the specified frequency ranges and wherein the EEG power sums are compared to EEG power sum data generated from epochs of the plurality of reference EEG records to determine the rank.

20. The method of claim 19, wherein the EEG power sum data generated from epochs of the plurality of reference EEG records populates a look-up table that divides the range of EEG power sums for each frequency range into ten ranks.

21. The method of any one of claims 15 to 20, wherein, for each epoch, the ranks are concatenated to form the code with each value of the code representing the level of EEG power in a respective specified frequency range.

22. The method of any one of claims 1 to 21 , further comprising preprocessing the specified section of the input EEG record.

23. The method of claim 22, wherein the preprocessing comprises low-pass and high-pass filtering of the input EEG record.

24. The method of any one of claims 1 to 23, further comprising, upon detection of unnatural brain states, adjusting sedation / anesthesia of a patient or prompting an investigative procedure.

25. The method of claim 24, wherein sedation / anesthesia of the patient is adjusted manually or automatically.

26. The method of any one of claims 1 to 25, further comprising displaying and / or outputting information representative of detected unnatural brain states.

27. An apparatus comprising:memory embodying computer executable instructions; andat least one processor configured to communicate with said memory and to execute said instructions to cause said apparatus to:mathematically process a specified section of an input electroencephalogram (EEG) record to extract features thereof; and comparing the extracted features against natural sleep EEG data to detect unnatural brain states.

28. The apparatus of claim 27, wherein the apparatus is caused to compare the extracted features against the natural sleep EEG data to identify EEG patterns that never or rarely occur in natural sleep.

29. The apparatus of claim 28, wherein the natural sleep EEG data is generated from EEG records of subjects during natural sleep and with no known brain disorder and not on any agents / drugs that alter brain state.

30. The apparatus of any one of claims 27 to 29, wherein the apparatus is further caused to determine the prevalence and / or reoccurrence of detected unnatural brain states.

31. The apparatus of any one of claims 27 to 30, wherein the extracted features are spectral features of the input EEG record.

32. The apparatus of any one of claims 27 to 31 , wherein the apparatus is caused to mathematically process epochs of the specified section of the input EEG record to extract features thereof.

33. The method of claim 32, wherein the epochs are continuous and nonoverlapping.

34. The method of claim 33, wherein the epochs are 3-second epochs.

35. The apparatus of any one of claims 27 to 34, wherein the specified section of the input EEG record is the entire input EEG record one or more portions of the input EEG record.

36. The apparatus of claim 27, wherein the extracted features are spectral features of the input EEG record and wherein the extracted features are analyzed to calculate power spectral density over specified frequency ranges.

37. The apparatus of claim 36, wherein during mathematically processing the specified section of the input EEG record, the apparatus is caused to:perform frequency domain analysis on epochs of a specified section of an input EEG record to determine EEG power at specified frequencies;for each epoch, calculate EEG power over the specified frequency ranges;for each epoch, assign, for each specified frequency range, a rank to the calculated EEG power, each rank being determined based on values of EEG power encountered in epochs of a plurality of reference EEG records;for each epoch, determine a code based on the assigned ranks; and analyze the codes determined for the epochs against natural sleep EEG data to detect codes representative of unnatural brain states.

38. The apparatus of claim 37, wherein during the analyzing, the apparatus is caused to compare the codes determined for the epochs to frequency of code occurrence data generated from EEG records of subjects during natural sleep and with no known brain disorder and not on any agents / drugs that alter brain state.

39. The apparatus of claim 38, wherein the frequency of code occurrence data generated from the EEG records of subjects populates a look-up table and includes at least one frequency of occurrence entry for each possible code.

40. The apparatus of claim 39, wherein the at least one frequency of occurrence entry at least comprises median and 95thpercentile of frequency of occurrence entries for each possible code.

41. The apparatus of any one of claims 37 to 40, wherein during the performing frequency domain analysis, the apparatus is caused to determine EEG power incrementally over the specified frequencies.

42. The apparatus of claim 41 , wherein the specified frequencies are in the range between 0.33 Hz to 60 Hz and EEG power is calculated at 0.33 Hz intervals over the range.

43. The apparatus of claim 41 or 42, wherein the specified frequency ranges comprise four frequency ranges.

44. The apparatus of claim 43, wherein the four frequency ranges comprise a 0.33 Hz to 2.33 Hz frequency range, a 2.67 Hz to 6.33 Hz frequency range, a 7.0 Hz to 14.0 Hz frequency range, and a 14.3 Hz to 35 Hz frequency range.

45. The apparatus of any one of claims 41 to 44, wherein during the assigning, the apparatus is caused to, for each epoch, sum the EEG powers over the specified frequency ranges and compare the EEG power sums to EEG power sum data generated from epochs of the plurality of reference EEG records to determine the rank.

46. The apparatus of claim 45, wherein the EEG power sum data generated from epochs of the plurality of reference EEG records populates a look-up table that divides the range of EEG power sums for each frequency range into ten ranks.

47. The apparatus of any one of claims 41 to 46, wherein, for each epoch, the apparatus is caused to concatenate the ranks to form the code with each value of the code representing the level of EEG power in a respective specified frequency range.

48. The apparatus of any one of claims 27 to 47, wherein the apparatus is caused to preprocess the specified section of the input EEG record.

49. The apparatus of claim 48, wherein during preprocessing, the apparatus is caused to low-pass and high-pass filter the input EEG record.

50. The apparatus of any one of claims 27 to 49, further comprising a display device and wherein the apparatus is further caused to display information representative of unnatural brain states on the display device for use in adjusting sedation / anesthesia or prompting an investigative procedure.

51. The apparatus of any one of claims 27 to 50, wherein the apparatus is caused to provide output to a drug delivery system for use in adjustingsedation / anesthesia.

52. An apparatus for distinguishing unnatural from natural patterns of electroencephalographic (EEG) activity comprising:memory embodying computer executable instructions; andat least one processor configured to communicate with said memory and to execute said instructions to cause said apparatus to carry out the method of any one of claims 1 to 26.

53. A non-transitory computer-readable medium embodying executable program instructions, which when executed by at least one processor, cause an apparatus to carry out the method of any one of claims 1 to 26.

54. A computer program comprising executable instructions, which when executed by at least one processor, cause an apparatus to carry out the method of any one of claims 1 to 26.

Images