System, device and method for determining brain state from cranial movement due to cerebral blood flow

Abstract

A headset configured to be attached to a human head includes an accelerometer that provides a signal indicative of head acceleration due to cerebral blood flow. An analyzer evaluates a plurality of samples representing a time series of acceleration, each sample corresponding to head movement due to cardiac contraction. The analyzer identifies a brain state based at least in part on the chaos level of the plurality of samples. An algorithm applied by the analyzer is formulated in part based on clinical data and testing of a plurality of subjects. In one embodiment, the plurality of samples evaluated by the analyzer are indicative of head acceleration in only one axis. In some circumstances, the analyzer further identifies a brain state based on a concurrent neurological test of the subject.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of priority to U.S. Provisional Patent Application Serial No. 62 / 793,767, filed January 17, 2019, entitled "SYSTEMS, DEVICES, AND METHODS FOR IDENTIFYING BRAIN STATE FROM CRANIAL MOVEMENT DUE TO CEREBRAL BLOOD FLOW," which is incorporated herein by reference in its entirety. [Technical field]

[0002] The present invention relates generally to medical devices, and more specifically to systems, devices and methods for determining brain status from skull movement due to blood flow in the brain. [Background technology]

[0003] Ballistocardiography is a method of measuring the movement of the human body in response to the heartbeat (cardiac contraction). The process measures the movement of the whole body. The heartbeat generates forces on the head and neck via the cerebral vasculature. The forces on the brain generated by the heartbeat are transmitted to the skull. Cranial accelerometry is a method used to measure the forces on the head and neck generated by the heartbeat forces. Cranial accelerometry has been used to estimate cerebral vasospasm after subarachnoid hemorrhage and to measure the biomechanics of the brain after concussion. Summary of the Invention

[0004] The present disclosure provides a system including an accelerometer, a fixation mechanism configured to fix the accelerometer to a human head including a brain such that the accelerometer is configured to generate an accelerometer signal indicative of head movement due to cerebral blood flow caused by heartbeats, and an analyzer configured to receive a plurality of data samples based on the accelerometer signal, each data sample indicative of head movement associated with a different heartbeat in an axis, and to measure a chaos level of the plurality of data samples and indicate that a large vessel occlusion (LVO) has occurred in the brain if the chaos level exceeds a threshold. In one embodiment, the accelerometer is a three-dimensional accelerometer capable of providing axial accelerometer signals in any of three orthogonal axes. In a further embodiment, the analyzer input is configured to receive another plurality of data samples based on an axial accelerometer signal indicative of head movement in another axis orthogonal to the one axis, each data sample of the another plurality of data samples indicative of head movement in conjunction with a different heartbeat in the another axis, and the analyzer is configured to measure a second chaos level of the second plurality of data samples and indicate that a large vessel occlusion (LVO) has occurred in the brain if the second chaos level exceeds a second threshold. In yet another embodiment, the analyzer input is configured to receive a third plurality of data samples based on a third axial accelerometer signal indicative of head movement in a third axis orthogonal to the one axis, each data sample of the third plurality of data samples indicative of head movement in conjunction with a different heartbeat in the third axis, and the analyzer is configured to measure a third chaos level of the third plurality of data samples and indicate that a large vessel occlusion (LVO) has occurred in the brain if the third chaos level exceeds a third threshold. In another embodiment, the chaos level is based at least in part on a difference between signal levels at a plurality of selected times for each of a plurality of acceleration versus time functions of the plurality of data samples. In a further embodiment, the chaos level is based at least in part on a calculation of the plurality of acceleration versus time functions. In yet another embodiment, the calculation is based on a difference function of each data sample, each difference function comprising the difference between the data sample and an average of all data samples.) a number of times. In yet another embodiment, the calculation further comprises modifying and summing each difference function to produce a number of results; summing the number of results and producing a sum of the results; and normalizing the sum of the results to a number of data samples. In another embodiment, the calculation comprises calculating the root mean square of the difference functions. In a further embodiment, the calculation further comprises summing the root mean square of the difference functions to produce a sum of the results; and normalizing the sum to a minimum and maximum average of the data samples. In another embodiment, the calculation comprises modifying and summing each difference function to produce a number of results; summing the number of results and producing a sum of the results; and normalizing the sum to a minimum and maximum average of the data samples. In another embodiment, the calculation comprises applying a correlation function for each data sample to an average of all data samples. In yet another embodiment, the calculation further comprises calculating a number of times different from the average by increasing a threshold. In yet another embodiment, the processor is configured to measure a chaos level based at least in part on a shape of the acceleration versus time function of each data sample. In another embodiment, the analyzer is configured to align the plurality of acceleration versus time functions in a time series based on information of the plurality of acceleration versus time functions. In yet another embodiment, the analyzer is configured to align the plurality of acceleration versus time plots in a time series based at least in part on an electrocardiogram (ECG) signal indicative of cardiac contractions. In another embodiment, the analyzer is configured to align the plurality of acceleration versus time plots in a time series based at least in part on a photoplethysmogram (PPG) signal indicative of cardiac contractions. In another embodiment, the system may further include a transmitter configured to transmit a wireless signal indicative of the plurality of data samples; and a receiver configured to receive the wireless signal and provide the data samples to the analyzer. In a further embodiment, the system may further include an analog-to-digital converter (ADC) configured to sample the accelerometer signal and generate a plurality of data samples.In yet another embodiment, the system may further include a controller configured to chronologically align the plurality of data samples prior to transmitting the wireless signal including the chronologically aligned data samples. In another embodiment, the analyzer is configured to diagnose the presence or absence of an LVO condition based at least in part on neurological examination information.

[0005] The present disclosure also provides a system comprising: a three-dimensional accelerometer configured to generate three orthogonal signals, each orthogonal signal indicative of an acceleration of a movement of the three-dimensional accelerometer in an axis orthogonal to the other two axes; a fixation mechanism configured to fix the three-dimensional accelerometer to a human head including a brain; and an input configured to receive three sets of data samples, each data sample being based on one of the orthogonal signals and indicative of a movement of the head in conjunction with a different cardiac contraction in one axis, the system including an analyzer configured to measure a chaos level of each set of data samples and indicate that a large vessel occlusion (LVO) has occurred in the brain when the chaos level of at least one set of data samples exceeds a threshold value.

[0006] The present disclosure also provides an apparatus comprising: an input configured to receive a plurality of data samples, each data sample indicative of head movement due to different cardiac contraction-induced blood flow in the brain in one axis; a processor configured to measure a chaos level of the plurality of data samples aligned in time with respect to the cardiac contractions; and an output configured to indicate that a large vessel occlusion (LVO) has occurred in the brain if the chaos level exceeds a threshold. In one embodiment, the accelerometer signal samples are indicative of the head movement in only one axis. In another embodiment, the accelerometer signal samples are indicative of an output of an accelerometer fixed to the head. In a further embodiment, the accelerometer is a three-dimensional accelerometer capable of providing an axial accelerometer signal in any of three orthogonal axes. In another embodiment, the chaos level is based at least in part on a difference between signal levels at a plurality of selected times for each of a plurality of acceleration versus time plots of the plurality of accelerometer signal samples. In a further embodiment, the chaos level is based at least in part on a calculation of a plurality of acceleration versus time functions. In yet another embodiment, the calculation includes calculating a difference function for each data sample multiple times, each difference function including the difference between the data sample and an average of all data samples. In yet another embodiment, the calculation further includes modifying and summing each difference function to calculate a plurality of results; summing the plurality of results to calculate a sum of results; and normalizing the sum of results to a number of data samples. In another embodiment, the calculation includes calculating the root mean square of the difference function. In a further embodiment, the calculation further includes summing the root mean square of the difference functions to calculate a sum of results; and normalizing the sum of results to a minimum and maximum of the average of the data samples. In another embodiment, the calculation includes modifying and summing each difference function to calculate a plurality of results; summing the plurality of results to calculate a sum of results; and normalizing the sum of results to a minimum and maximum of the average of the data samples. In another embodiment, the calculation includes applying a correlation function for each data sample to an average of all data samples.In yet another embodiment, the calculation includes determining a number of times that differ from an average by increasing a threshold. In another embodiment, the chaos level is based at least in part on a shape of an acceleration versus time plot of each accelerometer signal sample. In yet another embodiment, a relative timing relationship between the plurality of acceleration versus time plots is determined from the plurality of acceleration versus time plots. In yet another embodiment, the analyzer is configured to diagnose the presence or absence of an LVO condition based at least in part on neurological examination information.

[0007] The present disclosure provides an apparatus comprising an accelerometer configured to generate an accelerometer signal; a fixation mechanism configured to fix the accelerometer to a human head including a brain; and a communication interface configured to transmit an analysis signal to the analyzer, the analysis signal being based on an accelerometer signal from the accelerometer indicative of head movement due to cerebral blood flow caused by cardiac contractions and based on a plurality of accelerometer signal samples, each accelerometer signal sample corresponding to a different cardiac contraction and indicative of head movement in only one axis. In one embodiment, the accelerometer is a three-dimensional accelerometer capable of providing an axial accelerometer signal in any of three orthogonal axes. In yet another embodiment, the apparatus may further comprise an analog-to-digital converter (ADC) configured to sample the accelerometer signal and generate a plurality of data samples. In another embodiment, the apparatus may further comprise a controller configured to chronologically align the plurality of data samples prior to transmitting the wireless signal including the chronologically aligned data samples.

[0008] The present disclosure also provides a method comprising receiving a plurality of data samples, each data sample indicative of head movement due to different cardiac contraction-induced blood flow in the brain in one axis; measuring a chaos level of the plurality of data samples aligned in time with respect to the cardiac contractions; and indicating that a large vessel occlusion (LVO) has occurred in the brain if the chaos level exceeds a threshold. In one embodiment, the accelerometer signal samples are indicative of the head movement in only one axis. In another embodiment, the accelerometer signal samples are indicative of an output of an accelerometer fixed to the head. In a further embodiment, the accelerometer is a three-dimensional accelerometer capable of providing an axial accelerometer signal in any of three orthogonal axes. In another embodiment, measuring the chaos level comprises measuring a difference between signal levels at a plurality of selected times for each of a plurality of acceleration versus time plots of the plurality of accelerometer signal samples. In a further embodiment, measuring the chaos level comprises calculating the chaos level from a plurality of acceleration versus time functions. In yet another embodiment, calculating the chaos level comprises calculating a difference function for each data sample multiple times, each difference function comprising the difference between the data sample and an average of all data samples. In yet another embodiment, calculating the chaos level further comprises modifying and summing each difference function to produce a plurality of results; summing the plurality of results to produce a sum of results; and normalizing the sum of results to a number of data samples. In another embodiment, calculating the chaos level further comprises calculating the root mean square of the difference functions. In a further embodiment, calculating the chaos level further comprises summing the root mean square of the difference functions to produce a sum of results; and normalizing the sum of results to an average minimum and maximum of the data samples. In another embodiment, calculating the chaos level further comprises modifying and summing each difference function to produce a plurality of results; summing the plurality of results to produce a sum of results; and normalizing the sum of results to an average minimum and maximum of the data samples.In another embodiment, calculating the chaos level includes applying a correlation function for each data sample to an average of all data samples. In yet another embodiment, calculating the chaos level includes determining a number of times that differ from the average by increasing a threshold. In another embodiment, measuring the chaos level includes evaluating a shape of an acceleration versus time plot of each accelerometer signal sample. In yet another embodiment, the method further includes arranging the plurality of data samples in a time series based on information measured from the plurality of data samples. In yet another embodiment, the method further includes receiving neurological examination information; and diagnosing the presence or absence of an LVO condition based on the neurological examination information and the chaos level.

[0009] The present disclosure also provides a system including an accelerometer; a fixation mechanism configured to fix the accelerometer to a human head including a brain such that the accelerometer is configured to generate an accelerometer signal indicative of head movement due to cerebral blood flow caused by cardiac contractions; and an analyzer having a first input configured to receive a plurality of data samples based on the accelerometer signal, each data sample indicative of head movement associated with a different cardiac contraction in an axis, and a second input configured to receive neurological examination information from patients, including patients with large vessel occlusion (LVO) and patients without LVO. [Brief description of the drawings]

[0010] [Figure 1] 1 is an illustration of an example of a system for observing head acceleration and assessing brain state.

[0011] [Figure 2A] FIG. 2 is a block diagram of an example in which a headset includes a communication interface.

[0012] [Figure 2B] FIG. 1 is a block diagram of an example of a headset comprising an analyzer and a user interface.

[0013] [Diagram 3] 1 is a graph showing two sets of acceleration signals from one axis of an accelerometer over time.

[0014] [Figure 4] 1 is an illustration of an example of chaos in a cranial accelerometer during multiple cardiac cycles for three different conditions.

[0015] [Diagram 5] 1 is an illustration of an example of a method for measuring chaos in a cranial accelerometer during a heartbeat.

[0016] [Figure 6] 1 is a flowchart of an example of a method for identifying a brain state.

[0017] [Figure 7] 1 illustrates two exemplary data samples for head pulse recording.

[0018] [Figure 8] 1 is an illustration of an example of data samples representing all head pulse recording data samples.

[0019] [Figure 9] The graph shows the number of data samples on the y-axis and the number of data points outside one standard deviation on the x-axis.

[0020] [Figure 10] 1 shows an average of data samples from a head pulse recording and a graph of one of the data samples.

[0021] [Figure 11] 2 shows a graph of the average absolute value and its integral of data samples from a head pulse recording.

[0022] [Figure 12]Graphs of one data sample and the average over the entire set of data samples from the head pulse recording are shown. Detailed Description of the Invention

[0023] As used in this specification and the appended claims, the singular forms "a," "an," and "this" include plural referents unless the context clearly dictates otherwise. Thus, for example, reference to "an accelerometer" includes a plurality of such accelerometers, reference to "an interface" includes reference to one or more interfaces, and so forth.

[0024] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. Although methods and materials similar or equivalent to those described herein can be used in the practice of the disclosed methods and compositions, exemplary methods, devices, and materials are described herein.

[0025] Additionally, the use of "or" means "and / or" unless otherwise stated. Similarly, "include," "including," "including," and "including" are interchangeable and are not intended to be limiting.

[0026] It should be further understood that when the term "comprising" is used in describing various embodiments, those skilled in the art will recognize that in some specific instances, the embodiments can be alternatively described as "consisting essentially of" or "consisting of."

[0027] The accelerometer data versus time for multiple heart beats is called a head pulse record. A heart beat is the time of one cardiac cycle and corresponds to the RR interval from one heart beat to the next (the ECG QRS peak represents the systolic peak). The heart beats can be used to analyze the head pulse record into multiple data samples. Each data sample represents the accelerometer output versus time for one heart beat.

[0028] Cranial accelerometry is a method used to measure the forces on the head and neck generated by the force of the heartbeat. Cranial accelerometry has been used to estimate cerebral vasospasm after subarachnoid hemorrhage (Smith et al., Neurocrit Care, December 2015; 23(3):364-9) and to measure brain biomechanics after concussion (Auerbach et al., Clin J Sport Med., March 2015; 25(2):126-3).

[0029] Acute ischemic stroke is a common and often fatal disease worldwide. Acute stroke is the leading cause of death in China and the fifth leading cause of death in the United States. Approximately 85% of all strokes in the United States are ischemic, defined as a localized reduction in blood flow to the brain due to occlusion of an arterial blood vessel. Stroke forms can be divided into large vessel occlusion (LVO) stroke and small vessel occlusion (SVO) stroke. The optimal treatment for SVO stroke is intravenous infusion of tissue plasminogen activator within the first 4.5 hours of stroke onset. The optimal treatment for LVO stroke is intravenous infusion of tissue plasminogen activator combined with subsequent endovascular thrombectomy (EVT). EVT involves placing an arterial catheter in the body, introducing a thrombectomy device through the catheter into the occluded cerebral artery, and withdrawing the device along with the thrombus, opening the occluded cerebral blood vessel.

[0030] EVT is currently the standard of care for LVO in many developed countries. EVT can only be performed in medical centers that have specialized radiological equipment and trained staff and physicians to perform the procedure. These centers are called comprehensive stroke centers (CSCs) or thrombectomy capable centers (TCCs). They are distinct from primary stroke centers (PSCs), which are hospitals that cannot perform EVT but can treat patients with intravenous infusion of tissue plasminogen activator. Currently, there are 1110 PSCs and only 191 CSCs / TCCs in the United States. This means that only 17% of acute stroke centers in the United States can handle LVO patients. Once an LVO patient is brought to a PSC, the patient must be transported to a CSC / TCC. The transport time exacerbates the patient's brain damage.

[0031] The time from the onset of stroke symptoms (when the thrombus occludes a large blood vessel) to successful thrombectomy directly determines the clinical prognosis of stroke patients: the longer it takes to start carrying out said treatment, the worse the patient's situation will be, and the sooner said treatment is carried out, the more likely the patient will be cured.

[0032] The time to successful EVT depends primarily on the time it takes to transport the patient from where they were found to a CSC / TCC. To reduce morbidity, EVT is often performed within minutes of the patient's arrival at the CSC / TCC. However, transport time to a CSC / TCC varies widely depending on the local prehospital triage process.

[0033] Prehospital providers would be overwhelmed if they directed all suspected stroke patients to CSCs / TCCs, whereas directing suspected stroke patients to the nearest stroke hospital (usually a PSC) would require secondary transport from the PSC to a CSC / TCC, risking delays in EVT.

[0034] One strategy to maximize efficiency and improve stroke outcomes is to have LVO patients transferred to CSCs / TCCs and non-LVO patients transferred to PSCs. However, determining whether a patient has an LVO stroke before admission is problematic.

[0035] The standard of care to screen for LVO involves performing brain imaging to image the patency of intracranial vessels, most commonly computed tomography angiography (CTA), which requires a CT scanner. In some municipalities in the United States, CT scanners are deployed in ambulances, but this is not practical because of the cost and the lack of a physician available to interpret the images and communicate with the emergency department.

[0036] When performed by a neurologist or other stroke-trained medical professional, the accuracy of performing the neurological exam and interpreting the signs and symptoms is approximately 80%. This accuracy is not achieved when performed by prehospital medical professionals. Thus, even if a neurologist is present at the time of clinical response (either telemedicine or in person), approximately 20% of patients are incorrectly triaged.

[0037] In addition to using a portable CT scanner, biometrics that measure changes in the human brain during LVO can be combined with the above neurological examination components to achieve greater accuracy in detecting LVO.

[0038] The forces generated by the heart are transmitted to the human skull along large arterial branches, including the carotid and vertebral arteries. The minute head movements generated by the transmission of these forces are called head pulses. A head pulse recording is a single or set of waveforms over time that can be transmitted by devices sensitive to position, velocity, force, or action potentials, including accelerometers, action potential detectors, velocity sensors, system statistics, etc. These sensors are placed on the patient's head that can transmit the head pulse. These transmitters generate electrical signals that are recorded in analog form or converted to digital form and then displayed on an oscilloscope or digital monitor. The head pulse recording can be obtained in a healthy subject and used as a reference standard. Head pulses can be obtained in patients with various conditions and compared to head pulse characteristics of normal patients to measure specific waveform characteristics of disease processes. Although the examples discussed below are directed to the examination of LVO, the techniques discussed herein may be applied to other disease processes, including, for example, acute ischemic stroke, intracerebral hemorrhage, concussion, traumatic brain injury, cerebral edema, and prognosis after cardiac arrest. The technology may also be applied to the assessment of clinical conditions during anesthesia and sleep. The head pulse technology can be modified from known standards to diagnose certain neurodegenerative conditions including Alzheimer's disease, frontotemporal dementia, Parkinson's disease, synucleinopathies, progressive supranuclear palsy, prion diseases, multiple system atrophy, and other neurodegenerative diseases that cause brain atrophy or protein deposition in the atrophic brain.

[0039] Measurement of head pulses and analysis of the resulting waveform data in combination with several elements of the neurological examination are methods to improve the accuracy of the diagnosis of LVO stroke and therefore the decision to transfer a stroke patient to a center offering EVT.

[0040] By constructing a headset that contacts the human head and houses action potential, acceleration, velocity, and / or position sensors while recording the time of concurrent heartbeats, the head pulses can be acquired and analyzed. The head pulse recordings are then sent to a computer algorithm that incorporates these signals and concurrent clinical data (e.g., level of consciousness, limb strength, etc.) to instantly and probabilistically predict the presence or absence of disease (e.g., acute LVO stroke) or clinical state (e.g., sleep or anesthesia) and make that information available to a trained user of the device for clinical decision making.

[0041] In the examples discussed herein, a headset configured to be attached to a human head comprises at least one three-axis accelerometer providing a signal indicative of head motion due to cerebral blood flow. An analyzer evaluates a plurality of samples indicative of acceleration over time, each sample corresponding to a head acceleration due to a heartbeat. The analyzer identifies a brain state using an algorithm based at least in part on the head acceleration based at least in part on the chaos level of the plurality of samples. The algorithm applied to the analyzer is formulated in part based on clinical data, the clinical data being classified based on a cohort of subjects with or without LVO. In one embodiment, the plurality of samples evaluated by the analyzer are indicative of head motion in only one axis. In some circumstances, the analyzer further identifies a brain state based on a concurrent neurological examination of the subject.

[0042] FIG. 1 is an illustration of an example of a system 100 for observing head acceleration and assessing brain conditions. In the embodiments discussed herein, the system 100 diagnoses, or at least assesses the probability of, whether a large vessel occlusion (LVO) has occurred in the brain of a subject. Although the following embodiments are directed to diagnosing whether an LVO condition has led to a stroke in a subject, the devices and techniques described herein are applicable to other diagnostic and prognostic situations. In general, the analyzer evaluates a number of data samples indicative of head acceleration due to multiple heartbeats and measures the chaos level of the data samples. In the following embodiments, the data samples are digital representations of an accelerometer output signal indicative of acceleration as a function of time for each heartbeat. The chaos level is measured based on evaluation of the digital representations that are aligned in time with respect to the heartbeat. As discussed herein, chaos is a subjective term used to describe the lack of repetitively correlated accelerometer data samples. Chaos can be numerically defined by several measures, most often based on the dissimilarity between any single measurement after a heartbeat and an estimate of the average of all measurements from a head pulse recording.

[0043] In the embodiment of FIG. 1, the system 100 includes an accelerometer 102, a fixation mechanism 104, and an analyzer 106. In some situations, the analyzer 106 is part of a headset 108 that is fixed to the subject's head 110. As described below, in one embodiment, the analyzer 106 may be separate from the headset 108. Also, a communication interface 112 may transmit information to the analyzer 106. In other embodiments, data is stored in a memory in the headset 108 and transferred to the analyzer 106 via a wired or wireless connection. In other situations, data is stored in a removable memory (e.g., memory card, flash card, memory cartridge, memory stick, etc.) of the headset 108 and moved to the analyzer 106 for transfer. In situations where the analyzer is implemented with a processor in the headset, a separate memory device may be omitted. Thus, if there is no memory device in the headset, data may be analyzed by a processor in the headset. If the headset has a memory device, the data may be stored in the memory device within the headset and analyzed by a processor external to the headset and the data may be erased mechanically, or the data may be stored in the memory device within the headset and transmitted wirelessly to a processor external to the headset.

[0044] In some circumstances, such as when the analyzer 106 is part of the headset 108, the communications interface 112 may be omitted. To indicate this, the communications interface 112 is shown with a dashed line. The various functions and operations in the blocks described with reference to the system 100 may be implemented in any number of devices, circuits, or elements. The functions in two or more of the blocks shown in this figure may be combined into one device, and the functions described as being performed by any one device may be implemented by several devices or elements.

[0045] As is well known, heartbeat generates forces on the head and neck through the cerebral vasculature. The forces on the brain tissue generated by the heartbeat are translated to the skull. Thus, the subject's head 110 moves in response to each beat 114, 115, and 116 of the heart 118 as blood is pumped through the brain 120. In the embodiment herein, the acceleration 122 is detected and evaluated by the accelerometer 102 to diagnose whether the subject has experienced a large vessel occlusion (LVO). In one technique, several signal samples are collected, each corresponding to a heartbeat 114, 115, and 116. The analyzer 106 evaluates the signal samples to measure the chaos level of the collected signal samples. In general, a higher chaos level increases the likelihood of a stroke due to LVO. However, based on clinical data and head movement data of multiple subjects, improved algorithms can be developed to improve the accuracy of diagnosing LVO. As described in more detail below, any of a number of evaluation, computation, or signal processing techniques can be used to measure the chaos level. The chaos level may be expressed as a number that is compared to a threshold value, and a value above the threshold value is determined to indicate that an LVO stroke has occurred. In embodiments herein, the algorithm is applied to the data based in part on neurological examination and clinical data from multiple subjects for algorithm refinement. In the embodiment of FIG. 1, the analyzer 106 also evaluates concurrent examination and clinical data 123 to diagnose the presence or absence of an LVO condition. Thus, at least in some circumstances, one person performs the examination of the subject and inputs the results into the analyzer 106. Thus, the analyzer may include or be connected to input devices, such as, for example, a keyboard, mouse, touchpad, microphone, touchscreen, etc. The analyzer may also include or be connected to output devices, such as a visual display and a speaker.

[0046] The accelerometer 102 detects acceleration in at least one direction (axis) and generates an accelerometer signal based on head acceleration. In one embodiment, the accelerometer is a three-axis accelerometer, sometimes referred to as a three-dimensional (3D) accelerometer, which generates three signals. Each signal corresponds to acceleration in one of three axes (X, Y, and Z) 124, 125, and 126. Each axis is orthogonal to the other two axes. Thus, the three signals generated by the accelerometer are also orthogonal.

[0047] In some situations, signal samples from only one axis are collected and analyzed. In other situations, the chaos of a signal sample from one axis may be measured, the chaos of a signal sample from another axis may be evaluated, and the two chaos values ​​may then be evaluated to determine the likelihood of an LVO condition. For example, signal samples from each of three axes may be evaluated separately to measure the chaos of each axis, and the resulting values ​​may be further processed to measure the probability of an LVO condition. In other situations, two or more chaos values ​​may be averaged and then compared to a threshold. Thus, various techniques may be used to evaluate data from multiple axes.

[0048] In some situations, the headset 108 may include multiple accelerometers. Data from each accelerometer may be evaluated independently to measure the chaos of each axis of each accelerometer, and the resulting values ​​may be further processed to measure the probability of an LVO condition. In other situations, data from multiple axes from each accelerometer may be processed, and the resulting evaluation data may be further processed with evaluation data from other accelerometers. Thus, data from any number of axes and accelerometers may be processed and evaluated in numerous ways.

[0049] A power source 128 provides power to the components of the headset 108. In the embodiments herein, the power source 128 is a battery within the headset 108, which may include one or more cells. The battery may be disposable or rechargeable. If rechargeable, the headset may include circuitry and / or a charging interface to facilitate charging the battery. In some circumstances, a power source may be external to the headset. For example, an AC-DC power source or an external battery pack may be connected to the headset. The charging interface may also be an interface that facilitates wireless charging.

[0050] 2A is a block diagram of an example of the headset 108 including a communication interface 112. In operation, the headset 108 is placed on the subject's head 110 by the fixation mechanism 104. The accelerometer 102 detects head acceleration in at least one direction (axis) for several heartbeats and generates a number of signal samples 202-205, each signal sample associated with a heartbeat. The communication interface 112 transmits data 208 corresponding to the signal samples 202-205 to the analyzer 106. In this example, the data 208 is a digital representation of the signal samples 202-205 and is received by a communication interface 210 connected to or part of the analyzer 106.

[0051] Thus, data 208 in the example includes data samples 214-217 indicative of acceleration as a function of time, as measured by an accelerometer. The data samples may be a plot, graph, table, list of associated values, data string, or other representation of the acceleration versus time function of the accelerometer output. The data samples 214-217 may in some circumstances include other data related to acceleration and time functions. For example, the data 208 may indicate position as a function of time or velocity as a function of time. In the example described below, the data samples are representations of acceleration versus time that can be processed to measure chaos levels across multiple samples. In the example described below, the head movement is conveyed by a three-axis accelerometer, and the analog signal from each axis is digitized for processing by the analyzer. Such a time series of digitized data is called a head pulse recording.

[0052] Data representing head acceleration (or other data indicative of movement) can be communicated in any of several ways, depending on the particular implementation. In the embodiment of FIG. 2, the analog output of the accelerometer is periodically sent to an analyzer and converted to digital data indicative of acceleration over time for several heartbeats. In this embodiment, a controller 220 directs a parallel analog-to-digital converter (ADC) 222 to convert the periodic analog head pulse records generated by the accelerometer 102. The digital data is stored in a memory unit 224. Examples of suitable ADCs and memory units include a 24-bit ADC that simultaneously converts the analog signals to digital form and stores the data in a MicroSD memory via a serial peripheral interface (SPI). In this embodiment, signals from an electrocardiogram (ECG) 226 and a light-emitting diode (LED) sensor 228 that generates a photoplethysmogram (PPG) are used to provide a time reference to the acceleration signal. As is well known, a PPG is a waveform that resembles an arterial pulse. The ECG signal and / or the PPG signal may be omitted. In some situations, for example, the timing of the samples can be derived from the accelerometer waveform or the difference between data samples acquired by opposite head positions. This is a scalp plethysmography measurement. Other techniques can be used to obtain heartbeat timing information. For example, a sensor can be placed in an artery, such as placing an accelerometer in the carotid artery. Thus, the timing of the heartbeat can be obtained or derived from an ECG signal, a PPG signal, the head pulse recording, a sensor on an artery, or any combination thereof.

[0053] The controller 220 is any controller, processor, circuitry, logic circuitry, processing circuitry, or processor arrangement that manages the functions described herein and facilitates the overall functioning of the headset. An example of a suitable controller 220 is a microprocessor and supporting circuitry that operates according to an nRF 52 Cortex M-4 processor architecture.

[0054] Data 208 is retrieved from memory 224 and communicated to the analyzer 106. The communication interface 112 in the headset 108 includes at least a transmitter 230. And the communication interface 210 connected to the analyzer 106 includes at least a receiver 232. Signals between the two communication interfaces can be communicated via wires or cables, but in this embodiment a wireless interface is used. The communication interfaces 112 and 210 in this embodiment operate according to the Bluetooth communication standard. Other suitable examples of wireless communication include technologies according to the Wi-Fi and ZigBee communication standards. In some situations, the communication interfaces 112 and 210 may operate according to a proprietary wireless interface or protocol. In still other situations an optical interface can be used.

[0055] Thus, a transmitter 230 in the communication interface 112 transmits digital data 208 representing the captured samples 202-205 to a receiver 212 in a communication interface 210 connected to the analyzer 106. In some circumstances, the data is stored in a memory in the headset and transferred to the analyzer when connected to the analyzer 106 via a wired connection, or transmitted via a wireless interface. In other embodiments, the analog output of the accelerometer is converted to digital data that is transmitted to the analyzer in real time. Thus, in at least some embodiments, the digital data 208 is wirelessly transmitted to the analyzer 106, where the digital data 208 represents the level of acceleration as a function of time over several heartbeats. The resulting acceleration versus time data samples 214-217 for each heartbeat corresponding to the analog samples 202-205 are evaluated together in the analyzer 106. Briefly and clearly, FIG. 2 shows four data samples 214-217 representing four analog signal samples 202-205. However, any number of signal samples and / or data samples may be captured and evaluated. The collection of data samples may be referred to as a head pulse recording.

[0056] The analyzer 106 is any computer, processor, server, or other device that executes code, applies the algorithm 234 to the data samples 214-217, and displays the presence or absence of an LVO condition to the subject. In one embodiment, the analyzer 106 is a local desktop or laptop computer that executes a program. In other situations, the analyzer 106 is an application running on a server. Thus, the analyzer 106 may be implemented in the "cloud" where the data 208 is provided over a network, such as the Internet, an intranet, or other private communication network. In yet another embodiment, the analyzer may comprise an application that runs on a smartphone or tablet. In yet another embodiment, the algorithm may be embedded in the microprocessor 220 and generate an output via a series of lights on the headset. The algorithm 234 may be based on several comparisons, evaluations, and mathematical calculations, may be adjusted, modified, improved, or may be based on laboratory and clinical data 123. In the example discussed herein, the algorithm 234 applied by the analyzer 106 in the system 100 is refined and is based in part on the neurological exam and clinical data 123. In this refinement of the algorithm, the system 100 includes the same components as used in the field and is used to acquire data from several subjects. The analyzer 106 applies the algorithm 234 and diagnoses a binary or graded result indicating the presence or absence of an LVO condition in the subject. The refinement of the algorithm 234 is based on comparing the prediction of LVO by the algorithm with the presence or absence of LVO diagnosed by actual angiography. For example, if the result indicates the absence of an LVO condition and a reference result such as a computed tomography angiogram (CTA) indicates the presence of an LVO condition, the algorithm is adjusted so that the result for the same data set indicates the presence of an LVO condition.Examples of neurological exam and clinical data suitable for tuning the algorithm include various components of the National Institutes of Health Stroke Scale (NIHSS), measures of arm strength asymmetry (a value of 0 if both arms move equally or not at all, and a value of 1 if the arms move asymmetrically when instructed or imitated by the patient), or elements of other standard neurological exams common in the field of neurology.

[0057] In some situations, these results, data, and LVO status can be subjected to a learning machine, such as MATLAB or Classification Learner, using a classification tree. The results of the classification can be used to improve the algorithm. In some situations, the input to the classification algorithm is in the form of a spreadsheet incorporating the results of clinical tests and chaos measurements. For example, the clinical test inputs an asymmetric limb weakness variable with values ​​0 or 1 into the patient intake or into the numerical component of the NIHSS score. The various derivatives of chaos are input into the same spreadsheet data column as column data. Finally, the angiographic diagnosis of LVO is input as the result variable of the classification algorithm. Data from multiple patients, both with and without LVO, are sent to the classification algorithm, taking care to prevent overfitting. The output is a model that can be prospectively tested with patient data not used to construct the algorithm, to test the performance characteristics of the algorithm. In this embodiment, the algorithm is improved by successive bootstrapping of data.

[0058] FIG. 2B is a block diagram of an example of a headset 108 in which the system 100 is implemented. In this embodiment, the headset includes a user interface output device 230 and a user interface input device 232, as well as other components described with reference to FIG. 2A. The user interface output device 230 is any interface that provides information to a user and may include visual and / or audio output devices. Examples of visual output devices include visual displays, lights, and LEDs. If a light or LED is used, one or more LEDs or lights may be used to convey information to the user of the duration, color, frequency / duty cycle, and / or intensity of activation. If multiple LEDs or lights are used, information can be conveyed to the user based on the selection of a particular active light or LED in addition to the duration, color, frequency / duty cycle, and intensity. Multi-color LEDs can be used in some situations. Examples of suitable audio output devices include speakers and buzzers that provide information to the user of the tone of activation, duration, frequency / duty cycle, and / or intensity of the sound generated. In some circumstances, an audio output device may generate sound or music to convey information. The controller provides appropriate signals to the user interface output devices 230 to generate visual and / or audio output that provides information to the user. The user interface input devices 232 are any devices that allow a user to provide information to a system. The user interface input devices 232 provide signals or values ​​to the controller 220 that are indicative of information entered by a user. In some circumstances, the user interface input devices 232 may comprise one or more switches or buttons. Other examples include keypads, touchpads, and touch screen displays.In some situations, the user interface input devices 232 may include a microphone that provides signals to the controller corresponding to voice commands provided by a user and that can be interpreted by controller 220. In some situations, the user interface input devices 232 may include a combination of different types of input devices.

[0059] The system of the embodiment of FIG. 2B is similar to the system of the embodiment of FIG. 2A, except that some components are omitted, the analyzer is implemented in the headset 108, and the user interfaces 230 and 232 are included. A communication interface is omitted because it is not necessary to communicate data to the analyzer. The analyzer can be implemented using a processor, controller, processor arrangement, or other electrical circuit. In the embodiment of FIG. 2B, the controller 220 is shown separate from the analyzer 234, but both devices and other functions may be implemented on one processor. The data capture and analysis operations of the embodiment of FIG. 2B are similar to those discussed with reference to FIG. 2B, except that the digital data is 208. This represents that the acceleration signal generated by the accelerometer is provided directly to the analyzer without being sent to another device.

[0060] FIG. 3 is a graph of acceleration over time for two sets of data samples from one axis of an accelerometer. The two graphs each include four data samples of acceleration as a function of time. The data samples in FIG. 3 are not necessarily derived from captured data, but are intended to provide a visual interpretation of the chaos between the data sample sets of the head pulse recordings. For clarity and simplicity, only four data samples are shown in each set of graphs. As noted above, typically many data samples are captured and evaluated. In the first graph 301, the four data samples 302-305 represent possible acceleration plots captured by the headset for a subject without an LVO condition. In the second graph 306, the four data samples 307-310 represent possible acceleration plots captured by the headset for a subject with an LVO condition. In general, data samples 302-305 have less variation between data samples than data samples 307-310. The chaos of the data sample sets is related to the differences between the data samples. The difference may be in one or more of the amplitude, phase, and slope of the waveform. Figure 3 is a visual representation of the chaos difference between two sets of data samples. Another mathematical function or process applied by the algorithm can be used to quantify the chaos.

[0061] FIG. 4 is an illustration of an example of chaos in a cranial accelerometer of heartbeats for three different conditions. A first correlation plot 401 includes multiple traces 402 of acceleration as a function of time for a subject with an LVO stroke. The X-axis is time and the Y-axis is acceleration (gs). The waveforms of the multiple traces are time-aligned with an ECG trigger at time zero. The average of the multiple signals is shown as line 404, the high standard deviation of the bin values ​​is shown as line 405, and the low standard deviation is shown as line 403. A second correlation plot 410 includes multiple traces 411 of acceleration as a function of time for the subject of the first figure after thrombectomy. The first correlation plot 401 shows little correlation between the data trace morphology and the ECG. However, after thrombectomy, the second correlation plot 410 shows a higher correlation with the ECG. A third correlation plot 420 includes multiple data samples 421 of acceleration as a function of time for a subject with an acute small vessel stroke. A fourth correlation diagram 430 includes multiple data samples 431 of acceleration as a function of time for a subject with a stroke mimic that was in fact a migraine. Visual inspection of data samples 401 and 410 showed resolution of chaos after thrombectomy. The presence of chaos in 401 indicates that the subject is experiencing an LVO stroke. This forms the basis for biometric prediction of LVO occlusion. The patient represented by 421 has had a stroke at the time, but not due to an LVO. The absence of chaos in the data sample indicates that the patient does not need to be treated with thrombectomy. The patient represented by 430 has a migraine that mimics a stroke. The patient does not need to be treated for an LVO. In contrast to data samples 401, 420, and 430, a trained person can identify that the patient in 401 is different from the patients in 420 and 430. This is independent of measuring the pre-stroke baseline of each subject to reach this conclusion.

[0062] FIG. 5 is an illustration of an example of how a measure of chaos can be derived from multiple data samples. The horizontal axis is proportional to time and includes 2048 bins, where the data samples are scaled to the 2048 bin range based on the average heart rate. The vertical axis is acceleration measured in gs. An ECG trigger is used at time zero (0) to set the reference time. A single data sample 502 of acceleration in the vertical direction is shown as a solid line. A data sample 504 with the signals averaged over all data samples is shown as a dotted line. For each bin, the difference 506 between the single data sample and the averaged signal is shown as a shaded solid line.

[0063] Various techniques can be used to measure chaos in a data sample set. Below, some examples of chaos calculations are described, in which a head pulse recording is parsed into N data samples based on heart rate and further processed using one or more of the following calculations:

[0064] The first chaos calculation involves calculating the RMS difference between each pair of data samples for each data point, squaring the result, averaging it, and then calculating the square root.

[0065] The second chaos calculation calculates the mean, the standard deviation below the mean, and the standard deviation above the mean for all data samples from the head pulse recording. For each data sample, the number of data points that fall outside the standard deviation is tallied. The number of data samples is plotted on the y-axis, and the number of data points that fall outside the standard deviation is plotted on the x-axis. The ratio of values ​​from the resulting graph represents the chaos value.

[0066] The third chaos calculation calculates the RMS difference between the mean of all data samples in the head pulse and each data sample at the quartile between them. A histogram is calculated for each quartile and the histogram curve is fitted to a normal distribution. A sigma (the width of the normal fit) is used for each quartile and the ratio between the quartile sigmas is the chaos value.

[0067] The fourth chaos calculation calculates the average of all data samples from the head pulse recording to obtain the absolute value. The result is integrated over all data points. The ratio of the quartiles is taken as the chaos value.

[0068] The fifth chaos calculation involves calculating, for each data point, the RMS difference between each data sample and the average of all data samples from the head pulse recording, squaring the result, averaging it and calculating the square root.

[0069] The sixth chaos calculation counts the number of bins in the 2048 traces that differ from the average by increasing the threshold.

[0070] FIG. 6 is a flow chart of an example of a method for identifying brain states. Any combination of hardware and code may be used to perform the method. Examples of some suitable techniques include a computer, tablet, or smartphone running code to perform the steps described below. In many circumstances, other steps may be performed in addition to those described with reference to FIG. 6.

[0071] At step 602, a plurality of data samples are received, where each data sample is indicative of head acceleration due to blood flow in the brain in one axis. In this embodiment, each data sample is a digital representation of acceleration as a function of time. As described above, a suitable technique for capturing the data samples includes attaching a three-axis accelerometer to the subject's head, where an analog accelerometer signal is converted to a digital data sample. The data samples are aligned in time with respect to the heartbeat based on information from the data samples or from other signals related to the heartbeat. Examples of such signals include ECG and PPG signals as discussed above.

[0072] At 604, the chaos level of the data samples is measured. Chaos can be defined numerically by several measures, most often based on the dissimilarity of any single data sample after a heartbeat to an estimate of the average of all measurements following several heartbeats, or the dissimilarity of a single data sample to all other data samples. The value can be measured or calculated in a variety of ways. Calculations such as root mean square calculations can be applied to various functions and the results normalized to calculate a value for chaos.

[0073] At 606, an indication is provided to the user that an LVO stroke has occurred. In the example, if certain clinical exam conditions are present, an indication is provided if the chaos level exceeds a threshold. In the example herein, information regarding concurrent clinical neurological exam findings, such as limb strength and components of the NIHSS score, is integrated with the biometric data at 604. A classification algorithm from the data of multiple LVO and non-LVO patients is then used to predict a binary or detailed probability of LVO in the patient being examined. Thus, the algorithm can diagnose the presence or absence of LVO based on the chaos level and the neurological and clinical data acquired when the subject's acceleration data was obtained. Examples of indications include text or symbols provided via a visual display, or LED lights of various colors, although any indication method may be used when an LVO condition is present.

[0074] 7-11 are diagrams illustrating examples of evaluating data samples of a head pulse recording. FIG. 7 illustrates two exemplary data samples 700 and 701 of a head pulse recording. In the examples herein, the head pulse recording consists of many heartbeat intervals and a time series of acceleration values ​​covering all measurement axes. The data of the head pulse recording is parsed into RR intervals associated with the subject's heartbeat. Each heartbeat interval is resampled to contain 2048 data points, so that all RR intervals are equal. Each heartbeat interval is referred to as a data sample. The RMS difference between the two data samples 700 and 701 is calculated one data point at a time by subtracting the first data point A 703 on the first data sample 700 from the second data point B 704 on the second data sample 701, squaring the calculated difference, adding all 2048 such data point differences, dividing the sum by 2048, and taking the square root. The RMS chaos value is calculated from a head pulse recording consisting of N data samples. As shown in FIG. 8, each data sample is calculated relative to every other data sample for a total of N(N-1) / 2 data sample pairs, separately for each axis.

[0075] Figure 8 is an illustration of an example of data samples 801-805 representing all data samples from the entire head pulse recording as shown in Figure 7. Calculations 806-809 are performed between the first data sample and all remaining data samples. Calculations 810-812 are performed between the second data sample and all remaining data samples. Calculation 813 is performed between the penultimate data sample 804 and the last data sample 805. These N(N-1) / 2 results are then squared, averaged and square rooted to calculate the RMS chaos value for each axis.

[0076] Another example of a technique for measuring chaos involves calculating the distribution outside of the standard deviation. For each data point, the value of each data sample at that data point is used to calculate the standard deviation of the data sample. A series of such data points is a curve of data points above and below the 2048 average curve. Figure 4 shows several head pulse recordings that have been parsed into multiple data samples and plotted with the averages 404 and 413 shown, one of which has a standard deviation above 405 and 415, and another below 403 and 412. For each data sample, the number of data points that fall outside of one standard deviation can be tallied.

[0077] FIG. 9 shows a graph 901 of the resulting number of data samples on the y-axis and the number of data points outside the standard deviation on the x-axis. This is calculated for each acceleration axis recorded. The ratio of the value of the third quarter 904 divided by the value of the first quarter 902 is the chaos number. Other ratios may be derived from the plot using the unratioed values ​​902-905, or other parts of the graph 901 used for the numbers.

[0078] A quartile distribution is another example of a chaos measurement technique. Figure 10 shows a graph of the mean 1000 of data samples from a head pulse recording and one of the data samples 1001. For each quartile, between the mean 1000 and one of the data samples 1001, one data point at a time, we subtract data point A 1002 from data point B 1003 and square the calculated difference to calculate the quartile distribution. For each quartile, we create a histogram of the distribution 1004 of the squared differences. We subject each histogram to a normal distribution fit 1005, calculating a single number that is the histogram width 1006. The result, i.e. the width of the normal distribution 1006, is taken as the chaos value, and several ratios are also calculated. These include the ratio of the average of the second and third quartiles divided by the first quartile, the ratio of the second quartile divided by the first quartile, the ratio of the average of the second, third and fourth quartiles divided by the first quartile, and the ratio of the average of the first and second quartiles divided by the average of the third and fourth quartiles. These calculations are performed for each axis. Instead of using quartiles, other divisions may be used that distinguish between chaos in data samples at the beginning of the data sample interval and those at the middle and end of the data sample interval.

[0079] The integrated average acceleration is yet another example of a technique for measuring chaos. FIG. 11 shows a graph 1100 of the absolute value of the average of data samples from a head pulse recording. Graph 1101 is the integral of 1100. Any of the quartiles 1102-1105 are used in ratios to measure the chaos value. These include the average of the second and third quartiles divided by the first quartile, the second quartile divided by the first quartile, the average of the second, third, and fourth quartiles divided by the first quartile, and the average of the first and second quartiles divided by the average of the third and fourth quartiles. These calculations are performed for each axis. Instead of using the quartiles, other divisions may be used that distinguish between chaos in data samples at the beginning of a data sample interval and the middle and end of a data sample interval.

[0080] RMS average chaos is an example of another approach to measuring chaos. As shown in Figure 12, the RMS difference between data sample 1200 and the average of all data samples in the same head pulse recording 1201 is calculated one data point at a time, for example, by subtracting data point A 1203 from data point B 1204, squaring the calculated difference, adding up the differences of all 2048 such data points, dividing the sum by 2048, and taking the square root. The RMS chaos of a head pulse recording consisting of the RMS values ​​of N data samples is compared to the average 1201, respectively.

[0081] The techniques, devices, and systems discussed above have advantages over conventional systems for diagnosing stroke in patients. Patients with LVO stroke (approximately 40% of all ischemic strokes) benefit from rapid removal of the clot blocking the cerebral artery (thrombectomy). The sooner this is performed, the better the patient's prognosis. Conventional approaches require brain imaging by CT or MRI to diagnose whether a patient has suffered a stroke, as clinical examination of the patient alone is not accurate enough. In some cases, clinical assessment is not accurate in concluding whether an LVO condition exists, as some non-LVO strokes (small vessel strokes) and some non-stroke conditions, such as migraines and seizures, can mimic an LVO stroke.

[0082] As a result, the above techniques may be improved for diagnosis. Diagnosis of stroke due to LVO is one example of such a diagnosis. A device operating according to this example combines clinical test features with accelerograms to better identify LVO patients and more accurately diagnose. A relatively inexpensive portable device can provide a rapid and accurate diagnosis. Such a device can be used in emergency situations (pre-admission by emergency medical services, in the emergency department of a primary stroke center, and in hospitalized patients with a sudden change in neurological status) to facilitate triage and improve the prognosis of this form of stroke. The technique can be used to detect LVO stroke without the use of cross-sectional imaging of the brain. Also, the use of one or more triaxial sensors at one or more locations on the head provides a simple device that can be easily and quickly used on a patient without lifting the patient's head. Because of its simple nature, the device can be used without regard to other interventions. Furthermore, a portable, non-invasive, and non-significant risk-free device in the above implementation can diagnose the presence or absence of a stroke within 30-60 seconds. The manufacturing cost of such a device would be orders of magnitude lower than CT and MRI scanning devices. Moreover, such devices do not pose any risk of harm to the patient compared to the state of the art in the field: for example, CT causes radiation exposure and the risk of kidney failure due to the administration of contrast agents, and MRI is contraindicated for patients with metal implants due to the risk of damage from the strong magnetic field, and such devices can also be used at the bedside (battery-operated) and anywhere, including outside and within the hospital.

[0083] Typically, medical professionals are trained in the use of the headset and understand the usefulness of using this predictive tool to aid in the diagnosis of LVO. The user of the device is trained with a specific instruction manual. The system can be used in an ambulance, hospital, or other location. In most cases, the device is stored in a sealed container. The headset has an internal electronic serial number, which may be embossed on the outside of the device. Once a patient with a suspected stroke is identified, the user of the device removes the headset from its packaging and turns it on with a switch or by pulling an insulating tab, which connects it to the battery. The headset is then placed in the coronal plane of the subject.

[0084] In an embodiment where the headset is equipped with an LED user interface and the analyzer is in a computer, the device user attaches the PPG device to the patient's earlobe, finger, or other extremity, connects the PPG sensor wires to the headset, and observes the LEDs flashing with each heartbeat to verify that the PPG is working properly. Additionally or simultaneously, the user places ECG electrodes on each shoulder or upper limb of the patient and connects the ECG electrodes to the headset device via a plug connector. This provides an electrical pulse to the device with each heartbeat. The user can observe this via another flashing LED.

[0085] After the head pulse has been recorded sufficiently, the device will light a specific color LED to inform the user that the data is complete. If the recording is poor, another LED will be displayed. In response, the user can either reposition the headset, replace the headset, or calm the patient to avoid full head movement to allow for sufficient recording.

[0086] After obtaining sufficient head pulse recordings, the user obtains specialized neurological exam data from the patient and enters these findings into a connected computer or a computer that communicates wirelessly with the device. The analyzer in the attached computer executes the algorithm and provides an electronic message reporting the degree of certainty that the patient is having an LVO stroke. The user then uses this information to determine the next steps of care. Via text, email, website, or other direct services, this information can be sent to the receiving hospital, medical consultants, or other staff involved in the patient's emergency care.

[0087] In instances where the analyzer is implemented within a headset, lab test information can be input into the headset via wires, wirelessly, or via a physical switch, and an internal microprocessor within the headset executes the algorithm and displays the results on an output user device, such as a visual display on a headset LED arrangement.

[0088] The techniques, devices, and systems discussed herein aid in LVO stroke diagnosis and aid in triage decisions that improve outcomes for LVO stroke patients by reducing the time to delivery of effective treatment.

[0089] Head pulse measurements from 42 subjects with suspected stroke who underwent CTA at a major medical center were obtained and analyzed {Smith et al., ISC 2019}. Analysis of the waveforms from subjects without LVO revealed that each data sample was closely related to time and also very similar in time between each data sample. On the other hand, data samples from stroke patients with LVO were not closely related to the heartbeat and appeared "chaotic". The chaos was objectified with a series of statistics and sent to a machine learning algorithm along with components of the NIH Stroke Scale to derive a predictive model of LVO. Subjects with LVO were chaotic significantly more frequently (83% vs. 25%, p=0.0002) than those without LVO. When the model was combined with a simple neurological feature (asymmetry in upper limb muscle strength), it appropriately classified all LVO patients (Keenan et al., ISC 2019).

[0090] Combining data from upper limb muscle asymmetry and head pulse recordings can be an effective tool to diagnose the presence or absence of LVO. This uses a headset equipped with electronic components to measure the head pulse and a computer programmed with a model algorithm, and is a portable system that takes only 30–60 seconds.

[0091] As noted above, the techniques discussed herein can be adapted for use in other conditions. For example, the head pulse contains information regarding concussion and traumatic brain injury (TBI). Analysis of the frequency of the head pulse shows that the frequency of the brain's power component of the heartbeat is shifted to a harmonic of the baseline heart rate. This frequency shift indicates that the skull and brain are vibrating at a higher frequency than normal after concussion and TBI. Normal subjects without concussion or TBI do not have significant energy in the range of harmonics six times higher than the heart rate, whereas over 86% of subjects with concussion do. This allows for objective biometric identification of concussion and TBI. This biometric can be used to help determine when it is safe to return an athlete to play or a soldier to combat, as the brain is more vulnerable during this stage of concussion recovery. The biometric data indicates that the head pulse is abnormal beyond the time that the subject would feel better. Since feeling well (a subjective measure) is the gold standard for returning to play, using this objective bioinformation could reduce the frequency of second impact syndrome and possibly chronic traumatic encephalopathy. Using the head pulse to identify the cause of altered mental status in patients who arrive at a clinic or emergency department with no medical history can help identify patients who have suffered some form of head trauma.

[0092] Analysis of head pulse data allows differentiation of intracerebral hemorrhage from LVO stroke. Analysis of head pulse data from subjects with suspected stroke (no LVO but ICH) shows no chaotic characteristics of LVO. It is important to differentiate ICH from LVO because ICH patients may not require triage to a specialized center, whereas LVO patients require prompt treatment.

[0093] Using multiple accelerometers in an orthogonal relationship (XYZ axes) provides additional information regarding the head pulse in various pathologies. Concussion is specifically seen with a lateral sensor, while LVO is seen with a longitudinal sensor. Using bilateral triaxial accelerometers, heartbeat can be detected. This utilizes scalp plethysmography, where the scalp expands centripetally during systole. The discrepancy of opposing forces obtained from two or more sensors attached to opposite sides of the skull provides a waveform that is geared to the heartbeat. This derived trigger is used to parse the head pulse recording into data samples. Signal averaging of the data samples allows analysis of amplitude, phase, and frequency.

[0094] The amplitude of the main peak of the data samples on each axis is modulated by respiration. Respiration can be inferred from this signal. This signal can be further improved by combining the signals from each axis. Heart rate and RR intervals also vary with respiration. These signal measurements can be combined to provide a stable and accurate respiration signal.

[0095] The head pulse measurement device can be implemented with a MEMS chip, a digitizer, a battery, and an RF or infrared transmitter. These components can be integrated into one or two semiconductor ASICs. The sensor chip is typical of those found in smartphones, less than 2 mm square and less than 1 mm thick. With these dimensions, the entire device can be embedded in a glass temple arm, as part of a hearing aid, as a patch attached with adhesive, or injected as a capsule under the skin. The advantage of incorporating a non-invasive device into one or more devices that are monitoring almost constantly is that if a neurological event occurs, it is detected within seconds. To obtain the data necessary to gather characteristics of many neurological events of clinical interest, significant data is required before and after the event. In traditional clinical trials, obtaining such data is economically unfeasible. With very low-cost integrated devices, data can be obtained almost continuously from a huge number of patients at a very low cost. Many strokes occur during sleep, known as "wake-up strokes". If a continuous monitor could alert the patient or the patient's caregiver that a stroke is occurring, prognosis would be improved.

[0096] Static accelerometers provide additional information about the head position when the signal was acquired. The polarity of the accelerometers relative to the headset they are housed in is known. Assuming the headset is worn in a particular orientation on the head (usually coronal), the position of the head relative to gravity is known. This information can be integrated into the algorithm to improve the predictive value of various brain disorders. Additionally, geomagnetic sensors can be incorporated into the device to provide additional information about the orientation of the body and headset.

[0097] A GPS device can be incorporated into the headset to provide information about the device's geographic location, which is used to determine the patient's location and periodically transmit the GPS coordinates via cellular technology to provide mapping capabilities to the end user, thereby providing the treating physician with real-time information about the patient's location, improving the efficiency of managing patient care.

[0098] Accelerometers lose sensitivity in the low frequency range. New devices can be made to increase the sensitivity to lower frequencies. A liquid-filled sphere vibrates when perturbed by an external force or vibration. The vibration can be measured by either reflected or refracted light from the surface of the sphere. For a refracted ray entering the sphere off-axis, the ray first refracts as it enters the sphere, then is reflected off the back surface and refracts again before leaving the sphere. The position and angle of this exiting ray is a very sensitive measure of the shape of the sphere. A light source such as a laser is highly parallel, providing a perfect probe light source for measuring liquid-filled spheres. Measuring the return ray over time provides the time-resolved motion of the liquid-filled sphere sensor. Measurements at three orthogonal angles provide the time-resolved characteristics of the motion of the liquid-filled sphere sensor, and of any object in contact with the liquid-filled sphere sensor.

[0099] Clearly, other embodiments and modifications of the present invention will occur readily to those skilled in the art in light of these teachings. The above description is illustrative and not limiting. The present invention should be limited only by the following claims, which include all such embodiments and modifications when viewed in conjunction with the above specification and accompanying drawings. Therefore, the scope of the present invention should be determined not with reference to the above description, but with reference to the appended claims and their full scope of equivalents.

Claims

1. accelerometer; a fixation mechanism configured to fix the accelerometer to a human head including a brain such that the accelerometer is configured to generate an accelerometer signal indicative of head movement due to blood flow in the brain caused by heartbeat; and 1. A system comprising: an input configured to receive a plurality of data samples based on the accelerometer signal, each data sample indicative of a head movement in conjunction with a different heartbeat in an axis; and an analyzer configured to measure a chaos level of the plurality of data samples and indicate that a large vessel occlusion (LVO) has occurred in the brain when the chaos level exceeds a threshold, the data samples include accelerometer signal levels at time intervals, and the chaos level is based at least in part on a difference between signal levels at a plurality of selected times for each of a plurality of acceleration versus time functions of the plurality of data samples; the chaos level is based at least in part on a calculation of the plurality of acceleration versus time functions; The system, wherein the calculating includes calculating a difference function for each data sample multiple times, each difference function including a difference between the data sample and an average of all data samples.

2. The system of claim 1 , wherein the accelerometer is a three-dimensional accelerometer capable of providing axial accelerometer signals in any of three orthogonal axes.

3. 3. The system of claim 2, wherein the analyzer input is configured to receive a second plurality of data samples based on an axial accelerometer signal indicative of head movement in another axis orthogonal to the one axis, each data sample of the second plurality of data samples indicative of head movement in conjunction with a different heartbeat in the other axis, and the analyzer is configured to measure a second chaos level of the second plurality of data samples and indicate that a large vessel occlusion (LVO) has occurred in the brain if the second chaos level exceeds a second threshold, the second chaos level being based at least in part on a second difference function for each data sample of the other plurality of data samples, each second difference function comprising a difference between a data sample of the other plurality of data samples and an average of the second plurality of data samples.

4. 4. The system of claim 3, wherein the analyzer input is configured to receive a third plurality of data samples based on a third axis accelerometer signal indicative of head movement in a third axis orthogonal to the one axis, each data sample of the third plurality of data samples indicative of head movement in conjunction with a different heartbeat in the third axis, and the analyzer is configured to measure a third chaos level of the third plurality of data samples and indicate that a large vessel occlusion (LVO) has occurred in the brain if the third chaos level exceeds a third threshold, the third chaos level being based at least in part on a third difference function for each data sample of the third plurality of data samples, each third difference function comprising a difference between a data sample of the third plurality of data samples and an average of the second plurality of data samples.

5. The system of claim 1 , wherein the chaos level is based at least in part on a difference between signal levels at a plurality of selected times for each of a plurality of acceleration versus time functions of the plurality of data samples.

6. The calculation is 2. The system of claim 1, comprising calculating a difference function for each data sample multiple times, each difference function comprising a difference between the data sample and an average of all data samples.

7. The calculation is Modifying and summing each difference function to produce multiple results; summing the results to obtain a sum of the results; and Normalizing the sum of the results to several data samples, or Normalizing the resulting sum to the minimum and maximum average of the data samples. The system of claim 6 further comprising:

8. The calculation is determining the root mean square of said difference function; The system of claim 6 , comprising:

9. The calculation is summing the root mean square of the difference functions to obtain a sum; and normalizing said sum to the minimum and maximum of the average of said data samples; The system of claim 8 further comprising:

10. The system of claim 1 , wherein the analyzer is configured to measure a level of chaos based at least in part on a shape of an acceleration versus time function of each data sample.

11. The system of claim 1 , wherein the analyzer is configured to align a plurality of acceleration versus time functions in a time series based on information of the plurality of acceleration versus time functions.

12. 2. The system of claim 1, wherein the analyzer is configured to align the plurality of acceleration versus time plots in a time series based at least in part on an electrocardiogram (ECG) or photoplethysmogram (PPG) signal indicative of cardiac contractions.

13. a transmitter configured to transmit a wireless signal indicative of the plurality of data samples; and a receiver configured to receive the wireless signal and provide the data samples to the analyzer; The system of claim 1 further comprising:

14. an analog-to-digital converter (ADC) configured to sample the accelerometer signal and generate a plurality of data samples; The system of claim 13 further comprising:

15. a controller configured to time-sequentially align the plurality of data samples prior to transmitting the wireless signal including the time-sequentially aligned data samples; The system of claim 14 further comprising:

16. The system of claim 1 , wherein the analyzer is configured to diagnose the presence or absence of an LVO condition based at least in part on neurological examination information.

17. a three-dimensional accelerometer configured to generate three orthogonal signals, each orthogonal signal indicative of an acceleration of a motion of the three-dimensional accelerometer in an axis orthogonal to the other two axes; a fixation mechanism configured to fix the three-dimensional accelerometer to a human head including a brain; and an analyzer configured to receive three sets of data samples, each data sample being based on one of the orthogonal signals and indicative of a head movement associated with a different cardiac contraction in one axis; and configured to measure a chaos level of each set of data samples and indicate that a large vessel occlusion (LVO) has occurred in the brain when the chaos level of at least one set of data samples exceeds a threshold value. A system comprising: the data samples include accelerometer signal levels at time intervals, and the chaos level is based at least in part on a difference between signal levels at a plurality of selected times for each of a plurality of acceleration versus time functions of the plurality of data samples; the chaos level is based at least in part on a calculation of the plurality of acceleration versus time functions; The system, wherein the calculating includes calculating a difference function for each data sample multiple times, each difference function including a difference between the data sample and an average of all data samples.

18. an input configured to receive a plurality of data samples, each data sample indicative of a different cardiac contraction-induced cerebral blood flow head movement in an axis; an analyzer configured to measure a chaos level of a plurality of data samples aligned in time with respect to the cardiac contractions, the data samples include accelerometer signal levels at time intervals, and the chaos level is based at least in part on a difference between signal levels at a plurality of selected times for each of a plurality of acceleration versus time functions of the plurality of data samples; the chaos level is based at least in part on a calculation of the plurality of acceleration versus time functions; the calculating includes calculating a difference function for each data sample multiple times, each difference function including a difference between the data sample and an average of all data samples; and an output configured to indicate that a large vessel occlusion (LVO) has occurred in the brain when the chaos level exceeds a threshold. An apparatus comprising:

19. 20. The apparatus of claim 18, wherein the plurality of data samples are accelerometer signal samples indicative of movement of the head in only one axis.

20. 20. The apparatus of claim 18, wherein the plurality of data samples are accelerometer signal samples indicative of an output of a head-mounted accelerometer.

21. 21. The apparatus of claim 20, wherein the accelerometer is a three-dimensional accelerometer capable of providing an axial accelerometer signal in any of three orthogonal axes, a chaos level determined for the data samples corresponds to each axis, and the output is configured to indicate the presence of LVO in the brain when the chaos level in at least one axis exceeds a threshold.

22. The calculation is Modifying and summing each difference function to produce multiple results; summing the results to obtain a sum of the results; and Normalizing the sum of the results to several data samples, or Normalizing the resulting sum to the minimum and maximum average of the data samples.

20. The apparatus of claim 18, further comprising:

23. The calculation is determining the root mean square of said difference function; 20. The apparatus of claim 18, comprising:

24. The calculation is summing the root mean square of the difference functions to obtain a sum; and normalizing said sum to the minimum and maximum of the average of said data samples; 24. The apparatus of claim 23, further comprising:

25. 20. The apparatus of claim 18, wherein the chaos level is based at least in part on a shape of an acceleration versus time plot of each accelerometer signal sample.

26. 20. The apparatus of claim 18, wherein the analyzer is configured to diagnose the presence or absence of an LVO condition based at least in part on neurological examination information.

27. 1. A method of operating a system, the system comprising: receiving a plurality of data samples including accelerometer signal samples, each data sample indicative of a different cardiac contraction induced cerebral blood flow induced head movement in an axis; measuring a chaos level of a plurality of data samples aligned in time with respect to the cardiac contraction; and indicating that a large vessel occlusion (LVO) has occurred in the brain when the chaos level exceeds a threshold; the data samples include accelerometer signal levels at time intervals, and the chaos level is based at least in part on a difference between signal levels at a plurality of selected times for each of a plurality of acceleration versus time functions of the plurality of data samples; the chaos level is based at least in part on a calculation of the plurality of acceleration versus time functions; A method wherein the calculating comprises calculating a difference function for each data sample multiple times, each difference function comprising the difference between the data sample and an average of all data samples.

28. 28. The method of claim 27, wherein the accelerometer signal samples are indicative of movement of the head in only one axis.

29. 28. The method of claim 27, wherein the accelerometer signal samples are indicative of an output of a head-mounted accelerometer.

30. 30. The method of claim 29, wherein the accelerometer is a three-dimensional accelerometer capable of providing axial accelerometer signals in any of three orthogonal axes.

31. Measuring the chaos level includes measuring a difference between signal levels at a plurality of selected times on each of a plurality of acceleration versus time plots of a plurality of accelerometer signal samples.

28. The method of claim 27, comprising:

32. Measuring the chaos level Calculating the chaos level from a plurality of acceleration versus time functions.

32. The method of claim 31 , comprising:

33. Calculating the chaos level 33. The method of claim 32, comprising calculating a difference function for each data sample multiple times, each difference function comprising the difference between the data sample and an average of all data samples.

34. Calculating the chaos level Modifying and summing each difference function to produce multiple results; summing the results to obtain a sum of the results; and Normalizing the resulting totals to several data samples 34. The method of claim 33, further comprising:

35. Calculating the chaos level determining the root mean square of said difference function; 34. The method of claim 33, further comprising:

36. Calculating the chaos level summing the root mean square of the difference functions to obtain a sum; and normalizing said sum to the minimum and maximum of the average of said data samples; 36. The method of claim 35, further comprising:

37. Calculating the chaos level Modifying and summing each difference function to produce multiple results; summing the results to obtain a sum of the results; and normalizing the summed result to the minimum and maximum of the average of the data samples.

34. The method of claim 33, comprising:

38. Measuring the chaos level Evaluating the shape of the acceleration versus time plot of each accelerometer signal sample 28. The method of claim 27, comprising:

39. chronologically ordering the plurality of data samples based on information measured from the plurality of data samples.

28. The method of claim 27, further comprising:

40. receiving neurological examination information; and Diagnosing the presence or absence of an LVO condition based on the neurological examination information and the chaos level.

28. The method of claim 27, further comprising:

Images