This disclosure is directed to medical devices and, more particularly, to systems and methods for detecting stroke, brain ischemic events, and/or brain hypoxia events.
Stroke is a serious medical condition that can cause permanent neurological damage, complications, and death. Stroke may be characterized as the rapidly developing loss of brain functions due to a disturbance in the blood vessels supplying blood to the brain. The loss of brain functions can be a result of ischemia (lack of blood supply) caused by thrombosis or embolism, or hemorrhage (e.g., a ruptured blood vessel). During a stroke, the blood supply to an area of a brain may be decreased, which can lead to dysfunction of the brain tissue in that area.
Stroke is the number two cause of death worldwide and the number one cause of disability. Speed to treatment is the critical factor in stroke treatment as 1.9 M neurons are lost per minute on average during stroke. Stroke diagnosis and time between event and therapy delivery are the primary barriers to improving therapy effectiveness. Stroke has 3 primary etiologies; i) ischemic stroke (representing approximately 65% of all strokes), ii) hemorrhagic stroke (representing approximately 10% of all strokes), and iii) cryptogenic strokes (includes TIA, representing approximately 25% of all strokes). Strokes can be considered as having neurogenic and/or cardiogenic origins.
A variety of approaches exist for treating patients undergoing a stroke. For example, a clinician may administer anticoagulants, such as warfarin, or may undertake intravascular interventions such as thrombectomy procedures to treat ischemic stroke. As another example, a clinician may administer antihypertensive drugs, such as beta blockers (e.g., Labetalol) and ACE-inhibitors (e.g., Enalapril) or may undertake intravascular interventions such as coil embolization to treat hemorrhagic stroke. Lastly, if stroke symptoms have resolved on their own with negative neurological work-up, a clinician may administer long-term cardiac monitoring (external or implantable) to determine potential cardiac origins of cryptogenic stroke.
In general, the disclosure is directed to devices, systems, and techniques for determining characteristic(s) of a brain event, via a medical device, e.g., an implantable medical device (IMD) or external medical device, located on the head of a patient. For example, using electrodes, the medical device may sense electrical signals from a patient and generate EEG signal(s) based on the electrical signals. Processing circuitry of the system, e.g., of the medical device or another device configured to communicate with the medical device, may apply EEG signal(s) to a machine learning (ML) model to determine one or more characteristics of a brain event, the ML model being trained on training EEG data and simulated EEG data.
Brain events, such as a stroke, and potential treatments thereof, may vary significantly in type of the brain event, such as a type of stroke, in location in the brain of the event, and/or in infarct volume/size of the stroke. Non-invasive techniques for determining stroke have had difficulty accurately determining a type of stroke that occurred, a location of where a stroke has occurred in the brain of a patient, and/or an infarct volume/size of a stroke that occurred.
The techniques of this disclosure may provide one or more advantages. For example, applying a ML model trained using training EEG data and simulated EEG data may reduce or remove coverage gaps (e.g., “non-coverage” areas) in which the ML model is not able to accurately identify particular types, locations, magnitudes (e.g., infarct volume/size), and/or times of brain events due to insufficient training EEG data. The reduction or removal of coverage gaps of a ML model may improve detection by the system of a type of the brain event (e.g., a type of stroke, such as ischemic stroke, hemorrhagic stroke, cryptogenic stroke, or stroke mimic), a location in the brain of the brain event (e.g., stroke), and/or an infarct volume/size of the brain event (e.g., stroke). Accurately determining a type, location, and/or infarct volume/size of a brain event, such as a stroke, may lead to more particularized treatment to target the particularities of a brain event, which may lead to improved outcomes of treatments.
In accordance with techniques of this disclosure, processing circuitry of the system may generate a four-dimensional map of the brain of the patient based, at least in part, on the EEG signal(s), and determine whether the ML model has coverage gaps using the generated four-dimensional map. In some examples, processing circuitry may generate additional simulated EEG data corresponding to the coverage gaps to fill in the coverage gaps and send the additional simulated EEG data to further train the ML model to update the ML model, and apply the EEG signal(s) to the updated ML model to determine characteristic(s) of the brain event.
In one example, the disclosure describes a system comprising: a memory; a plurality of electrodes; sensing circuitry configured to sense, via at least two electrodes of the plurality of electrodes, electrical signals from a patient; and generate, based on the electrical signals, one or more electroencephalography (EEG) signals; and processing circuitry configured to: receive, from the sensing circuitry, one or more EEG signals; and apply the one or more EEG signals to a machine learning (ML) model to determine one or more characteristics of a brain event, the ML model being trained on training EEG data and simulated EEG data.
In another example, this disclosure describes a method for sensing, by sensing circuitry and via at least two electrodes of a plurality of electrodes, electrical signals from a patient; generating, by the sensing circuitry and based on the electrical signals, one or more electroencephalography (EEG) signals; receiving, by processing circuitry and from the sensing circuitry, the one or more EEG signals; and applying the one or more EEG signals to a machine learning (ML) model to determine one or more characteristics of a brain event, the ML model being trained on training EEG data and simulated EEG data.
In another example, this disclosure describes a computer-readable medium comprising instructions that, when executed, cause processing circuitry to perform sensing, via at least two electrodes of a plurality of electrodes, electrical signals from a patient; generating, based on the electrical signals, one or more electroencephalography (EEG) signals; receiving the one or more EEG signals; and applying the one or more EEG signals to a machine learning (ML) model to determine one or more characteristics of a brain event, the ML model being trained on training EEG data and simulated EEG data.
The summary is intended to provide an overview of the subject matter described in this disclosure. It is not intended to provide an exclusive or exhaustive explanation of the systems, device, and methods described in detail within the accompanying drawings and description below. Further details of one or more examples of this disclosure are set forth in the accompanying drawings and in the description below. Other features, objects, and advantages will be apparent from the description and drawings, and from the claims.
Many aspects of the present disclosure can be better understood with reference to the following drawings. The components in the drawings are not necessarily to scale. Instead, emphasis is placed on clearly illustrating the principles of the present technology.
It can be difficult to determine whether a patient is suffering from a stroke or has suffered from a stroke. Current diagnostic techniques typically involve evaluating a patient for visible symptoms, such as paralysis or numbness of the face, arm, or leg, as well as difficultly walking, speaking, balancing, seeing, or understanding (e.g., the B.E.F.A.S.T visible stroke indication of Balance, Eyes, Face, Arm, Speech, Time to call for emergency help). However, these techniques may result in undiagnosed strokes, particularly more minor strokes that leave patients relatively functional upon cursory evaluation. Other diagnostic techniques may include evaluating imaging, such as a computerized tomography (CT) scan or magnetic resonance imagining (MRI), which usually needs to be performed in a medical office or hospital.
Even for relatively minor strokes, it is important to treat the patient as soon as possible because treatment outcomes for stroke patients are highly time-dependent. However, such treatments may be frequently underutilized and/or relatively ineffective due to the failure to timely identify whether a patient is undergoing or has recently undergone a stroke. This is a particular risk with more minor strokes that leave patients relatively functional upon cursory evaluation.
In addition, it can be difficult to determine, non-invasively, a type of stroke or seizure that occurred, a location of where a stroke or seizure has occurred in the brain of a patient, and/or an infarct volume/size of a stroke. Determining a type of stroke or seizure, a location in the brain of a stroke or seizure, and/or an infarct volume/size of a stroke may lead to more particularized treatment to target a particular type of stroke or seizure, a particular location in the brain of a stroke or seizure, and/or particular infarct volume/size of a stroke.
Accordingly, there is a need for improved methods for determining what types of brain events occurred, such as a type of stroke or seizure. There is also a need for improved methods for determining a location in the brain of the patient where a stroke or seizure occurred. There is also a need for improved methods for determining an infarct volume/size of a stroke that occurred. This disclosure describes various systems, devices, and techniques for determining one or more characteristics of a brain event using one or more devices configured to sense patient signals including electroencephalogram (EEG) signals, which may be located on or near the head of the patient. In some examples, a brain event, as described herein, may include one or more of a stroke, a brain ischemic event, a brain hypoxia event, or a seizure.
Due to an insufficient amount of training EEG data to detect a variety of types, locations, magnitudes, and/or times of brain events, a ML model trained on training EEG data, by itself, may have coverage gaps (e.g., “non-coverage” areas) in which the ML model is not able to accurately identify particular types, locations, magnitudes (e.g., infarct volume/size), and/or times of brain events while just being trained on the training EEG data itself.
This disclosure describes techniques for applying EEG signal(s) to a ML model to determine one or more characteristics of a brain event, the ML model being trained on training EEG data and simulated EEG data. In some examples, particular types of simulated EEG data may be selected to supplement the training EEG data to train the ML to fill in the coverage gaps of the ML model being trained on training EEG data itself. Applying a ML model trained on training EEG data and simulated EEG data helps improve the coverage of the ML model to identify particular types, locations, magnitudes (e.g., infarct volume/size), and/or times of brain events with greater accuracy. Determining brain event characteristic(s) of a brain event enables a more particularized treatment to target a brain event with particular brain event characteristics, which will lead to more accurate diagnoses, better patient outcomes and/or more improved healthcare efficiency.
This disclosure describes techniques of applying EEG signal(s) to generate a multi-dimensional, such as a four-dimensional (4D) map, and using the multi-dimensional map (e.g., 4D map) to determine whether the ML model has coverage gaps. In response to determining the ML model has coverage gaps, this disclosure describes generating additional simulated EEG data corresponding to the coverage gaps to fill in the gaps. In some examples, the additional simulate data may be used to further train the ML model to update the ML model. The updated ML model would have at least some of the coverage gaps filled in. The EEG signal(s) may be applied to the updated ML model to determine characteristic(s) of the brain event. In some examples, using the updated ML model with the coverage gaps filled in by the simulated data may help the ML model to identify more particular types, locations, magnitudes (e.g., infarct volume/size), and/or times of brain events the ML model may not have been able to with the coverage gaps.
Characteristics of a brain event may include one or more of a type of brain event, such as a type of stroke or a type of seizure, a brain event location in the brain, such as a location of a stroke or a location of a seizure in the brain, a last known normal (LKN) time before onset of the brain event, a time of the onset of the brain event, and/or infarct volume/size of the stroke. In some examples, a type of stroke may include one or more of ischemic stroke, hemorrhagic stroke, cryptogenic stroke, stroke mimic/non-stroke.
In some examples, a plurality of electrodes may be placed on a patient's head to sense electrical signal from a patient and generate EEG signal(s) based on the electrical signals. In other examples, as described herein, a medical device (e.g., an IMD or external medical device wearable by the patient), may be configured to determine characteristic(s) of a brain event from a location on or near the head of the patient. Using electrodes, the IMD may sense electrical signals from a patient and generate EEG signal(s) based on the electrical signals. Processing circuitry of the system, such as processing circuitry of an IMD or of another device configured to communicate with the IMD, may apply the EEG signal(s) to a ML model to determine one or more characteristic(s) of a brain event based, the ML model being trained on training EEG data and simulated EEG data. Determining characteristic(s) of a brain event enables a more particularized treatment to target a brain event with particular brain event characteristics.
Conventional EEG electrodes are typically positioned over a large portion of a user's scalp. While electrodes in this region are well positioned to detect electrical activity from the patient's brain, there are certain drawbacks. Sensors in this location interfere with patient movement and daily activities, making them impractical for prolonged monitoring. Additionally, implanting traditional electrodes under the patient's scalp is difficult and may lead to significant patient discomfort. To address these and other shortcomings of conventional EEG sensors, some examples of the present technology may include an IMD configured to record electrical signals at a region near the patient's head, such as adjacent a rear portion of the patient's neck or base the patient's skull or near the patient's temple. In these positions, implantation under the patient's skin is relatively simple, and a temporary application of a wearable sensor device (e.g., coupled to a bandage, garment, band, or adhesive member) does not unduly interfere with patient movement and activity. Although primarily described in the context of leadless sensor devices, in some examples, a sensor device may include electrode extensions. The electrode extensions may increase a size of a vector for sensing signals via the electrodes, such as brain and cardiac signals, and/or may position electrodes closer to a source of the brain and cardiac signals, which may enhance the sensitivity of algorithms using such signals to detect and/or predict patient conditions.
However, the EEG signals detected via electrodes disposed at or adjacent the back of a patient's neck may include relatively high noise amplitude. For example, the electrical signals associated with brain activity may be intermixed with electrical signals associated with cardiac activity (e.g., ECG signals) or signals including components associated with mechanical activity of the heart and skeletal muscle activity (e.g., EMG signals) and artifacts from other electrical sources such as patient movement or external interference. Accordingly, in some embodiments, the sensor data may be filtered or otherwise manipulated to separate the brain activity data (e.g., EEG signals) and ECG signals (or other cardiac signals) from each other and other electrical signals (e.g., EMG signals, etc.). In some examples, IMD or an external device may employ machine learning/adaptive neural network techniques to improve the signal extraction capability (e.g., to filter out or reduce the contribution of ECG signals from the EEG signals). One such methodology is described in “ECG Artifact Removal of EEG signal using Adaptive Neural Network” as published in IEEE Xplore 27 May 2019, which is hereby incorporated by reference in its entirety. Similarly, electrical signals associated with skeletal muscle activity may also be filtered from the EEG sensor data to remove such artifacts.
Aspects of the technology described herein can be embodied in a special purpose computer or data processor that is specifically programmed, configured, or constructed to perform one or more of the computer-executable instructions explained in detail herein. Aspects of the technology can also be practiced in distributed computing environments where tasks or modules are performed by remote processing devices, which are linked through a communication network (e.g., a wireless communication network, a wired communication network, a cellular communication network, the Internet, a short-range radio network (e.g., such as via Bluetooth)). In a distributed computing environment, program modules may be located in both local and remote memory storage devices.
Computer-implemented instructions, data structures, screen displays, and other data under aspects of the technology may be stored or distributed on computer-readable storage media, including magnetically or optically readable computer disks, as microcode on semiconductor memory, nanotechnology memory, organic or optical memory, or other portable and/or non-transitory data storage media. In some embodiments, aspects of the technology may be distributed over the Internet or over other networks (e.g., a Bluetooth network) on a propagated signal on a propagation medium (e.g., an electromagnetic wave(s), a sound wave) over a period of time, or may be provided on any analog or digital network (packet switched, circuit switched, or other scheme).
While conventional EEG electrodes are placed over the patient's scalp, the present technology advantageously enables recording of clinically useful brain activity data via electrodes positioned at the target region 104 at the rear of the patient's neck or head, or other cranial locations, such as temporal locations, described herein. This anatomical area is well suited to both implantation of IMD 106 and to temporary placement of a sensor device over the patient's skin. In contrast, conventional EEG electrodes positioned over the scalp are cumbersome, and implantation over the patient's skull is challenging and may introduce significant patient discomfort. As noted elsewhere here, conventional EEG electrodes are typically positioned over the scalp to more readily achieve a suitable signal-to-noise ratio for detection of brain activity. However, by using certain digital signal processing, and a special-purpose classifier algorithm, clinically useful brain activity data can be obtained using sensors disposed at the target region 104. Specifically, the electrodes can detect electrical activity that corresponds to brain activity in the P3, Pz, and/or P4 regions (as shown in
While conventional approaches to stroke and/or seizure detection utilizing EEG have relied on data from a large number of EEG electrodes, this disclosure describes that clinically useful stroke and/or seizure determinations can be made utilizing relatively few electrodes, such as via the electrodes carried by IMD 106. For example, IMD 106 may extract features from EEG signals indicative of brain activity or cardiac activity. IMD 106 may then determine whether or not the patient has experienced a stroke and/or seizure based on these extracted features. In some examples, IMD 106 takes the form of a LINQ™ Insertable Cardiac Monitor (ICM), available from Medtronic, Inc., of Minneapolis, Minnesota. The example techniques may additionally, or alternatively, be used with a medical device not illustrated in
Clinicians sometimes diagnose a patient (e.g., patient 102) with medical conditions and/or determine whether a condition of patient 102 is improving or worsening based on one or more observed physiological signals collected by physiological sensors, such as electrodes, optical sensors, chemical sensors, temperature sensors, acoustic sensors, and motion sensors. In some cases, clinicians apply non-invasive sensors to patients in order to sense one or more physiological signals while a patient is in a clinic for a medical appointment. However, in some examples, events that may change a condition of a patient, such as administration of a therapy, may occur outside of the clinic. As such, in these examples, a clinician may be unable to observe the physiological markers needed to determine whether an event, such as a stroke or seizure, has changed a medical condition of the patient and/or determine whether a medical condition of the patient is improving or worsening while monitoring one or more physiological signals of the patient during a medical appointment. In the example illustrated in
In some examples, IMD 106 includes a plurality of electrodes. The plurality of electrodes is configured to detect signals that enable processing circuitry of IMD 106 to determine current values of stroke and/or seizure metrics associated with the brain and/or cardiovascular functions of patient 102. In some examples, the plurality of electrodes of IMD 106 are configured to detect a signal indicative of an electric potential of the tissue surrounding the IMD 106. Moreover, IMD 106 may additionally or alternatively include one or more optical sensors, accelerometers, impedance sensors, respiration sensors, temperature sensors, chemical sensors, light sensors, pressure sensors, and acoustic sensors, in some examples. Such sensors may detect one or more physiological parameters indicative of a patient condition.
External device 108 may be a hand-held computing device with a display viewable by the user and an interface for providing input to external device 108 (e.g., a user input mechanism). In some examples, external device 108 may be a smartphone, smart watch, smart glasses, or other personal smart device. In some examples, external device 108 may be a smart device of patient 102. For example, external device 108 may include a small display screen (e.g., a liquid crystal display (LCD) or a light emitting diode (LED) display) that presents information to the user. In addition, external device 108 may include a touch screen display, keypad, buttons, a peripheral pointing device, voice activation, or another input mechanism that allows the user to navigate through the user interface of external device 108 and provide input. If external device 108 includes buttons and a keypad, the buttons may be dedicated to performing a certain function, e.g., a power button, the buttons and the keypad may be soft keys that change in function depending upon the section of the user interface currently viewed by the user, or any combination thereof.
In other examples, external device 108 may be a larger workstation or a separate application within another multi-function device, rather than a dedicated computing device. For example, the multi-function device may be a notebook computer, tablet computer, workstation, one or more servers, cellular phone, personal digital assistant, or another computing device that may run an application that enables the computing device to operate as a secure device.
When external device 108 is configured for use by the clinician, external device 108 may be used to transmit instructions to IMD 106. Example instructions may include requests to set electrode combinations for sensing and any other information that may be useful for programming into IMD 106. To program IMD 106, the clinician may configure and store operational parameters for IMD 106 within IMD 106 with the aid of external device 108. In some examples, external device 108 assists the clinician in the configuration of IMD 106 by providing a system for identifying potentially beneficial operational parameter values.
Whether external device 108 is configured for clinician or patient use, external device 108 is configured to communicate with IMD 106 and, optionally, another computing device (not illustrated by
Processing circuitry 110, in some examples, may include one or more processors that are configured to implement functionality and/or process instructions for execution within IMD 106. For example, processing circuitry 110 may be capable of processing instructions stored in a storage device. Processing circuitry 110 may include, for example, microprocessors, graphical processing units (GPUs), tensor processing units (TPUs), digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or equivalent discrete or integrated logic circuitry, or a combination of any of the foregoing devices or circuitry. Accordingly, processing circuitry 110 may include any suitable structure, whether in hardware, software, firmware, or any combination thereof, to perform the functions ascribed herein to processing circuitry 110.
Processing circuitry 110 may represent processing circuitry located within any one or both of IMD 106 and external device 108. In some examples, processing circuitry 110 may be entirely located within a housing of IMD 106. In other examples, processing circuitry 110 may be entirely located within a housing of external device 108. In other examples, processing circuitry 110 may be located within any one or combination of IMD 106, external device 108, and another device or group of devices that are not illustrated in
Medical device system 100A of
In some examples, IMD 106 includes one or more accelerometers. An accelerometer of IMD 106 may collect an accelerometer signal which reflects a measurement of any one or more of a motion of patient 102, a posture of patient 102 and a body angle of patient 102. In some cases, the accelerometer may collect a three-axis accelerometer signal indicative of patient 102's movements within a three-dimensional Cartesian space. For example, the accelerometer signal may include a vertical axis accelerometer signal vector, a lateral axis accelerometer signal vector, and a frontal axis accelerometer signal vector. The vertical axis accelerometer signal vector may represent an acceleration of patient 102 along a vertical axis, the lateral axis accelerometer signal vector may represent an acceleration of patient 102 along a lateral axis, and the frontal axis accelerometer signal vector may represent an acceleration of patient 102 along a frontal axis. In some cases, the vertical axis substantially extends along a torso of patient 102 when patient 102 from a neck of patient 102 to a waist of patient 102, the lateral axis extends across a chest of patient 102 perpendicular to the vertical axis, and the frontal axis extends outward from and through the chest of patient 102, the frontal axis being perpendicular to the vertical axis and the lateral axis.
IMD 106 may measure a set of parameters including an impedance (e.g., subcutaneous impedance measured via electrodes depicted in
In some examples, one or more sensors (e.g., electrodes, motion sensors, optical sensors, temperature sensors, or any combination thereof) of IMD 106 may generate a signal that indicates a parameter of a patient. In some examples, the signal that indicates the parameter includes a plurality of parameter values, where each parameter value of the plurality of parameter values represents a measurement of the parameter at a respective interval of time. The plurality of parameter values may represent a sequence of parameter values, where each parameter value of the sequence of parameter values are collected by IMD 106 at a start of each time interval of a sequence of time intervals. For example, IMD 106 may perform a parameter measurement in order to determine a parameter value of the sequence of parameter values according to a recurring time interval (e.g., every day, every night, every other day, every twelve hours, every hour, or any other recurring time interval). In this way, IMD 106 may be configured to track a respective patient parameter more effectively as compared with a technique in which a patient parameter is tracked during patient visits to a clinic, since IMD 106 is implanted within patient 102 and is configured to perform parameter measurements according to recurring time intervals without missing a time interval or performing a parameter measurement off schedule. Processing circuitry 110 may determine these different parameters separately from the stroke and/or seizure metrics or determine the stroke and/or seizure metrics based at least partially on one or more other parameter measurements.
IMD 106 may be referred to as a system or device. In one example, IMD 106 may include a memory, a plurality of electrodes carried by the housing of IMD 106, sensing circuitry configured to sense, via at least two electrodes of the plurality of electrodes, electrical signals from patient 10 and generate, based on the electrical signals, physiological information. IMD 106 may also include processing circuitry configured to receive, from the sensing circuitry, the physiological information and determine, based on the physiological information, a stroke and/or seizure metric indicative of a respective stroke and/or seizure status of the patient. The processing circuitry may be configured to then store the stroke and/or seizure metric in the memory. The housing of IMD 106 carries the plurality of electrodes and contains, or houses, both of the sensing circuitry and the processing circuitry. In this manner, IMD 106 may be referred to as a leadless sensing device because the electrodes are carried directly by the housing instead of by any leads that extend from the housing. In some examples, however, sensor device 106 may include one or more sensing leads extending therefrom and into the tissue of the patient; such lead(s) may be employed instead of or in addition to the electrodes of sensor device 106, and may perform any of the functions attributed herein to the electrodes.
The physiological data can include electrical brain activity data and/or electrical heart activity data. In some examples, the plurality of electrodes are configured to detect brain activity data corresponding to activity in at least one of a P3, Pz, or P4 brain region, which is at the back of the head or upper neck region as shown in
In some examples, IMD 106 may include a single sensing circuitry configured to generate, from the sensed electrical signals, information that includes both the electrical brain activity data (e.g., electroencephalogram (EEG) data) and the electrical heart activity data (e.g., electrocardiogram (ECG) data or cardiac contraction). In other examples, the processing circuitry of IMD 106 may include separate hardware that generates different information from the sensed electrical signals. For example, IMD 106 may include first circuitry configured to generate the electrical brain activity from the electrical signals and second circuitry different from the first circuitry and configured to generate the electrical heart activity data from the electrical signals. Even with the first and second circuitry configured to generate different information, or data, in some examples, sensed electrical signals may be conditioned or processed by one or more electrical components (e.g., filters or amplifiers) prior to being processed by the first and second circuitry. In some examples, electrical brain activity data may include features, such as spectral features, indicative of the strength of signals in various frequency bands or at various frequencies. In some examples, electrical heart activity data may include features such as the timing and/or amplitude of P-waves, R-waves, or any other features representative of heart function.
Each of the stroke and/or seizure metrics may be indicative of the likelihood (or risk) that patient 102 has experienced, or is experiencing, a stroke and/or seizure, respectively. For example, each stroke metric may include a numerical value representative of the probability that patient 102 has experienced a stroke. In some examples, a stroke and/or seizure metric may be an indication that the patient 102 has experienced a respective and/or seizure. In some examples, a stroke and/or seize metric may indicate a classification of a respective stroke and/or seizure. In some examples, a stroke and/or seize metric may indicate one or more characterizations of a respective stroke and/or seizure. IMD 106 may then compare the metric to a respective threshold or monitor a relative change in the metric value over time to determine whether or not patient 102 has experienced or is experiencing a stroke. For example, each seizure metric may include a numerical value representative of the probability that patient 102 has experienced a seizure. IMD 106 may then compare the metric to a respective threshold or monitor a relative change in the metric value over time to determine whether or not patient 102 has experienced or is experiencing a seizure. In other examples, the stroke and/or seizure metric may be a binary value that indicates no event occurred or that an event did occur. In some examples, IMD 106 may generate each stroke and/or seizure metric based on sensed data other than the sensed electrical signals from the carried electrodes on the housing of IMD 106.
In one example, IMD 106 may include one or more accelerometers within the housing. The accelerometer may be configured to generate accelerometer signal, which may be stored processed as motion/posture data, representative of posture and/or motion of patient 102. IMD 106 may then be configured to determine one or more of a posture of a patient or activity level of a patient based on the generated accelerometer signal.
In some examples, the physiological information generated from the sensed electrical signals may include ECG information. IMD 106 may extract various features from the ECG information, such as heart rate, heart rate variability, etc.
IMD 106 may generate the stroke and/or seizure metrics at the same or different frequencies. For example, for a patient who has suffered a stroke in the recent past, such as the past three months, IMD 106 may generate stroke metrics hourly or daily. In some examples, for a patient who has not suffered a stroke, IMD 106 may generate stoke metrics at longer intervals, such as daily or weekly. These time periods are examples, and the generation of stroke metrics are not limited to the periods discussed above. In some examples, these frequencies may refer to the frequency at which the sensing circuitry generates appropriate information from which the stroke and/or seizure metric is determined. In other examples, IMD 106 may continually generate physiological information from which stroke and/or seizure metrics can be determined. However, the frequency may refer to how often the processing circuitry generates the stroke and/or seizure metric from the physiological information. Continually generating physiological information may include sensing physiological signal and other generation of physiological information on a periodic and/or triggered basis without user intervention.
Each of IMDs 106 may include a respective set of electrodes and be configured to sense respective EEG signals via the respective electrode (and/or other physiological parameters via other sensors or the electrodes as described herein). In the example, illustrated in
In the example of
The configuration of housing 201 can facilitate placement either over the user's skin in a bandage-like form or for subcutaneous implantation. As such, a relatively thin housing 201 can be advantageous. Additionally, housing 201 can be flexible in some embodiments, so that housing 201 can at least partially bend to correspond to the anatomy of the patient's neck (e.g., with left and right lateral portions 207 and 209 of housing 201 bending anteriorly relative to the central portion 205 of housing 201).
In some examples, housing 201 can have a length L of from about 15 to about 50 mm, from about 20 to about 30 mm, or about 25 mm. Housing 201 can have a width W from about 2.5 to about 15 mm, from about 5 to about 10 mm, or about 7.5 mm. In some embodiments, housing 201 can have a thickness of the thickness is less than about 10 mm, about 9 mm, about 8 mm, about 7 mm, about 6 mm, about 5 mm, about 4 mm, or about 3 mm. In some embodiments, the thickness of housing 201 can be from about 2 to about 8 mm, from about 3 to about 5 mm, or about 4 mm. Housing 201 can have a volume of less than about 1.5 cc, about 1.4 cc, about 1.3 cc, about 1.2 cc, about 1.1 cc, about 1.0 cc, about 0.9 cc, about 0.8 cc, about 0.7 cc, about 0.6 cc, about 0.5 cc, or about 0.4 cc. In some embodiments, housing 201 can have dimensions suitable for implantation through a trocar introducer or any other suitable implantation technique.
As illustrated, electrodes 213 carried by housing 201 are arranged so that all three electrodes 213 do not lie on a common axis. In such a configuration, electrodes 213 can achieve a variety of signal vectors, which may provide one or more improved signals, as compared to electrodes that are all aligned along a single axis. This can be particularly useful in a sensor device 210 configured to be implanted at the neck or head while detecting electrical activity in the brain. In some embodiments, this electrode configuration also provides for improved cardiac ECG sensitivity by integrating 3 potential signal vectors. In some examples, processing circuitry may create virtual signal vectors through a weighted sum or two or more physical signal vectors, such as the physical signal vectors available from electrodes 213 of sensor device 210 or the electrodes of any other sensor device described herein.
In the example shown in
In operation, electrodes 213 are used to sense electrical signals (e.g., EEG signals and/or ECG signals) which may be submuscular or subcutaneous. The sensed electrical signals may be stored in a memory of the sensor device 210, and signal data may be transmitted via a communications link to another device (e.g., external device 108 of
Such sensor devices may include one or more extensions extending in a first, inferior direction, toward the neck or shoulders of the patient. Extensions extending in this first direction may position electrodes to facilitate cardiac signal, e.g., ECG, sensing. Such sensor devices may include one or more extensions extending in a second, superior direction, opposite the first direction, toward the upper cranium and scalp of the patient. Extensions extending in this second direction may facilitate brain signal, e.g., EEG, sensing. Each extension may include one or more electrodes to provide one or more sensing vectors of one or more orientations with another electrode on the same extension, a different extension, or a housing of the sensor device.
In the example shown in
In the example shown in
Proximal electrode 313A and distal electrode 313B are used to sense electrical signals (e.g., EEG signals, ECG signals, other brain and/or cardiac signals, or impedance) which may be submuscular or subcutaneous. Electrical signals may be stored in a memory of sensor device 310, and signal data may be transmitted via integrated antenna 326 to another medical device, which may be another implantable device or an external device, such as external device 108 (
In the example shown in
In the example shown in
Processing circuitry 402 may include fixed function circuitry and/or programmable processing circuitry. Processing circuitry 402 may include any one or more of a microprocessor, a GPU, a TPU, a controller, a DSP, an ASIC, an FPGA, or equivalent discrete or analog logic circuitry. In some examples, processing circuitry 402 may include multiple components, such as any combination of one or more microprocessors, one or more GPUs, one or more TPUs, one or more controllers, one or more DSPs, one or more ASICs, or one or more FPGAs, as well as other discrete or integrated logic circuitry. The functions attributed to processing circuitry 402 herein may be embodied as software, firmware, hardware or any combination thereof. Processing circuitry 402 may be an example of or component of processing circuitry 110 (
Sensing circuitry 406 and communication circuitry 404 may be selectively coupled to electrodes 418A-418C via switching circuitry 408, as controlled by processing circuitry 402. Sensing circuitry 406 may monitor signals from electrodes 418A-418C in order to monitor electrical activity of the brain (e.g., to produce an EEG) and/or heart (e.g., to product an ECG) from which processing circuitry 402 may generate stroke and/or seizure metrics. Sensing circuitry 406 may also sense physiological characteristics such as subcutaneous tissue impedance, the impedance being indicative of at least some aspects of patient 102's respiratory patterns and the EMG or ECG being indicative of at least some aspects of patient 102's cardiac patterns. Sensing circuitry 406 also may monitor signals from sensors 414, which may include motion sensor(s) 416, and any additional sensors, such as light detectors or optical sensors, pressure sensors, or acoustic sensors, that may be positioned on or in sensor device 400.
In some examples, a subcutaneous impedance signal collected by IMD 400 may indicate a respiratory rate and/or a respiratory intensity of patient 102 and an EGM or ECG collected by IMD 400 may indicate a heart rate of patient 102 and an atrial fibrillation (AF) burden of patient 102 or other arrhythmia. In some examples, a respiration component may additionally (using a blended sensor technique) or alternatively be sensed in other signals, such as a motion sensor signal, optical signal, or as a component (e.g., baseline shift) of the cardiac signal sensed via electrodes 418. Sensing circuitry 406 also may monitor signals from sensors 414, which may include motion sensor(s) 416, and any additional sensors, such as light detectors or pressure sensors, that may be positioned on IMD 400. Sensors 414 may also or alternatively detect heart sounds, respiration (e.g., rate or timing), impedance, or blood pressure. Therefore, sensors 414 may also or alternatively include sensors such as one or more microphones, pressure sensors, electrodes, etc. IMD 400 may utilize any of these sensors 414 to determine one or more physiological signals of the patient that may be employed to detect a stroke and/or seizure. In some examples, sensing circuitry 406 may include one or more filters and amplifiers for filtering and amplifying signals received from one or more of electrodes 418A-418C and/or motion sensor(s) 416. In some examples, sensing circuitry 406 may include separate hardware (e.g., separate circuits) configured to condition and process sensed electrical signals from which stroke and/or seizure metrics are generated. In this manner each separate circuit may perform one or more filters and amplifiers configured to extract relevant features or signal components from the sensed electrical signals. Moreover, processing circuitry 402 may selective control each separate circuit depending on whether a stroke and/or seizure metric should be generated.
Communication circuitry 404 may include any suitable hardware, firmware, software or any combination thereof for communicating with another device, such as external device 108 or another IMD or sensor, such as a pressure sensing device. Under the control of processing circuitry 402, communication circuitry 404 may receive downlink telemetry from, as well as send uplink telemetry to, external device 108 or another device with the aid of an internal or external antenna, e.g., antenna 405. In some examples, communication circuitry 404 may receive downlink telemetry from, as well as send uplink telemetry to, external device 108 or another device via tissue conductance communication (TCC) using two or more of electrodes 418, e.g., as selected by processing circuitry 402 via switching circuitry 408. In addition, processing circuitry 402 may communicate with a networked computing device via an external device (e.g., external device 108) and a computer network, such as the Medtronic CareLink® Network.
A clinician or other user may retrieve data from IMD 400 using external device 108, or by using another local or networked computing device configured to communicate with processing circuitry 402 via communication circuitry 404. The clinician may also program parameters of IMD 400 using external device 108 or another local or networked computing device.
In some examples, storage device 410 may be referred to as a memory and include computer-readable instructions that, when executed by processing circuitry 402, cause IMD 400 and processing circuitry 402 to perform various functions attributed to IMD 400 and processing circuitry 402 herein. Storage device 410 may include any volatile, non-volatile, magnetic, optical, or electrical media, such as a random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), electrically-erasable programmable ROM (EEPROM), flash memory, or any other digital media. Storage device 410 may also store data generated by sensing circuitry 406, such as physiological information, or data generated by processing circuitry 402, such as stroke and/or seizure metrics.
Power source 412 is configured to deliver operating power to the components of IMD 400. Power source 412 may include a battery and a power generation circuit to produce the operating power. In some examples, the battery is rechargeable to allow extended operation. In some examples, recharging is accomplished through proximal inductive interaction between an external charger and an inductive charging coil within external device 108. Power source 412 may include any one or more of a plurality of different battery types, such as nickel cadmium batteries and lithium ion batteries. A non-rechargeable battery may be selected to last for several years, while a rechargeable battery may be inductively charged from an external device, e.g., on a daily or weekly basis.
As described herein, IMD 400 may be configured to sense electrical signals and generate stroke and/or seizure metrics indicative of whether or not patient 102 has experienced a stroke and/or seizure. In some examples, the processing circuitry 402 is configured to analyze data from one or more electrode combinations using electrodes 418 to extract brain activity data and to discard or reduce any contribution from heart or muscle activity. In some examples, the electrodes 418 are configured to be disposed over the patient's skin. In such embodiments, the electrodes 418 can include protrusions (e.g., microneedles or other suitable structures) configured to at least partially penetrate the patient's skin so as to improve detection of subcutaneous electrical activity.
In some examples, sensing circuitry 406 senses a brain signal via electrodes 418. The brain signal may represent the electrical activity of the brain, and may be an EEG. Processing circuitry 402 may determine parameter values from the brain signal, such values determined based on magnitudes of the signal in one or more frequency bands. Sensing circuitry 406 may include filters and other circuitry to isolate the brain signal of interest.
In some examples, sensing circuitry 406 senses a cardiac signal, and processing circuitry 402 may determine parameter values from the cardiac signal. Example parameter values as described herein, such as heart rate or heart rate variability, may be determined based on detection of occurrence of cardiac beats in the cardiac signal. Sensing circuitry 406 may be configured to sense a variety of different signals within which cardiac beats may be identified and values of cardiac parameters may be determined.
For example, sensing circuitry 406 may be configured to sense a cardiac signal representing the electrical activity (e.g., depolarizations and repolarizations) of the heart, such as a subcutaneous ECG signal, via electrodes 418. As another example, sensing circuitry 406 may be configured to sense a cardiac signal representing mechanical activity of the heart via electrodes 418. A component of a signal sensed via electrodes 418, e.g., on or under the scalp of the patient, may vary based on vibration, blood flow, or impedance changes associated with cardiac contractions. Filtering to isolate this component may include 0.5 to 3 Hz bandpass filtering, although other filtering types, ranges, and cutoffs are possible. In some examples, sensing circuitry 406 may be configured to sense a cardiac signal representing mechanical activity of the heart via other sensors 414, such as optical sensors, pressure sensors, or motion sensors 416.
For example, sensing circuitry 406 and/or processing circuitry 402 may detect cardiac pulses via an optical sensor. Processing circuitry 402 may determine heart rate or heart rate variability based on the detection of cardiac pulses via the optical sensor, e.g., in combination with an ECG signal or in the absence of an ECG signal, such as if ECG signal quality is poor. An optical sensor signal may additionally or alternatively be used for other purposes, such as to determine blood oxygenation, local tissue perfusion, or blood pressure, any of which may be useful for detection or prediction of stroke and/or discrimination of ischemic and hemorrhagic stroke.
One or more electrodes 418 may be positioned, e.g., during implantation of sensor device 400, to facilitate sensing of a cardiac signal via the electrodes. In some examples, sensor device 400 may include one or more electrode extensions 265, 272, 276, 284, 285, 286 to facilitate positioning of one or more electrodes 418, e.g., via tunneling under the scalp, at desired locations for sensing the brain and/or cardiac signals. Desired locations for sensing brain and cardiac signals using electrodes 418 may be determined prior to implantation of sensor device 406 for a particular patient using external sensing equipment, such as standard multi-electrode ECG and EEG equipment, either on the particular patient, or experimentally on a number of subjects. In some examples, the one or more housing-based electrodes 418 of sensor device 400 are positioned at a desired location for sensing a brain signal and the one or more extension-based electrodes 418 are positioned at a desired location for sensing a cardiac signal, or vis-a-versa. With reference to
In some examples, processing circuitry 402 may utilize both electrical, e.g., ECG, and pulsatile cardiac signals in an integrated fashion for the detection, prediction, and/or classification of conditions. In some examples, such integration may result in an “enhanced” ECG signal. For example, processing circuitry 402 may identify features within an ECG signal based on the timing of pulses in a pulsatile signal. In some examples, processing circuitry 402 may account for a delay in pulsatile timing relative to the ECG in such integration.
For example, an optical sensor signal (e.g., a photoplethysmographic signal) can be used as a timing base for ensemble averaging or other means to improve the signal-to-noise ratio for a cardiac signal. The optical sensor signal can therefore be considered a surrogate cardiac signal and/or be used to derive an enhanced cardiac signal, which may be particularly useful when the ECG has poor quality. A first or second derivative of an optical sensor signal can be used as a trigger for ensemble averaging, e.g., the ECG signal, by, for example, determining the time associated with a maximum/minimum value of the first or second derivative and/or a zero-crossing of the first or second derivative. Sharp, high-frequency points can be used as trigger points to increase the resolution of the ensemble signal, whereas lower-frequency trigger points may smear or distort the ensemble average. The cardiac waveforms that are aligned with the trigger points can be stored and averaged to generate the ensemble signal.
Processing circuitry 402 may be configured to calculate physiological characteristics relating to one or more electrical signals received from the electrodes 418, such as stroke and/or seizure metrics. For example, processing circuitry 402 may be configured to algorithmically determine the presence or absence of a stroke and/or seizure (via generation of a stroke and/or seizure metric) or other neurological condition from the electrical signal. In certain examples, processing circuitry 402 may make a stroke and/or seizure determination for each electrode 418 (e.g., channel) or may make a stroke and/or seizure determination using electrical signals acquired from two or more selected electrodes 418.
In some examples, processing circuitry 402 may employ patient movement information as a part of stroke and/or seizure detection. For example, motion sensor 416 may include one or more accelerometers discussed above. In some examples, motion sensors 46 may be configured to detect patient posture, patient activity, and/or patient movement, which includes detection of patient falling.
In some examples, sensing circuitry 406 senses electrical signals from the patient. Sensing circuitry 406 may sense these electrical signals from a sensing vector determined by the electrodes 418 selected for sensing. In this manner, sensing circuitry 406 may use different vectors (e.g., different electrode combinations) in order to obtain different electrical information from the patient. Sensing circuitry 406 may generate one or more EEG signals based on the sensed electrical signals. Generating one or more EEG signals may include various filtering, amplification, transforms, digitization, or any other conditioning and processing that generates an EEG signal can be analyzed by processing circuitry 402.
Processing circuitry 402 may apply EEG signal(s) to a ML model to determine one or more characteristics of a brain event, the ML model being trained on training EEG data and simulated EEG data.
In some examples, processing circuitry 402 may generate a four-dimensional map of the brain of the patient based, at least in part, on the EEG signal(s), and determine whether the ML model has coverage gaps using the generated four-dimensional map. In some examples, processing circuitry 402 may generate additional simulated EEG data corresponding to the coverage gaps to fill in the coverage gaps and send the additional simulated EEG data to further train the ML model to update the ML model, and apply the EEG signal(s) to the updated ML model to determine characteristic(s) of the brain event.
Processing circuitry 502, in one example, may include one or more processors that are configured to implement functionality and/or process instructions for execution within external device 500. For example, processing circuitry 502 may be capable of processing instructions stored in storage device 510. Processing circuitry 502 may include, for example, microprocessors, GPUs, TPUs, DSPs, ASICs, FPGAs, or equivalent discrete or integrated logic circuitry, or a combination of any of the foregoing devices or circuitry. Accordingly, processing circuitry 502 may include any suitable structure, whether in hardware, software, firmware, or any combination thereof, to perform the functions ascribed herein to processing circuitry 502.
Communication circuitry 504 may include any suitable hardware, firmware, software or any combination thereof for communicating with another device, such as IMD 400. Under the control of processing circuitry 502, communication circuitry 504 may receive downlink telemetry from, as well as send uplink telemetry to, IMD 400, or another device. In other examples, communication circuitry 504 may also employ TCC for communicating with other devices.
Storage device 510 may be configured to store information within external device 500 during operation. Storage device 510 may include a computer-readable storage medium or computer-readable storage device. In some examples, storage device 510 includes one or more of a short-term memory or a long-term memory. Storage device 510 may include, for example, RAM, dynamic random access memories (DRAM), static random access memories (SRAM), magnetic discs, optical discs, flash memories, or forms of electrically programmable memories (EPROM) or EEPROM. In some examples, storage device 510 is used to store data indicative of instructions for execution by processing circuitry 502. Storage device 510 may be used by software or applications running on external device 500 to temporarily store information during program execution. In some examples, storage device 510 may include an artificial intelligence model 512 (e.g., machine learning, neural networks, etc.) stored on the storage device 510.
Data exchanged between external device 500 and IMD 400 may include operational parameters. External device 500 may transmit data including computer readable instructions which, when implemented by IMD 400, may control IMD 400 to change one or more operational parameters and/or export collected data. For example, processing circuitry 502 may transmit an instruction to IMD 400 which requests IMD 400 to export collected data (e.g., data corresponding to one or more of the physiological information, stroke metrics, seizure metrics, or accelerometer signal) to external device 500. In turn, external device 500 may receive the collected data from IMD 400 and store the collected data in storage device 510. Additionally, or alternatively, processing circuitry 502 may export instructions to IMD 400 requesting IMD 400 to update electrode combinations for stimulation or sensing.
A user, such as a clinician or patient 102, may interact with external device 500 through user interface 506. User interface 506 includes a display (not shown), such as an LCD or LED display or other type of screen, with which processing circuitry 502 may present information related to IMD 400 (e.g., stroke and/or seizure metrics). In addition, user interface 506 may include an input mechanism to receive input from the user. The input mechanisms may include, for example, any one or more of buttons, a keypad (e.g., an alphanumeric keypad), a peripheral pointing device, a touch screen, or another input mechanism that allows the user to navigate through user interfaces presented by processing circuitry 502 of external device 500 and provide input. In other examples, user interface 506 also includes audio circuitry for providing audible notifications, instructions or other sounds to patient 102, receiving voice commands from patient 102, or both. Storage device 510 may include instructions for operating user interface 506 and for managing power source 508.
Power source 508 is configured to deliver operating power to the components of external device 500. Power source 508 may include a battery and a power generation circuit to produce the operating power. In some examples, the battery is rechargeable to allow extended operation. Recharging may be accomplished by electrically coupling power source 508 to a cradle or plug that is connected to an alternating current (AC) outlet. In addition, recharging may be accomplished through proximal inductive interaction between an external charger and an inductive charging coil within external device 500. In other examples, traditional batteries (e.g., nickel cadmium or lithium ion batteries) may be used. In addition, external device 500 may be directly coupled to an alternating current outlet to operate.
In some examples, external device 500 may provide an alert to the patient or another entity (e.g., a call center) based on a stroke and/or seizure indication provided by IMD 400. In some examples, user interface 506 may provide an interface for presenting an alert of the detection, prediction, or classification of the condition, e.g., stroke, and for a user, e.g., the patient, a caregiver, or a clinician, to provide input overriding the detection, prediction, or classification. In this manner, systems as described herein may avoid unnecessary emergency activity resulting from a false detection by the system.
In some examples, external device 500 may receive patient data, such as one or more of the EEG signal, the accelerometer signal, and/or clock signal, such as from IMD 400, to further adjudicate a generated stroke metric satisfying a stroke criteria threshold in IMD 400. For example, external device 500 may confirm or deny whether a stroke was detected by IMD 400, confirm or deny what type of stroke was detected by IMD 400, and/or confirm or deny a location of stroke that was detected by IMD 400. In some examples, external device 500 may receive patient data, such as one or more of the EEG signal, the accelerometer signal, and/or clock signal, such as from IMD 400, to further adjudicate a generated seizure metric satisfying a seizure criteria threshold in IMD 400. For example, external device 500 may confirm or deny whether a seizure was detected by IMD 400, confirm or deny what type of seizure was detected by IMD 400, and/or confirm or deny a location of seizure that was detected by IMD 400. In some examples, external device may apply an artificial intelligence model 512 (e.g., machine learning, neural networks, etc.) stored in storage device 510 to the received patient data to confirm or deny whether a stroke or seizure was detected by IMD 400 and/or confirm or deny what type of stroke or seizure was detected by IMD 400, and/or confirm or deny a location of stroke or seizure that was detected by IMD 400.
Access point 600 may include a device that connects to network 602 via any of a variety of wired or wireless network connections, such as telephone dial-up, digital subscriber line (DSL), or cable modem connections. In other examples, access point 600 may be coupled to network 602 through different forms of connections, including wired or wireless connections. In some examples, access point 600 may be a user device, such as a tablet or smartphone, that may be co-located with the patient. As discussed above, IMD 106 may be configured to transmit data, such as any one or combination of an EEG signal, an accelerometer signal, and a tissue impedance signal to external device 108. In addition, access point 600 may interrogate IMD 106, such as periodically or in response to a command from the patient or network 602, in order to retrieve parameter values determined by processing circuitry of IMD 106, or other operational or patient data from IMD 106. Access point 600 may then communicate the retrieved data to server 604 via network 602.
In some cases, server 604 may be configured to provide a secure storage site for data that has been collected from IMD 106, and/or external device 108. In some cases, server 604 may assemble data in web pages or other documents for viewing by trained professionals, such as clinicians, via computing devices 610A-610N. One or more aspects of the illustrated system of
Server 604 may include processing circuitry 606. Processing circuitry 606 may include fixed function circuitry and/or programmable processing circuitry. Processing circuitry 606 may include any one or more of a microprocessor, a GPU, a TPU, a controller, a DSP, an ASIC, an FPGA, or equivalent discrete or analog logic circuitry. In some examples, processing circuitry 606 may include multiple components, such as any combination of one or more microprocessors, one or more controllers, one or more DSPs, one or more ASICs, or one or more FPGAs, as well as other discrete or integrated logic circuitry. The functions attributed to processing circuitry 606 herein may be embodied as software, firmware, hardware or any combination thereof.
In some examples, processing circuitry 606 may perform one or more techniques described herein based on an EEG signal, EGM signal, impedance signal, an accelerometer signal, or other sensor signals received from IMD 106, or parameter values determined based on such signals by IMD 106 and received from IMD 106, as examples. For example, processing circuitry 606 may perform one or more of the techniques described herein to apply EEG signal(s) to a ML model to determine one or more characteristics of a brain event, the ML model being trained on training EEG data and simulated EEG data. For example, processing circuitry 606 may perform one or more of the techniques described herein to generate a four-dimensional map of the brain of the patient based, at least in part, on the EEG signal(s), and determine whether the ML model has coverage gaps using the generated four-dimensional map. In some examples, processing circuitry 402 may generate additional simulated EEG data corresponding to the coverage gaps to fill in the coverage gaps and send the additional simulated EEG data to further train the ML model to update the ML model, and apply the EEG signal(s) to the updated ML model to determine characteristic(s) of the brain event.
Server 604 may include memory 608. Memory 608 includes computer-readable instructions that, when executed by processing circuitry 606, cause IMD 106 and processing circuitry 606 to perform various functions attributed to IMD 106 and processing circuitry 606 herein. Memory 608 may include any volatile, non-volatile, magnetic, optical, or electrical media, such as RAM, ROM, NVRAM, EEPROM, flash memory, or any other digital media.
In some examples, one or more of computing devices 610A-610N (e.g., device 610A) may be a tablet or other smart device located with a clinician, by which the clinician may program, receive alerts from, and/or interrogate IMD 106. For example, the clinician may access data corresponding to any one or combination of sensed physiological signals, an accelerometer signal, stroke metrics, seizure metrics, and other types of signals collected by IMD 106, or parameter values determined by IMD 106 based on such signals, through device 610A, such as when patient 102 is in between clinician visits, to check on a status of a medical condition, such as a status of the brain event characteristic(s). In some examples, the clinician may enter instructions for a medical intervention for patient 102 into an app in device 610A, such as based on a status of a patient condition determined by IMD 106, external device 108, processing circuitry 110, or any combination thereof, or based on other patient data known to the clinician. Device 610A then may transmit the instructions for medical intervention to another of computing devices 610A-610N (e.g., device 610B) located with patient 102 or a caregiver of patient 102. For example, such instructions for medical intervention may include an instruction to change a drug dosage, timing, or selection, to schedule a visit with the clinician, or to seek medical attention. In further examples, device 610B may generate an alert to patient 102 based on a status of a medical condition of patient 102 determined by IMD 106, which may enable patient 102 proactively to seek medical attention prior to receiving instructions for a medical intervention. In this manner, patient 102 may be empowered to take action, as needed, to address his or her medical status, which may help improve clinical outcomes for patient 102.
In further examples, device 610B may be configured to transmit alert messages to computing devices 610C associated with one or more care providers via network 602. Care providers may include emergency medical systems (EMS) and hospitals, and may include particular departments within a hospital, such as an emergency department, catheterization lab, a stroke response department, or seizure response department. In further examples, device may device 610B may provide patient-specific care recommendations (e.g., potential drug therapy for prevention or therapy of stroke and/or seizure). The ability of the system to detect the stroke and/or seizure with adequate sensitivity and specificity may, for example, guide EMS care giver to what they can expect when they arrive on the scene and how best to treat the presenting or soon to present stroke and/or seizure.
Processing circuitry 402 receives the EEG signal(s) from sensing circuitry 406 (704). Processing circuitry 402 applies the EEG signal(s) to an artificial intelligence model (e.g., machine learning, neural networks, etc.) to determine one or more characteristic(s) of a brain event (706). While the following description will refer to the artificial intelligence model as a machine learning (ML) model, other types of artificial intelligence models may be used in place of the ML model.
Due to an insufficient amount of training EEG data to detect a variety of types, locations, magnitudes, and/or times of brain events, a ML model trained on training EEG data, by itself, may have coverage gaps (e.g., “non-coverage” areas) in which the ML model is not able to accurately identify particular types, locations, magnitudes (e.g., infarct volume/size), and/or times of brain events while just being trained on the training EEG data itself and/or training data. In the techniques described herein, the ML model is trained on training EEG data and simulated EEG data. In some examples, particular types of simulated EEG data may be selected to supplement the training EEG data to train the ML to fill in the coverage gaps of the ML model being trained on training EEG data itself. In some examples, simulated EEG data may be generated by making a four-dimensional (x, y, z, and time) head model of EEG potentials over time over a mesh of many points on the head, using RBF interpolation to create a high-resolution map from a set of electrodes on the head, such as a limited set of electrodes Then, two points on that map may be chosen that are close to the location of an implanted device, such as IMD 106, and the difference of potentials for those two points are taken for every timestep to generate a new, simulated EEG data.
In some examples, the simulated EEG data may be generated by taking the bipolar difference of two virtual points, which approximate the location of electrodes in an implanted device, such as IMD 106, on a high-resolution four-dimensional head map of electrical potentials over time, generated by interpolating EEG data sampled at discrete electrode locations onto a human head mesh.
In some examples, characteristics of a brain event may include one or more of a type of brain event, such as a type of stroke or a type of seizure, a brain event location in the brain, such as a location of a stroke or a location of a seizure in the brain, a last known normal (LKN) time before onset of the brain event, a time of the onset of the brain event, and/or infarct volume/size of a stroke. In some examples, a type of stroke may include one or more of ischemic stroke, hemorrhagic stroke, cryptogenic stroke, stroke mimic/non-stroke.
In some examples, processing circuitry 402 may store the brain event characteristic(s) in memory (708). In some examples, processing circuitry 402 may control communication circuitry to transmit the brain event characteristic(s) information to external device 108 (710).
Brain synaptic input leads to ionic current flow and because of the brain's high local lateral connectivity, small regions of tissue will share excitation patterns. These regions of shared excitation may be approximated electrically as Equivalent Current Dipoles (ECD). In some examples, dipole models may be used to generate 4D maps, such as 900, 910, as shown in
A ML model using training EEG data, by itself, may have coverage gaps (e.g., “non-coverage” areas) in which the ML model is not able to accurately identify particular types, locations, magnitudes (e.g., infarct volume/size), and/or times of brain events while being trained on the training EEG data itself. In some examples, ML models using training EEG data and insufficient simulated data may have coverage gaps.
Processing circuitry 402 may use the 4D map to determine whether there are coverage gaps in the ML model (804). In some examples, coverage gaps may be compared to a coverage gap threshold. If the processing circuitry determines coverage gaps satisfy the coverage gap threshold processing circuitry 402 determines the currently trained ML model has coverage gaps. If processing circuitry determines the coverage gaps do not satisfy the coverage gap threshold (e.g., small or do not exist) processing circuitry 402 determines the currently trained ML model does not have coverage gaps.
For example, if the determined coverage gaps satisfy the coverage gap threshold, processing circuitry 402 determines the machine learning has coverage gaps and moves to item (806), and if the determined coverage gaps do not satisfy the coverage gap threshold, processing circuitry 402 determines the machine learning does not have coverage gaps and moves to item (805).
In some examples, if processing circuitry 402 determines the machine learning does not have coverage gaps and moves to item (“NO” branch of block 804) because processing circuitry 402 determined coverage gaps do not satisfy the coverage gap threshold, processing circuitry 402 may determine the one or more characteristics of the brain event based, at least in part, on the ML model (805).
In some examples, if processing circuitry 402 determines the machine learning does have coverage gaps and moves to item (“YES” branch of block 804) because processing circuitry 402 determined coverage gaps satisfies the coverage gap threshold, processing circuitry 402 may generate additional simulated EEG data that correspond to the coverage gaps to fill in the coverage gaps of the training EEG data and/or initial simulated EEG data (806). Processing circuitry 402 may send the generated additional simulated EEG data to further train the ML model to update the ML model to a model that has greater coverage than the initial ML model with coverage gaps due to limited training data (808). Processing circuitry 402 may apply the EEG signal(s) of the patient to the updated ML model (810). Processing circuitry may determine the one or more characteristics of the brain event based, at least in part, on the updated ML model (812).
When processing circuitry 402 transmits the brain event characteristic(s) to an external device, the external device may be associated with emergency services in some examples. In some examples, the external device may include global position system (GPS) capability or other location detection technology (e.g., WiFi triangulation) such that the external device can identify, store, and/or communicate the geographic location at which the brain event occurred. The external device may then transmit the location information and/or brain event characteristic(s) to another device or system via cell phone tower, satellite, or other technology. The other system may be an emergency service such as 911 or other medical service. If the technique of
It should be understood that various aspects disclosed herein may be combined in different combinations than the combinations specifically presented in the description and accompanying drawings. It should also be understood that, depending on the example, certain acts or events of any of the processes or methods described herein may be performed in a different sequence, may be added, merged, or left out altogether (e.g., all described acts or events may not be necessary to carry out the techniques). In addition, while certain aspects of this disclosure are described as being performed by a single module, unit, or circuit for purposes of clarity, it should be understood that the techniques of this disclosure may be performed by a combination of units, modules, or circuitry associated with, for example, a medical device. Also, the techniques could be fully implemented in one or more circuits or logic elements. The techniques of this disclosure may be implemented in a wide variety of devices or apparatuses, including an IMD, an external programmer, a combination of an IMD and external programmer, an integrated circuit (IC) or a set of ICs, and/or discrete electrical circuitry, residing in an IMD and/or external programmer.
In one or more examples, the described techniques may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored as one or more instructions or code on a computer-readable medium and executed by a hardware-based processing unit. Computer-readable media may include non-transitory computer-readable media, which corresponds to a tangible medium such as data storage media (e.g., RAM, ROM, EEPROM, flash memory, or any other medium that can be used to store desired program code in the form of instructions or data structures and that can be accessed by a computer).
Instructions may be executed by one or more processors, such as one or more digital signal processors (DSPs), general purpose microprocessors, application specific integrated circuits (ASICs), field programmable logic arrays (FPGAs), or other equivalent integrated or discrete logic circuitry. Accordingly, the term “processor” or “processing circuitry” as used herein may refer to any of the foregoing structure or any other physical structure suitable for implementation of the described techniques. Also, the techniques could be fully implemented in one or more circuits or logic elements.
The following examples are illustrative of the techniques described herein. Various examples have been described. These and other examples are within the scope of the following claims.
Example 1. A system comprising: a memory; a plurality of electrodes; sensing circuitry configured to: sense, via at least two electrodes of the plurality of electrodes, electrical signals from a patient; and generate, based on the electrical signals, one or more electroencephalography (EEG) signals; and processing circuitry configured to: receive, from the sensing circuitry, one or more EEG signals; and apply the one or more EEG signals to a machine learning (ML) model to determine one or more characteristics of a brain event, the ML model being trained on training EEG data and simulated EEG data.
Example 2. The system of claim 1, wherein the one or more characteristics of the brain event includes a type of stroke.
Example 3. The system of claim 2, wherein the type of stroke includes one or more of an ischemic stroke, hemorrhagic stroke, cryptogenic stroke, or stroke mimic.
Example 4. The system of any of claims 1-3, wherein the one or more characteristics of the brain event includes a location of the brain event in a brain of the patient.
Example 5. The system of any of claims 1-4, wherein the one or more characteristics of the brain event include a magnitude of the brain event.
Example 6. The system of any of claims 1-5, wherein the simulated EEG data fills in coverage gaps of the training EEG data.
Example 7. The system of any of claims 1-6, wherein the processing circuitry is configured to: apply the one or more EEG signals to the ML model to generate a four-dimensional map of the brain of the patient.
Example 8. The system of claim 7, wherein the processing circuitry is configured to: determine, using the four-dimensional map, whether a coverage gap of the training EEG data and the simulated EEG data satisfy a gap threshold; in response to determining the coverage gap satisfies the gap threshold, generate additional simulated EEG data to fill in the coverage gap; send the additional simulated EEG data to further train and update the ML model; and apply the one or more EEG signals to the updated ML model to determine the one or more characteristics of the brain event. Send the determined the one or more brain event characteristics of the brain event based, at least in part, on the updated generated four-dimensional map.
Example 9. The system of any of claims 7-8, wherein the processing circuitry is configured to generate the four-dimensional map of the brain of the patient using a dipole model.
Example 10. The system of any of claims 1 and 4-9, wherein the brain event comprises a stroke, a brain ischemic event, a brain hypoxia event, or a seizure.
Example 11. A method comprising: sensing, by sensing circuitry and via at least two electrodes of a plurality of electrodes, electrical signals from a patient; generating, by the sensing circuitry and based on the electrical signals, one or more electroencephalography (EEG) signals; receiving, by processing circuitry and from the sensing circuitry, the one or more EEG signals; and applying the one or more EEG signals to a machine learning (ML) model to determine one or more characteristics of a brain event, the ML model being trained on training EEG data and simulated EEG data.
Example 12. The method of claim 11, wherein the one or more characteristics of the brain event includes a type of stroke.
Example 13. The method of claim 12, wherein the type of stroke includes one or more of an ischemic stroke, hemorrhagic stroke, cryptogenic stroke, or stroke mimic.
Example 14. The method of any of claims 11-13, wherein the one or more characteristics of the brain event includes a location of the brain event in a brain of the patient.
Example 15. The method of claim 12, wherein the one or more characteristics of the brain event include a magnitude of the brain event.
Example 16. The method of any of claims 11-15, wherein the simulated EEG data fills in coverage gaps of the training EEG data.
Example 17. The method of any of claims 11-16 further comprising: applying the one or more EEG signals to the ML model to generate a four-dimensional map of the brain of the patient.
Example 18. The method of claim 17 further comprising: determining, using the four-dimensional map, whether a coverage gap of the training EEG data and the simulated EEG data satisfy a gap threshold; in response to determining the coverage gap satisfies the gap threshold, generating additional simulated EEG data to fill in the coverage gap; sending the additional simulated EEG data to further train and update the ML model; and applying the one or more EEG signals to the updated ML model to determine the one or more characteristics of the brain event. Sending the determine the one or more brain event characteristics of the brain event based, at least in part, on the updated generated four-dimensional map.
Example 19. The method of any of claims 17-18 further comprising: generating the four-dimensional map of the brain of the patient using a dipole model.
Example 20. The method of claim 18 further comprising: determining optimized electrode separation and implant location using the four-dimensional map of the brain of the patient.
Example 21 The method of any of claims 11 and 14-20, wherein the one or more brain event characteristics include a magnitude of the brain event.
Example 22. The method of any of claims 11-21, wherein the brain event comprises a stroke, a brain ischemic event, a brain hypoxia event, or a seizure.
Example 23. A computer-readable medium comprising instructions that, when executed, cause processing circuitry to execute any of the methods recited in claims 11-21.
This application claims the benefit of and priority to U.S. Provisional Patent Application Ser. No. 63/516,183 filed Jul. 28, 2023, the entire disclosure of which is incorporated by reference herein.
Number | Date | Country | |
---|---|---|---|
63516183 | Jul 2023 | US |