Automated adjustment monitoring using deep learning

Abstract

In some examples, a system is configured to use a neural network algorithm of a cerebral autoregulation model to determine a cerebral autoregulation status of a patient based at least in part on blood pressure of the patient over a period of time and regional cerebral oxygen saturation of the patient over the period of time.

Description

[0001] This application claims priority to U.S. Patent Application No. 17 / 210,222, filed March 23, 2021, entitled “Automation Monitoring Using Deep Learning,” the entire contents of which are incorporated herein by reference. Technical Field

[0002] This disclosure relates to monitoring the autoregulation status of patients. Background Technology

[0003] Clinicians can monitor one or more physiological parameters in a patient, for example, to monitor their autoregulation status. Autoregulation is a response mechanism by which an organism regulates blood flow across a wide range of systemic blood pressure variations through complex myogenic, neurogenic, and metabolic mechanisms. During autoregulation, small arteries dilate or constrict in an attempt to maintain adequate blood flow. Autoregulation can occur in various organs and organ systems, such as the brain, kidneys, and gastrointestinal tract. In an example of brain autoregulation, when intracranial pressure decreases, cerebral arterioles dilate to attempt to maintain blood flow. When intracranial pressure increases, cerebral arterioles constrict to reduce blood flow that could potentially cause brain damage. Summary of the Invention

[0004] This disclosure describes example apparatus, systems, and techniques for using machine learning to determine a patient's brain autoregulation state. For example, a system may be configured to receive a blood pressure signal indicating a patient's blood pressure over a time period and an oxygen saturation signal indicating the patient's regional brain oxygen saturation over the same time period. The system may input the patient's blood pressure and regional brain oxygen saturation, along with additional patient-associated data, into a brain autoregulation model to determine the patient's brain autoregulation state.

[0005] The brain autoregulation model may include a neural network algorithm trained via machine learning using training data from a patient population. This training data may include blood pressure and regional brain oxygen saturation data from the patient population, along with labeled ground truth values, enabling the neural network algorithm to learn the relationship between the patient's blood pressure and regional brain oxygen saturation, and the association between this relationship and the state of brain autoregulation.

[0006] In some examples, the training data used to train the neural algorithm may also include additional data derived from blood pressure data and regional brain oxygen saturation data, such as gradients of the blood pressure data, gradients of the regional brain oxygen saturation data, and a brain oxygenation index specifying the correlation between the blood pressure data and the regional brain oxygen saturation data. Furthermore, in some examples, the training data may also include additional data such as bypass markers used to indicate whether a patient is undergoing cardiopulmonary bypass surgery.

[0007] Compared to mechanical algorithms that may be based solely on a patient's blood pressure data, the technology disclosed herein enables brain autoregulation monitoring devices to more accurately determine a patient's brain autoregulation state and reduces false positives and false negatives by using a brain autoregulation model that includes a neural network algorithm trained via machine learning to determine the patient's brain autoregulation state. Therefore, the technology disclosed herein offers technical advantages.

[0008] In one example, this disclosure describes a method comprising: receiving a blood pressure signal indicating a patient's blood pressure over a time period and an oxygen saturation signal indicating the patient's regional brain oxygen saturation over the time period; determining the patient's brain autoregulation state based at least in part on the patient's blood pressure and the patient's regional brain oxygen saturation over the time period using a neural network algorithm of a brain autoregulation model; and sending a signal indicating the patient's brain autoregulation state to an output device.

[0009] In another example, this disclosure describes a system comprising: a blood pressure sensing device; an oxygen saturation sensing device; and a processing circuit system configured to: receive from the blood pressure sensing device a blood pressure signal indicating a patient's blood pressure over a certain time period, and from the oxygen saturation sensing device an oxygen saturation signal indicating the patient's regional brain oxygen saturation over the time period; determine the patient's brain autoregulation state, at least in part, based on the patient's blood pressure and the patient's regional brain oxygen saturation over the time period, using a neural network algorithm of a brain autoregulation model; and send a signal indicating the patient's brain autoregulation state to an output device.

[0010] In another example, this disclosure describes a non-transitory computer-readable and storable medium comprising instructions that, when executed by a processing circuitry system, cause the processing circuitry system to: receive a blood pressure signal indicating a patient's blood pressure over a time period and an oxygen saturation signal indicating the patient's regional brain oxygen saturation over the time period; determine the patient's brain autoregulation state using a neural network algorithm of a brain autoregulation model, at least in part based on the patient's blood pressure and the patient's regional brain oxygen saturation over the time period; and send a signal indicating the patient's brain autoregulation state to an output device.

[0011] In another example, this disclosure describes an apparatus comprising: means for receiving a blood pressure signal indicating a patient's blood pressure over a time period and an oxygen saturation signal indicating a patient's regional brain oxygen saturation over the time period; means for determining a patient's brain autoregulation state based at least in part on the patient's blood pressure and the patient's regional brain oxygen saturation over the time period using a neural network algorithm of a brain autoregulation model; and means for transmitting a signal indicating the patient's brain autoregulation state to an output device.

[0012] Details of one or more examples are set forth in the accompanying drawings and the description below. Other features, objects, and advantages will become apparent from this specification, the accompanying drawings, and the claims. Attached Figure Description

[0013] Figure 1 It shows a conceptual block diagram of an example brain autoregulation monitoring system.

[0014] Figure 2 It demonstrates what can be done to Figure 1 Details of an example training system for performing training on the brain autoregulation model are shown.

[0015] Figure 3 Showing Figure 1 and Figure 2 An example deep learning architecture for an example brain autoregulation model.

[0016] Figure 4 Showing Figure 1 and Figure 2 An example deep learning architecture for an example brain autoregulation model.

[0017] Figure 5 Demonstrated the use of Figure 3 and Figure 4 Example graphs showing the results of classifying the brain's autoregulation state using a deep learning architecture.

[0018] Figure 6An example user interface that includes information on brain autoregulation is shown.

[0019] Figure 7 This is a flowchart illustrating an example method for monitoring a patient's brain autoregulation state.

[0020] Figure 8 This is a flowchart illustrating an example method for monitoring a patient's brain autoregulation state. Detailed Implementation

[0021] Figure 1 This is a conceptual block diagram illustrating an example brain autoregulation monitoring system 100. (For example...) Figure 1 As shown, the brain autoregulation monitoring system 100 includes a processing circuit system 110, a memory 120, a control circuit system 122, a user interface 130, sensing circuit systems 140 and 142, and sensing devices 150 and 152. Figure 1 In the example shown, the user interface 130 may include a display 132, an input device 134, and / or a speaker 136, which may be any suitable audio device configured to generate and output noise and may include any suitable circuitry. In some examples, the autoregulation monitoring system 100 may be configured to determine and output (e.g., displayed on the display 132) the autoregulation state of the patient 101, for example, during medical procedures or for longer-term monitoring (such as in the intensive care unit (ICU) or for fetal monitoring). Clinicians may receive information about the patient's autoregulation state via the user interface 130 and adjust treatment or therapy for the patient 101 based on the autoregulation state information.

[0022] When patient 101 exhibits impaired autoregulation, they may experience inappropriate blood flow, which can be undesirable. Patient 101's autoregulation system may be impaired if the blood pressure gradient and oxygen saturation gradient show a consistent trend over a period of time (e.g., changing in the same direction). Patient 101's intact autoregulation occurs within a certain blood pressure range defined between the lower limit of autoregulation (LLA) and the upper limit of autoregulation (ULA). For example, below the lower limit of autoregulation (LLA), decreased blood flow to the corresponding organ may lead to local ischemia and adverse effects on the organ. Above the upper limit of autoregulation (ULA), increased blood flow to the corresponding organ may lead to congestion, which may cause swelling or edema of the organ. For example, during medical procedures, clinicians can monitor a patient's autoregulation and take one or more measures to maintain or keep the patient in an intact autoregulatory state, such as by raising or lowering the patient's blood pressure.

[0023] The processing circuit system 110 described herein, along with other processors, processing circuit systems, controllers, control circuit systems, etc., may include one or more processors. The processing circuit system 110 may include any combination of integrated circuit systems such as one or more microprocessors, digital signal processors (DSPs), application-specific integrated circuit systems (ASICs), or field-programmable gate arrays (FPGAs), discrete logic circuit systems, and analog circuit systems. In some examples, the processing circuit system 110 may include multiple components, such as one or more microprocessors, one or more DSPs, one or more ASICs, or one or more FPGAs, and any combination of other discrete or integrated logic circuit systems and / or analog circuit systems.

[0024] Control circuitry system 122 may be operatively coupled to processing circuitry system 110. Control circuitry system 122 is configured to control the operation of sensing devices 150 and 152. In some examples, control circuitry system 122 may be configured to provide timing control signals to coordinate the operation of sensing devices 150 and 152. For example, sensing circuitry systems 140 and 142 may receive one or more timing control signals from control circuitry system 122, which they may use to turn the corresponding sensing devices 150 and 152 on and off, such as to periodically collect calibration data using sensing devices 150 and 152. In some examples, processing circuitry system 110 may use timing control signals to synchronize operation with sensing circuitry systems 140 and 142. For example, processing circuitry system 110 may synchronize the operation of analog-to-digital converters and demultiplexers with sensing circuitry systems 140 and 142 based on timing control signals.

[0025] Memory 120 can be configured to store, for example, monitored physiological parameter values ​​of patient 101, such as blood pressure values, oxygen saturation values, regional brain oxygen saturation (rSO2) values, one or more brain autoregulation state values, one or more non-brain autoregulation state values, physiological parameters, mean arterial pressure (MAP) values, or any combination thereof. Memory 120 can also be configured to store these data, such as autoregulation state values ​​including modified and unmodified values, thresholds and rates, smoothing functions, Gaussian filters, confidence measures, expected autoregulation functions, historical patient blood pressure data, and / or estimates of autoregulation limits. Thresholds and rates, smoothing functions, Gaussian filters, confidence measures, expected autoregulation functions, and historical patient blood pressure data can remain constant throughout the use of device 100 and among multiple patients, or these values ​​can vary over time.

[0026] In some examples, memory 120 may store program instructions, such as neural network algorithms. The program instructions may include one or more program modules executable by processing circuitry system 110. For example, memory 120 may store a brain autoregulation model 124, which may be a model trained via machine learning to determine the brain autoregulation state of patient 101. When executed by processing circuitry system 110, these program instructions (such as program instructions for brain autoregulation model 124) enable processing circuitry system 110 to provide the functionality assigned to it herein. The program instructions may be embodied in software, firmware, and / or RAMware. Memory 120 may include any one or more of volatile, non-volatile, magnetic, optical, or electrical media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), electrically erasable programmable ROM (EEPROM), flash memory, or any other digital media.

[0027] User interface 130 may include display 132, input device 134, and speaker 136. In some examples, user interface 130 may include fewer or additional components. User interface 130 is configured to present information to a user (e.g., a clinician). For example, user interface 130 and / or display 132 may include a monitor, cathode ray tube display, flat panel display (such as a liquid crystal (LCD) display), plasma display, light-emitting diode (LED) display, and / or any other suitable display. In some examples, user interface 130 may be part of a multi-parameter monitor (MPM) or other physiological signal monitor used in a clinical or other setting, a personal digital assistant, a mobile phone, a tablet computer, a laptop computer, any other suitable computing device, or any combination thereof, with a built-in or stand-alone display.

[0028] In some examples, the processing circuitry 110 may be configured to present a graphical user interface to a user via a user interface 130 (such as a display 132). The graphical user interface may include indications of one or more physiological parameters of a patient via the display 132, such as blood pressure, oxygen saturation, information about autoregulation status (e.g., brain autoregulation values ​​and / or non-brain autoregulation values), pulse rate information, respiratory rate information, other patient physiological parameters, or combinations thereof. The user interface 130 may also include circuitry and other components, such as a speaker 136, configured to generate audio output and project that audio output to the user.

[0029] In some examples, the processing circuitry 110 may also receive input signals from other sources (not shown), such as a user. For example, the processing circuitry 110 may receive input signals from an input device 134, such as a keyboard, mouse, touchscreen, button, switch, microphone, joystick, touchpad, or any other suitable input device or combination of input devices. The input signals may contain information about the patient 101, such as physiological parameters, treatments provided to the patient 101, etc. The processing circuitry 110 may use additional input signals in any determination or operation performed by the processing circuitry 110.

[0030] In some examples, if the processing circuitry 110 determines that the patient 101's brain autoregulation is impaired, the processing circuitry 110 may present a notification indicating the impairment. This notification may include visual, auditory, tactile, or somatosensory notifications (e.g., alarm signals) indicating the patient 101's brain autoregulation state. In some examples, the processing circuitry 110 and the user interface 130 may be part of the same device or housed within a housing (e.g., a computer or monitor). In other examples, the processing circuitry 110 and the user interface 130 may be separate devices configured to communicate via a wired or wireless connection (e.g., a communication interface).

[0031] Oxygen saturation sensing circuitry 140 and blood pressure sensing circuitry (collectively referred to as sensing circuitry 140 and 142) may be configured to receive physiological signals sensed by respective sensing devices 150 and 152 and transmit these physiological signals to processing circuitry 110. Sensing devices 150 and 152 may include any sensing hardware configured to sense physiological parameters of a patient, such as, but not limited to, one or more electrodes, a photoreceiver, a blood pressure cuff, etc. The sensed physiological signals may include signals indicating physiological parameters of the patient 101, such as, but not limited to, blood pressure, regional oxygen saturation, blood volume, heart rate, and respiration. For example, sensing circuitry 140 and 142 may include, but are not limited to: blood pressure sensing circuitry, oxygen saturation sensing circuitry, regional oxygen saturation sensing circuitry, regional cerebral oxygen saturation sensing circuitry, blood volume sensing circuitry, heart rate sensing circuitry, temperature sensing circuitry, electrocardiogram (ECG) sensing circuitry, electroencephalogram (EEG) sensing circuitry, or any combination thereof.

[0032] In some examples, sensing circuit systems 140 and 142 and / or processing circuit system 110 may include signal processing circuit system 112, which is configured to perform any suitable analog conditioning on the sensed physiological signal. For example, sensing circuit systems 140 and 142 may pass an unaltered (e.g., raw) signal to processing circuit system 110. Processing circuit system 110 (e.g., signal processing circuit system 112) may be configured to modify the raw signal into a usable signal by, for example, filtering (e.g., low-pass, high-pass, band-pass, notch, or any other suitable filtering), amplification, performing operations on the received signal (e.g., taking the derivative, averaging), performing any other suitable signal conditioning (e.g., converting a current signal to a voltage signal), or any combination thereof. In some examples, the conditioned analog signal may be processed by an analog-to-digital converter of signal processing circuit system 112 to convert the conditioned analog signal into a digital signal. In some examples, signal processing circuit system 112 may operate on the analog or digital form of the signal to separate the different components of the signal. In some examples, signal processing circuitry 112 may perform any suitable digital conditioning on the converted digital signal, such as low-pass, high-pass, band-pass, notch filtering, averaging, or any other suitable filtering, amplification, performing operations on the signal, performing any other suitable digital conditioning, or any combination thereof. In some examples, signal processing circuitry 112 may reduce the number of samples in the digital detector signal. In some examples, signal processing circuitry 112 may remove dark contributions or environmental contributions from the received signal. Additionally or alternatively, sensing circuitry 140 and 142 may include signal processing circuitry 112 for modifying one or more original signals and transmitting one or more modified signals to processing circuitry 110.

[0033] In some examples, the oxygen saturation sensing device 150 is a regional oxygen saturation sensor configured to generate an oxygen saturation signal indicating blood oxygen saturation within a region of the venous, arterial, and / or capillary system of patient 101. For example, the oxygen saturation sensing device 150 may be configured to be placed on the skin of patient 101 (such as on patient 101's forehead) to determine the regional oxygen saturation of a specific tissue region (e.g., the frontal cortex of patient 101 or another brain location). The oxygen saturation sensing device 150 may include a transmitter 160 and a detector 162. The transmitter 160 may include at least two light-emitting diodes (LEDs), each configured to emit light of a different wavelength, such as red or near-infrared light. As used herein, the term "light" may refer to energy generated by a radiation source and may include any wavelength within one or more of the electromagnetic radiation spectrum of ultrasound, radio, microwave, millimeter waves, infrared, visible light, ultraviolet, gamma rays, or X-rays. In some examples, the light-driving circuitry (e.g., within sensing device 150, sensing circuitry 140, control circuitry 122, and / or processing circuitry 110) can provide a light-driving signal to drive emitter 160 and cause emitter 160 to emit light. In some examples, the LEDs of emitter 160 emit light in the range of approximately 600 nanometers (nm) to approximately 1000 nm. In a particular example, one LED of emitter 160 is configured to emit light at approximately 730 nm, and another LED of emitter 160 is configured to emit light at approximately 810 nm. Other wavelengths of light may be used in other examples.

[0034] Detector 162 may include a first detection element positioned relatively "close" (e.g., proximal) to emitter 160 and a second detection element positioned relatively "far" (e.g., distal) to emitter 160. In some examples, the first and second detection elements may be selected to be particularly sensitive to a selected target energy spectrum of light source 160. Multiple wavelengths of light intensity can be received at both the "close" and "far" detectors 162. For example, if two wavelengths are used, the two wavelengths can be compared at each location, and the resulting signals can be compared to derive an oxygen saturation value associated with additional tissue (such as brain tissue) through which the light received at the "far" detector passes as it travels through a region of the patient (e.g., the patient's skull).

[0035] In operation, light can reach detector 162 after passing through tissues of patient 101, including skin, bone, other superficial tissues (e.g., non-brain tissue and superficial brain tissue) and / or deep tissues (e.g., deep brain tissue). Detector 162 can convert the intensity of the received light into an electrical signal. The light intensity can be directly correlated with the light absorptivity and / or reflectivity in the tissue. Surface data from the skin and skull can be subtracted to generate a time-varying oxygen saturation signal of the target tissue. This technique can be referred to as near-infrared spectroscopy (NIRS), and the oxygen saturation signal can be called an NIRS signal.

[0036] Oxygen saturation sensing device 150 can provide an oxygen saturation signal (e.g., a regional oxygen saturation signal) to processing circuitry system 110 or any other suitable processing device to enable it to assess the autoregulation status of patient 101. Further details of examples of determining oxygen saturation based on optical signals can be found in commonly assigned U.S. Patent No. 9,861,317, entitled “Methods and Systems for Determining Regional Blood Oxygen Saturation,” published January 9, 2018.

[0037] In operation, the blood pressure sensing device 152 and the oxygen saturation sensing device 150 may be placed on the same or different parts of the patient 101's body. For example, the blood pressure sensing device 152 and the oxygen saturation sensing device 150 may be physically separate from each other and may be placed separately on the patient 101. As another example, the blood pressure sensing device 152 and the oxygen saturation sensing device 150 may, in some cases, be part of the same sensor or supported by a single sensor housing. For example, the blood pressure sensing device 152 and the oxygen saturation sensing device 150 may be part of an integrated blood oxygenation system configured to noninvasively measure blood pressure (e.g., based on time delay in a plethysmography (PPG) signal) and regional oxygen saturation. One or both of the blood pressure sensing device 152 or the oxygen saturation sensing device 150 may be further configured to measure other parameters, such as hemoglobin, respiratory rate, respiratory effort, heart rate, saturation pattern detection, response to stimuli (such as bispectral index (BIS) or electromyography (EMG) response to electrical stimulation), etc. Although Figure 1 The example brain autoregulation monitoring system 100 is shown, but Figure 1 The components shown are not intended to be limiting. Additional or alternative components and / or implementation methods may be used in other examples.

[0038] The blood pressure sensing device 152 can be any sensor or device configured to generate a blood pressure signal indicating blood pressure at a collection site of patient 101. For example, the blood pressure sensing device 152 may include a blood pressure cuff configured to noninvasively monitor blood pressure, a sensor configured to noninvasively generate a PPG signal, or an arterial catheter for invasively monitoring blood pressure in an artery of patient 101. In some examples, the blood pressure signal may include at least a portion of a waveform of the collected blood pressure. In some examples, the collection site may include at least one of the femoral artery, radial artery, dorsalis pedis artery, brachial artery, or combinations thereof of patient 101. In some examples, the blood pressure sensing device 152 may include multiple blood pressure sensing devices. For example, each of the multiple blood pressure sensing devices may be configured to acquire a corresponding blood pressure at a corresponding collection site among multiple collection sites of patient 101. The multiple collection sites may include similar or different arteries of patient 101.

[0039] In some examples, the blood pressure sensing device 152 may include one or more pulse oximeters. The acquired blood pressure can be determined by processing the time delay between two or more feature points within a single PPG signal obtained from a single pulse oximeter. Further examples of determining blood pressure based on a comparison of the time delay between certain components of a single PPG signal obtained from a single pulse oximeter are described in jointly assigned U.S. Patent Application Publication No. 2009 / 0326386, filed September 30, 2008, entitled “Systems and Methods for Non-Invasive Blood Pressure Monitoring.” In other cases, the blood pressure of patient 101 may be continuously and noninvasively monitored via multiple pulse oximeters placed at multiple locations on patient 101. As described in commonly assigned U.S. Patent No. 6,599,251, entitled “Continuous Non-invasive Blood Pressure Monitoring Method and Apparatus”, issued on July 29, 2003, multiple PPG signals can be obtained from multiple pulse oximeters, and the PPG signals can be compared with each other to estimate the blood pressure of patient 101.

[0040] Regardless of the form of the blood pressure sensing device 152, it can be configured to generate a blood pressure signal indicating the change in blood pressure (e.g., arterial blood pressure) of the patient 101 over time. In an example where the blood pressure sensing device 152 includes multiple blood pressure sensing devices, the blood pressure signal can include multiple blood pressure signals, each indicating the blood pressure at a corresponding sampling site of the patient 101. The blood pressure sensing device 152 can provide the blood pressure signal to the sensing circuitry system 142, the processing circuitry system 110, or any other suitable processing device to enable it to assess the autoregulatory state of the patient 101.

[0041] According to various aspects of this disclosure, the processing circuitry 110 can be configured to receive a blood pressure signal generated by the sensing circuitry 142 and the sensing device 152, indicating the patient 101's blood pressure over a certain time period, and an oxygen saturation signal generated by the sensing circuitry 140 and the sensing device 150, indicating the patient 101's regional oxygen saturation over that time period. The time period in which the processing circuitry 110 is located can be a previous 30 seconds, 60 seconds, 90 seconds, 120 seconds, or any other suitable time period.

[0042] A blood pressure signal indicating the blood pressure of patient 101 over a certain period of time can indicate the mean arterial pressure (MAP) of patient 101, which is the average (i.e., mean) blood pressure of patient 101 during a single cardiac cycle. Therefore, in some examples, the blood pressure signal can indicate the MAP of patient 101 for each cardiac cycle during that time period.

[0043] The regional oxygen saturation (rSO2) of patient 101 indicated by the oxygen saturation signal can be the regional oxygen saturation of patient 101's brain. In some examples, the oxygen saturation sensing device 150 may include multiple sensors placed on different parts of patient 101, such as sensors placed on or near the right side of patient 101's head, and sensors placed on or near the left side of patient 101's head. In this example, the oxygen saturation signal indicating the regional oxygen saturation of patient 101 may include a first oxygen saturation signal (from an oxygen saturation sensor placed on the right side of patient 101's head) indicating a first regional brain oxygen saturation of patient 101, and a second oxygen saturation signal (from an oxygen saturation sensor placed on the left side of patient 101's head) indicating a second regional brain oxygen saturation of patient 101.

[0044] The processing circuitry system 110 can be configured to determine physiological data associated with the patient's blood pressure and / or regional oxygen saturation during the time period, based at least in part on received signals indicative of the patient's blood pressure and / or regional oxygen saturation during the time period. For example, the processing circuitry system 110 can determine the patient's cerebral oxygenation index (COx) during the time period based at least in part on the linear correlation between the patient's blood pressure and regional oxygen saturation during the time period. For example, the processing circuitry system 110 can determine the cerebral oxygenation index based on the correlation between blood cerebral oxygen saturation (rSO2) and mean arterial pressure (MAP).

[0045] In some examples, the processing circuitry system 110 can determine the gradient of the patient 101's MAP during a given time period (also known as a window), which can be a change in the patient 101's MAP during that time period. The processing circuitry system 110 can also determine the gradient of the patient 101's regional cerebral oxygen saturation during that time period, which can be a change in the patient 101's regional cerebral oxygen saturation during that time period.

[0046] The processing circuitry system 110 can be configured to use a brain autoregulation model 124 to determine the brain autoregulation state of patient 101, at least in part, based on the patient 101's blood pressure and regional oxygen saturation during the time period. As further described in detail throughout this disclosure, the brain autoregulation model 124 may include a neural network algorithm trained via machine learning to determine the patient 101's brain autoregulation state. The processing circuitry system 110 can execute the brain autoregulation model 124 and can use the patient 101's MAP and regional oxygen saturation during the time period as inputs to the brain autoregulation model 124 to generate an output from the brain autoregulation model 124 indicative of the patient 101's brain autoregulation state.

[0047] In some examples, the processing circuitry 110 may also input additional information associated with the patient 101's blood pressure and / or regional oxygen saturation during the time period into the brain autoregulation model 124 to determine the patient 101's brain autoregulation state based on this additional information. For example, the processing circuitry 110 may input one or more of the following: the gradient of the patient 101's blood pressure (e.g., MAP) during the time period, the gradient of the patient 101's regional brain oxygen saturation during the time period, the patient 101's COx during the time period, or a bypass marker indicating that the patient 101 is undergoing cardiopulmonary bypass surgery during the time period.

[0048] In some examples, if the processing circuitry 110 receives an oxygen saturation signal in the form of a NIRS signal, it can input the raw NIRS signal indicating regional brain oxygen saturation of patient 101 into the brain autoregulation model 124. In some examples, if the processing circuitry 110 receives an oxygen saturation signal including a first oxygen saturation signal (from an oxygen saturation sensor placed on the right side of patient 101's head) indicating a first regional brain oxygen saturation of patient 101 and a second oxygen saturation signal (from an oxygen saturation sensor placed on the left side of patient 101's head) indicating a second regional brain oxygen saturation of patient 101, it can input both the first and second regional brain oxygen saturation signals into the brain autoregulation model 124. For example, the processing circuitry 110 can input two separate values ​​into the brain autoregulation model 124, or it can input the average of the two values ​​into the brain autoregulation model 124. In some examples, the processing circuitry 110 can also input patient 101's blood oxygen saturation (SpO2) into the brain autoregulation model 124.

[0049] In some examples, the processing circuitry 110 may also input additional information about patient 101 into the brain autoregulation model 124 and determine the patient 101's brain autoregulation state based on this additional information. For example, such additional information may include morphological features associated with patient 101's blood pressure during the time period and / or morphological features associated with patient 101's regional oxygen saturation during the time period. The additional information may also include patient 101's blood pressure other than MAP during the time period, such as patient 101's systolic or diastolic blood pressure during the time period. The additional information may also include patient demographic data about patient 101, such as the patient's age, information about patient 101's diet and lifestyle (e.g., whether patient 101 is a smoker), etc. The patient's brain autoregulation state may vary based on the variables indicated by the additional information.

[0050] Processing circuitry 110 can be configured to execute brain autoregulation model 124 to output an indication of the brain autoregulation state of patient 101 based on information input to brain autoregulation model 124. For example, brain autoregulation model 124 can output a value indicating whether the brain autoregulation state of patient 101 is one of the following: intact, damaged, or unknown. An intact brain autoregulation state indicates that the brain autoregulation control mechanism of patient 101 is functioning normally, while a damaged brain autoregulation state indicates that the brain autoregulation control mechanism of patient 101 is not functioning normally. Intact brain autoregulation occurs within a blood pressure range defined between LLA and ULA. Determining the brain autoregulation state using brain autoregulation model 124 allows processing circuitry 110 to quickly determine the brain autoregulation function of patient 101, for example, before determining the LLA and ULA specific to patient 101 or without needing to determine them.

[0051] In some examples, the brain autoregulation model 124 is a neural network algorithm trained via machine learning, which is used to determine the brain autoregulation state of patient 101 by taking multiple signals, including a blood pressure signal indicating the blood pressure of patient 101 and an oxygen saturation signal indicating the regional oxygen saturation of patient 101, as inputs.

[0052] Neural network algorithms, or artificial neural networks, can include trainable or adaptive algorithms that utilize nodes with defined rules. For example, a corresponding node among multiple nodes can generate an output based on an input using functions such as nonlinear functions or if-then rules. A corresponding node among multiple nodes can be connected along an edge to one or more different nodes among multiple nodes, such that the output of the corresponding node includes the input of the different nodes. These functions can include: parameters that can be determined or adjusted using a training set of inputs and desired outputs; predetermined associations between multiple signals or values ​​(such as blood pressure signals or(multiple) blood pressure values ​​from patient 101 or a patient group and oxygen saturation signals or(multiple) oxygen saturation values ​​of patient 101 or a patient group measured simultaneously with the blood pressure signals); and learning rules such as backpropagation learning rules. Backpropagation learning rules can train the neural network algorithm to minimize one or more error measurements by changing parameters, using one or more error measurements that compare the desired output with the output generated by the neural network algorithm.

[0053] The example neural network includes multiple nodes, at least some of which have node parameters. An input can be provided to the first node of the neural network algorithm that includes at least an oxygen saturation signal generated by oxygen saturation sensing device 150 or oxygen saturation sensing circuit system 140 and indicating regional brain oxygen saturation of patient 101 over a certain time period, and a blood pressure signal generated by blood pressure sensing device 152 or blood pressure sensing circuit system 142 and indicating blood pressure of patient 101 over that time period. In some examples, the input may include multiple inputs, each input to a corresponding node. The first node may include a function configured to determine the output based on the inputs and one or more adjustable node parameters. In some examples, the neural network may include a propagation function configured to determine the input of the next node based on the output and bias value of the previous node. In some examples, the learning rule may be configured to modify one or more node parameters to produce a favorable output. For example, the favorable output may be constrained by one or more thresholds and / or used to minimize one or more error measurements. The favorable output may include the output of a single node, a group of nodes, or multiple nodes.

[0054] The neural network algorithm can iteratively modify node parameters until the output includes a favorable output. In this way, the processing circuit system 110 can be configured to iteratively evaluate the output of the neural network algorithm and iteratively modify at least one node parameter based on the evaluation of the output of the neural network algorithm to determine the autoregulatory state of the brain of a patient (such as patient 101) based on the modified neural network algorithm.

[0055] In some examples, using brain autoregulation model 124, processing circuitry system 110 can determine the brain autoregulation state of patient 101 by determining a first value associated with a confidence level that the patient 101's brain autoregulation state is intact and a second value associated with a confidence level that the patient 101's brain autoregulation state is impaired. Therefore, processing circuitry system 110 can execute brain autoregulation model 124 to determine the brain autoregulation state of patient 101 based at least in part on the first value associated with a confidence level that the patient 101's brain autoregulation state is intact and the second value associated with a confidence level that the patient 101's brain autoregulation state is impaired.

[0056] For example, using the brain autoregulation model 124, the processing circuitry 110 can determine a value between 0 and 1 for each of the first and second values. If the brain autoregulation model 124 indicates that the first value is greater than the second value, then the brain autoregulation model 124 can output a value indicating that the patient 101's brain autoregulation state is intact. Conversely, if the brain autoregulation model 124 determines that the second value is greater than the first value, then the brain autoregulation model 124 can output a value indicating that the patient 101's brain autoregulation state is impaired. For example, if the brain autoregulation model 124 determines that the value associated with the confidence level that the patient 101's brain autoregulation state is intact is 0.8, and the value associated with the confidence level that the patient 101's brain autoregulation state is impaired, then the brain autoregulation model 124 can output a value indicating that the patient 101's brain autoregulation state is intact.

[0057] In some examples, if the brain autoregulation model 124 does not have sufficient confidence in classifying the patient 101's brain autoregulation state as intact or impaired, the brain autoregulation model 124 may output a value indicating that the patient 101's brain autoregulation state is unknown. In some examples, if the difference between the value associated with the confidence level of the patient 101's brain autoregulation state as intact and the value associated with the confidence level of the patient 101's brain autoregulation state as impaired is less than or equal to a threshold, the brain autoregulation model 124 may output a value indicating that the patient 101's brain autoregulation state is unknown. For example, if the value associated with the confidence level of the patient 101's brain autoregulation state as intact is 0.6, the value associated with the confidence level of the patient 101's brain autoregulation state as impaired is 0.4, and the threshold is 0.2, then the brain autoregulation model 124 may output a value indicating that the patient 101's brain autoregulation state is unknown.

[0058] In some examples, the brain autoregulation model 124 may perform post-processing on the values ​​determined by the brain autoregulation model 124 to determine the brain autoregulation state of patient 101. For example, the brain autoregulation model 124 may average the value associated with a confidence level that the patient 101's brain autoregulation state is intact with a previously determined value associated with a confidence level that the patient 101's brain autoregulation state is intact at the same blood pressure as the patient 101's current blood pressure. Similarly, the brain autoregulation model 124 may average the value associated with a confidence level that the patient 101's brain autoregulation state is impaired with a previously determined value associated with a confidence level that the patient 101's brain autoregulation state is impaired at the same blood pressure as the patient 101's current blood pressure. In other words, when the brain autoregulation model 124 determines values ​​associated with a confidence level that the patient 101's brain autoregulation state is intact and values ​​associated with a confidence level that the patient 101's brain autoregulation state is impaired, these values ​​may be associated with the patient 101's blood pressure at the time these values ​​are determined, and the processing circuitry system 110 may store the association between the patient 101's blood pressure and the confidence level values ​​in the memory 120.

[0059] In some examples, when the brain autoregulation model 124 determines values ​​associated with a confidence level that the patient 101's brain autoregulation state is intact and values ​​associated with a confidence level that the patient 101's brain autoregulation state is impaired at a given blood pressure, the brain autoregulation model 124 can look up confidence values ​​associated with the same blood pressure stored in memory 120. The brain autoregulation model 124 can average the values ​​associated with a confidence level that the patient 101's brain autoregulation state is intact with previously determined values ​​associated with a confidence level that the patient 101's brain autoregulation state is intact at the same blood pressure as the patient 101's current blood pressure, and can average the values ​​associated with a confidence level that the patient 101's brain autoregulation state is impaired with previously determined values ​​associated with a confidence level that the patient 101's brain autoregulation state is impaired at the same blood pressure as the patient 101's current blood pressure.

[0060] Therefore, the brain autoregulation model 124 can compare the average value associated with a confidence level that the patient 101's brain autoregulation state is intact with the average value associated with a confidence level that the patient 101's brain autoregulation state is impaired to determine the patient 101's brain autoregulation state. If the brain autoregulation model 124 determines that the average value associated with a confidence level that the patient 101's brain autoregulation state is intact is greater than or equal to the average value associated with a confidence level that the patient 101's brain autoregulation state is impaired, and if the difference between these two values ​​is greater than a threshold, then the brain autoregulation model 124 can determine that the patient 101's brain autoregulation state is intact. If the brain autoregulation model 124 determines that the average value associated with a confidence level that the patient 101's brain autoregulation state is impaired is greater than or equal to the average value associated with a confidence level that the patient 101's brain autoregulation state is intact, and if the difference between these two values ​​is greater than a threshold, then the brain autoregulation model 124 can determine that the patient 101's brain autoregulation state is impaired. If the difference between the two averages is less than or equal to the threshold, then the brain autoregulation model 124 can determine that the brain autoregulation state of patient 101 is unknown.

[0061] Once the processing circuitry system 110 has determined the autoregulation state of patient 101, it can generate information indicating the autoregulation state of patient 101 and output this information to an output device, such as user interface 130. The processing circuitry system 110 delivers this information to user interface 130. In some examples, this information may enable user interface 130 (e.g., display 132, speaker 136, and / or separate displays(not shown)) to present a graphical user interface including information indicating the autoregulation state of patient 101 (such as autoregulation state values) and / or indications of impaired autoregulation state of the brain. In some examples, the autoregulation state indication may include text, color, and / or audio presented to the user. The processing circuitry system 110 may be further configured to present indications of one or more autoregulation state values, one or more autoregulation limits (e.g., LLA and / or ULA), blood pressure(s), oxygen saturation(s), etc., on the graphical user interface. In addition to or as an alternative to a graphical user interface, the processor circuitry 110 may be configured to generate and present information indicating the determined autoregulatory state of the patient 101 via a speaker 136. For example, in response to detecting an impaired autoregulatory state in the patient 101, the processor circuitry 110 may generate an audible alarm via the speaker 136.

[0062] In some examples, the brain autoregulation monitoring system 100 (e.g., processing circuitry system 110 or user interface 130) may include a communication interface to enable the brain autoregulation monitoring system 100 to exchange information with external devices. The communication interface may include any suitable hardware, software, or both, which may allow the brain autoregulation monitoring system 100 to communicate with electronic circuitry systems, devices, networks, servers or other workstations, displays, or any combination thereof. For example, the processing circuitry system 110 may receive blood pressure values, oxygen saturation values, or predetermined data (such as predetermined brain autoregulation state values, predetermined non-brain autoregulation state values, or predetermined adjustment values) from an external device via the communication interface.

[0063] The components of the brain autoregulation monitoring system 100 are illustrated and described for illustrative purposes only, and these components are shown and described as individual components. In some examples, the functionality of some components may be combined in a single component. For example, the functionality of processing circuitry system 110 and control circuitry system 122 may be combined in a single processor system. Additionally, in some examples, the functionality of some components of the brain autoregulation monitoring system 100 shown and described herein may be divided among multiple components. For example, some or all of the functionality of control circuitry system 122 may be performed in processing circuitry system 110 or sensing circuitry systems 140 and 142. In other examples, the functionality of one or more components may be performed in a different order, or may not be necessary.

[0064] Figure 2 It demonstrates what can be done to Figure 1 Details of the example training system 200 for performing training on the brain autoregulation model 124 shown. Figure 2 Only one specific example of the training system 200 has been shown, and many other example devices with more, fewer, or different components can also be configured to operate according to the techniques disclosed herein.

[0065] Although the components of the training system 200 are in Figure 2 In some examples, these components are shown as part of a single device, but in others, they may be located within and / or part of different devices. For example, in some examples, training system 200 may represent a "cloud" computing system. Therefore, in these examples, Figure 2 The modules shown can span multiple computing devices. In some examples, training system 200 can represent one of multiple servers in a server cluster that makes up a "cloud" computing system. In other examples, training system 200 can be... Figure 1 An example of the brain autoregulation device 100 shown.

[0066] like Figure 2As illustrated in the example, training system 200 includes one or more processors 202, one or more communication units 204, and one or more storage devices 208. Storage device 208 further includes a brain autoregulation model 124, a training module 212, and training data 214. Each of components 202, 204, and 208 can be interconnected (physically, communicatively, and / or operatively) for inter-component communication. Figure 2 In the example, components 202, 204, and 208 can be coupled via one or more communication channels 206. In some examples, communication channel 206 may include a system bus, network connection, inter-process communication data structure, or any other channel for transmitting data. The brain autoregulation model 124, training module 212, and training data 214 can also transmit information to each other and to other components in the training system 200.

[0067] exist Figure 2 In the example, one or more processors 202 (each including a processing circuitry system) may implement functions and / or execute instructions within the training system 200. For example, one or more processors 202 may receive and execute instructions stored in storage device 208 that perform the functions of training module 212. During execution, these instructions executed by one or more processors 202 may cause the training system 200 to store information in storage device 208. One or more processors 202 may execute instructions of training module 212 to train brain autoregulation model 124 using training data 214. That is, training module 212 may be operated by one or more processors 202 to perform various actions or functions of the training system 200 described herein.

[0068] exist Figure 2 In the example, one or more communication units 204 may be operated to communicate with external devices (e.g., via one or more networks by sending and / or receiving network signals on those networks). Figure 1 The training system 200 can communicate with the device 100. For example, the training system 200 can use the communication unit 204 to transmit and / or receive radio signals on a radio network such as a cellular radio network. Similarly, the communication unit 204 can transmit and / or receive satellite signals on a satellite network such as a Global Positioning System (GPS) network. Examples of the communication unit 204 include network interface cards (such as Ethernet cards), optical transceivers, radio frequency transceivers, or any other type of device capable of transmitting and / or receiving information. Other examples of the communication unit 204 may include a near-field communication (NFC) unit, Radio, shortwave radio, cellular data radio, wireless network (e.g., Radio, and Universal Serial Bus (USB) controller.

[0069] exist Figure 2 In some examples, one or more storage devices 208 may be operable for storing information to be processed during the operation of the training system 200. In some examples, the storage device 208 may represent temporary memory, meaning that the primary purpose of the storage device 208 is not long-term storage. For example, the storage device 208 of the training system 200 may be volatile memory configured for short-term storage of information and therefore will not retain the stored contents in the event of a power outage. Examples of volatile memory include random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), and other forms of volatile memory known in the art.

[0070] In some examples, storage device 208 also represents one or more computer-readable storage media. That is, storage device 208 can be configured to store a larger amount of information than temporary storage. For example, storage device 46 may include non-volatile memory that retains information through power-on / off cycles. Examples of non-volatile memory include magnetic hard disks, optical disks, floppy disks, flash memory, or various forms of electrically programmable memory (EPROM) or electrically erasable programmable memory (EEPROM). In any case, in Figure 2 In the example, storage device 208 may store program instructions and / or data associated with brain autoregulation model 124, training module 212 and training data 214.

[0071] exist Figure 2 In the example, training system 200 may execute training module 212 to train brain autoregulation model 124 using training data 214, so as to more accurately and / or more quickly determine a patient's brain autoregulation state by associating one or more features with the patient's brain autoregulation state through training brain autoregulation model 124. Brain autoregulation model 124 may include a deep learning architecture such as a recurrent neural network or a convolutional neural network, which includes multiple layers to progressively extract higher-level features from the input of brain autoregulation model 124.

[0072] In some examples, training module 212 trains brain autoregulation model 124 to use the patient's blood pressure (such as the patient's MAP) and the patient's regional brain oxygen saturation as inputs, and to determine the patient's brain autoregulation state (e.g., whether the brain autoregulation state is intact, damaged, or unknown) based on the patient's MAP and regional brain oxygen saturation. Specifically, training module 212 trains brain autoregulation model 124 to use inputs such as the patient's MAP and regional brain oxygen saturation during a certain time period, and to determine the patient's brain autoregulation state immediately following that time period based on the inputs during that time period. Therefore, if the time period is 30 seconds, training module 212 can train brain autoregulation model 124 to determine the patient's brain autoregulation state immediately following that 30-second time period based on inputs such as the patient's MAP and regional brain oxygen saturation during the 30-second time period.

[0073] For example, training module 212 can train brain autoregulation model 124 by providing data such as MAP and regional cerebral oxygen saturation over time periods such as hours or days, and can train brain autoregulation model 124 by providing truth labels for labeling time points in the data to indicate whether the brain autoregulation state is intact or impaired. For example, if training module 212 is training brain autoregulation model 124 to determine the patient's brain autoregulation state immediately following a 30-second time period based on inputs during that time period, training module 212 can label data at 30 seconds to indicate the brain autoregulation state associated with data from time 0 to 30 seconds (e.g., MAP and regional cerebral oxygen saturation), label data at 31 seconds to indicate the brain autoregulation state associated with data from time 1 to 31 seconds, label data at 32 seconds to indicate the brain autoregulation state associated with data from time 2 to 32 seconds, and so on.

[0074] In some examples, training module 212 can train brain autoregulation model 124 by subtracting the mean of the MAP from the patient's MAP and by subtracting the mean regional brain oxygen saturation from the patient's regional oxygen saturation. For example, for a time period of 0 to 30 seconds, training module 212 can subtract the mean of the MAP of the entire training data from the MAP of the time period of 0 to 30 seconds, and can subtract the mean of the regional brain oxygen saturation of the entire training data from the regional oxygen saturation of the time period of 0 to 30 seconds, and can use the resulting values ​​as input to train brain autoregulation model 124.

[0075] Subtracting the mean MAP from the patient's MAP and subtracting the mean regional cerebral oxygen saturation from the patient's regional oxygen saturation can help prevent overfitting of the trained neural network in brain autoregulation model 124 based on the absolute value of MAP, such as preventing brain autoregulation model 124 from being trained to learn that a MAP of less than 50 is generally associated with impaired brain autoregulation. Alternatively, training module 212 can train brain autoregulation model 124 to determine the patient's brain autoregulation state based on the relationship between trends in MAP and trends in regional cerebral oxygen saturation during that time period.

[0076] In addition to MAP and regional brain oxygen saturation, training module 212 also trains brain autoregulation model 124 to use data from each time period as input, such as, but not limited to, one or more of the following: the gradient of MAP during that time period (e.g., as a time series), the gradient of regional brain oxygen saturation during that time period (e.g., as a time series), the brain oxygenation index during that time period, a marker indicating whether the patient is undergoing a medical procedure that may affect blood pressure values ​​(e.g., cardiopulmonary bypass surgery) during that time period, the raw NIRS signal indicating regional brain oxygen saturation during that time period, oxygen saturation signals from oxygen saturation sensors placed on the left and right sides of the patient's head (as two separate signals or as a combined (e.g., average) signal), morphological features associated with blood pressure during that time period and / or morphological features associated with regional oxygen saturation during that time period, hemodynamically relevant signals (such as systolic or diastolic blood pressure) during that time period, patient-associated demographic data, etc.

[0077] In some examples, training module 212 also uses the patient's blood oxygen saturation (SpO2) as input when training brain autoregulation model 124. Using the patient's blood oxygen saturation to train autoregulation model 124 allows brain autoregulation model 124 to respond to changes in regional brain oxygen saturation caused by changes in blood oxygen saturation unrelated to cerebral blood flow. For example, training module 212 can train brain autoregulation model 124 to learn that changes in regional brain oxygen saturation caused by changes in blood oxygen saturation unrelated to cerebral blood flow are not necessarily an indication of a change in the patient's autoregulatory state.

[0078] In some examples, the training data 214 used to train the brain autoregulation model 124 includes only data from patient 101 and excludes data from other subjects. In other examples, the training data 214 may include data from a patient population, which may or may not include patient 101. In some examples, once the training module 212 has trained the brain autoregulation model 124 using the training data 214, the training module 212 can test the brain autoregulation model 124 using a set of unexplored test data by determining how well the brain autoregulation model 124 classifies the brain autoregulation state based on the test data matches the expected brain autoregulation state classification of the test data. In this way, the training module 212 can evaluate and further improve the brain autoregulation model 124.

[0079] Once training module 212 has completed training the brain autoregulation model 124, the brain autoregulation model 124 can be installed, uploaded, or otherwise transferred to autoregulation monitoring system 100. In some examples, training module 212 may upload or otherwise transfer a copy of the brain autoregulation model 124 to another server or cloud, and autoregulation monitoring system 100 may connect the brain autoregulation model 124 via a network such as the Internet, VPN, or LAN.

[0080] In some examples, training module 212 may be able to train brain autoregulation model 124 to determine the autoregulatory state of organs other than the brain, such as the kidneys, gastrointestinal tract, etc. Specifically, in addition to regional brain oxygen saturation data, training module 212 may also use additional signals from these other organs as training data 214 to train brain autoregulation model 124 to determine the autoregulatory state of organs other than the brain. Physiological parameters from which the autoregulatory state of non-brain organs can be determined more directly may be relatively difficult to measure. Therefore, determining the autoregulatory state of non-brain organs based on brain autoregulatory state values ​​would be useful, for example, as described in U.S. Patent No. 10,932,673, published March 2, 2021, by Addison et al., entitled “Non-Cerebral Organ Autoregulatory State Determination,” which is incorporated herein by reference in its entirety.

[0081] Figure 3 Showing Figure 1 and Figure 2 Example brain autoregulation model 124, example deep learning architecture 300. Although deep learning architecture 300 is in Figure 3The LSTM deep learning architecture shown is used to train a Long Short-Term Memory (LSTM) model, but any other deep learning architecture can be equally well adapted to train a brain autoregulation model.

[0082] like Figure 3 As shown, the deep learning architecture 300 may include a sequence input layer 302, a bidirectional long short-term memory (BiLSTM) layer 304, a dropout layer 306, a BiLSTM layer 308, a dropout layer 310, a BiLSTM layer 312, a fully connected layer 314, a softmax layer 316, and a classification layer 318. The sequence input layer 302 can be connected to the BiLSTM layer 304. The BiLSTM layer 304 may have 16 hidden units and can be connected to the dropout layer 306. The dropout layer 306 may have a dropout ratio of 0.01 and can be connected to the BiLSTM layer 308. The BiLSTM layer 308 may have 8 hidden units and can be connected to the dropout layer 310. The dropout layer 310 may have a dropout ratio of 0.01 and can be connected to the BiLSTM layer 312. The BiLSTM layer 312 may have 4 hidden units and can be connected to the fully connected layer 314. The fully connected layer 314 can be connected to the softmax layer 316. The softmax layer 316 can be connected to the classification layer 318.

[0083] Sequence input layers (such as sequence input layer 302) feed sequence data into the neural network. Thus, sequence input layer 302 receives features for training the deep learning architecture 300, such as the MAP and regional brain oxygen saturation of one or more patients over time.

[0084] Dropout layers (such as dropout layers 306 and 310) randomly set input elements to zero with a given probability. By randomly setting input elements to zero, dropout layers allow elements to be ignored during the training phase. Selectively ignoring elements during the training phase can prevent overfitting of the training data.

[0085] BiLSTM layers (such as BiLSTM layers 304, 308, and 312) learn bidirectional long-term dependencies between time series or sequence data at time steps. These dependencies can be useful for the network to learn from the complete time series at each time step.

[0086] Fully connected layers (such as fully connected layer 314) multiply the input (e.g., from BiLSTM layer 312) by a weight matrix and then add a bias vector. Softmax layers (such as softmax layer 316) apply a softmax function to the input (e.g., from fully connected layer 314). The softmax function can be used as the final activation function in a neural network classifier (e.g., brain autoregulation model 124) to normalize the output of fully connected layer 314 to the probability of predicting the output class.

[0087] The classification layer (such as classification layer 318) assigns the input to one of two or more mutually exclusive classes using the probability of the predicted output class output by the softmax layer 316 for the input of the deep learning architecture 300, and can output the output class of the input to the brain autoregulation model 124 as a result of training the brain autoregulation model 124 with the deep learning architecture 300.

[0088] Figure 4 Showing Figure 1 and Figure 2 Example brain autoregulation model 124, example deep learning architecture 400. Although deep learning architecture 400 is in Figure 4 The example shown is a CNN used to train a convolutional neural network (CNN) model, but any other deep learning architecture can be equally well-suited for training the brain's autoregulation model 124.

[0089] like Figure 4 As shown, a portion of the deep learning architecture 400 includes a convolutional layer 404, a max pooling layer 406, an additive layer 408, a batch normalization layer 410, a rectified linear unit (ReLU) layer 412, a dropout layer 414, a convolutional layer 416, a batch normalization layer 418, a ReLU layer 420, a dropout layer 422, a convolutional layer 424, a max pooling layer 426, and an additive layer 428. Figure 4 The portion of the deep learning architecture 400 shown may only be a part (not all) of the hidden layers of the CNN (i.e., deep learning architecture 400), and the deep learning architecture may include... Figure 4 Other additional layers not shown.

[0090] Two-dimensional convolutional layers (such as convolutional layers 404, 416, and 424) apply a sliding convolutional filter to the input of the layer. The layer convolves the input by moving the filter vertically and horizontally along the input and calculating the dot product of the weights and the input, and then adding a bias term.

[0091] Max-pooling layers (such as max-pooling layers 406 and 426) perform downsampling by dividing the input into rectangular pooling regions and computing the maximum value in each region. Max-pooling layers follow convolutional layers for downsampling, thereby reducing the number of connections to the layers following the max-pooling layer and reducing the number of parameters to be learned in the layers following the max-pooling layer. Max-pooling layers can also reduce overfitting in neural network models.

[0092] Additive layers (such as adder layers 408 and 428) add inputs from multiple neural network layers element-wise. In the example of deep learning architecture 400, adder layer 408 can add inputs from convolutional layer 404 and max pooling layer 406, and adder layer 428 can add inputs from convolutional layer 424 and max pooling layer 426.

[0093] Batch normalization layers (such as batch normalization layers 410 and 418) normalize each input channel on a mini-batch basis. Batch normalization layers can speed up CNN training and reduce sensitivity to network initialization. Batch normalization layers can be used between convolutional layers and ReLU layers; therefore, in deep learning architecture 400, batch normalization layer 410 is used between convolutional layer 404 and ReLU layer 412, and batch normalization layer 418 is used between convolutional layer 416 and ReLU layer 420.

[0094] ReLU layers (such as ReLU layers 412 and 420) perform a threshold operation on each element of the input, where any value less than zero is set to zero. ReLU layers can allow for faster and more efficient training of deep learning architectures on large and complex datasets.

[0095] Dropout layers (such as dropout layers 414 and 422) randomly set input elements to zero with a given probability. By randomly setting input elements to zero, dropout layers allow elements to be ignored during the training phase. Selectively ignoring elements during the training phase can prevent overfitting of the training data.

[0096] Figure 5 Demonstrated the use of Figure 3 and Figure 4 Example graphs showing the results of classifying the brain's autoregulation states using a deep learning architecture. (Example:...) Figure 5 As shown in Figure 500A, the use of... Figure 3 The results of classifying brain autoregulation states using the LSTM deep learning architecture 300 are shown in Figure 500B. Figure 4 The results of classifying the brain's autoregulation state using the CNN deep learning architecture 400.

[0097] Each of Graphs 500A and 500B displays a Receiver Operating Characteristic (ROC) curve, a graphical representation of the diagnostic ability of a binary classifier system when discriminating changes in its discrimination threshold. The y-axis of each of Graphs 500A and 500B represents the true positive rate (TPR) for classifying brain autoregulation states using the brain autoregulation model 124, and the y-axis of each of Graphs 500A and 500B represents the false positive rate (FPR) for classifying brain autoregulation states using the brain autoregulation model 124.

[0098] The receiver operating characteristic (ROC) curve in Figure 500A and Figure 502A demonstrate the use of LSTM (such as...) Figure 3 The brain autoregulation model 124, trained using an LSTM deep learning architecture 300, classifies a single subject (e.g., patient 101) in the test set. The area under the ROC curve (A502A, also known as AUROC) can be equal to the probability that the classifier ranks a randomly selected positive instance higher than a randomly selected negative instance, which is 0.93. The accuracy of the brain autoregulation model 124 trained using LSTM can be 0.72, the sensitivity can be 0.95, and the specificity can be 0.71.

[0099] The Receiver Operating Characteristic (ROC) curve in graph 500B and graph 502B demonstrate the use of CNNs (such as...) Figure 4 The brain autoregulation model 124, trained using a CNN deep learning architecture 400, classifies a single subject (e.g., patient 101) in a test set. The area under the ROC curve (AUC) is equal to the probability that the classifier ranks a randomly selected positive instance higher than a randomly selected negative instance, which is 0.97. The accuracy of the brain autoregulation model 124 trained using CNN is 0.93, the sensitivity is 0.86, and the specificity is 0.94.

[0100] Figure 6 An example user interface incorporating information about brain autoregulation is shown. Figure 6 As shown, the graphical user interface (GUI) 600 is an example of an interface that the processing circuitry system 110 of the brain autoregulation monitoring system 100 can output, which is displayed on the display 132 to provide information on the brain autoregulation state of the patient 101. The graphical user interface 600 includes a graph of the patient 101's blood pressure value 602 changing over time.

[0101] The safe zone 604 in GUI 600 can display areas above LLA, while the unsafe zone 606 in GUI 600 can display areas below LLA. By showing whether patient 101's blood pressure value 602 is within the safe zone 604 or the unsafe zone 606, GUI 600 demonstrates the relationship between patient 101's blood pressure value 602 and patient 101's autoregulatory state. As shown by GUI 600, blood pressure value 602 is in the safe zone 604 until it drops below LLA at time t1, reaching the unsafe zone 606, and then until it recovers above LLA at time t2, thus returning to the safe zone 604.

[0102] Figure 7 This is a flowchart illustrating an example method for monitoring a patient's brain autoregulation state. Although regarding the processing circuitry system 110 of the brain autoregulation monitoring system 100 ( Figure 1 )right Figure 7 As described, in other examples, different processing circuitry systems may be performed individually or in combination with processing circuitry system 110. Figure 7 Any part of the technology.

[0103] like Figure 7 As shown, the method includes a processing circuit system 110 collecting patient data (702) about patient 101 over a certain time period. For example, patient data can be collected over time periods such as 30 seconds, 60 seconds, 90 seconds, and 120 seconds. For example, if the time period is 30 seconds, the processing circuit system 110 can collect patient data from time t to time t+30 seconds.

[0104] To collect patient data, the processing circuitry system 110 can receive a blood pressure signal indicating the patient 101's blood pressure (such as the patient 101's MAP) and an oxygen saturation signal indicating the patient 101's regional oxygen saturation (such as the patient 101's regional cerebral oxygen saturation). For example, sensing devices 150 and 152 can generate blood pressure and oxygen saturation signals, both of which are received by the processing circuitry system 110 as described above.

[0105] In some examples, the processing circuitry system 110 performs cleaning (704) on the collected patient data. For example, the processing circuitry system 110 may remove invalid values ​​(such as any data portion that is outside the valid value range), perform smoothing on the data values, perform interpolation on the values, etc.

[0106] The processing circuitry system 110 provides the collected patient data (e.g., cleaned patient data) to the brain autoregulation model 124 (706). In some examples, to provide the collected patient data to the brain autoregulation model 124, the processing circuitry system 110 may provide the patient 101's blood pressure (e.g., MAP) during the specified time period and the patient 101's regional brain oxygen saturation during at least the same time period as input to the brain autoregulation model 124. In some examples, if the regional brain oxygen saturation signal received by the processing circuitry system 110 comprises two or more regional brain oxygen saturation values, the processing circuitry system 110 may provide each of the two or more regional brain oxygen saturation values ​​as input to the brain autoregulation model 124, or may provide the average of the two or more regional brain oxygen saturation values ​​as input to the brain autoregulation model 124.

[0107] In some examples, the processing circuitry 110 can derive one or more values ​​from the patient 101's blood pressure and regional cerebral oxygen saturation during the time period, and can provide these values ​​as input to the brain autoregulation model 124. For example, the processing circuitry 110 can determine the gradient of the patient 101's MAP during the time period, and can determine the gradient of the patient 101's regional cerebral oxygen saturation during the time period, and can provide the determined gradients as input to the brain autoregulation model 124.

[0108] In some examples, the processing circuitry system 110 may provide additional data associated with patient 101 as input to the brain autoregulation model 124. For example, the processing circuitry system 110 may provide the following as input to the brain autoregulation model 124: a bypass marker indicating whether patient 101 is undergoing cardiopulmonary bypass surgery or other medical procedures during the time period; morphological characteristics of patient 101's blood pressure and / or regional brain oxygen saturation during the time period (e.g., raw blood pressure signal, raw PPG signal, etc., including the peak, location, area, etc. of the raw blood pressure signal and raw PPG signal); patient 101's systolic blood pressure and / or diastolic blood pressure during the time period; and patient 101's demographic data, such as height, weight, age, sex, disease status, and body mass index.

[0109] The processing circuitry 110 executes the brain autoregulation model 124 to output an indication (708) of the brain autoregulation state of the patient 101 based on the input data. That is, the processing circuitry 110 can use the brain autoregulation model 124 to classify the brain autoregulation state of the patient 101 into one of the following states: intact, damaged, or unknown.

[0110] In some examples, the brain autoregulation model 124 can determine confidence values ​​associated with the patient 101 being in an intact state and confidence values ​​associated with the patient 101 being in a damaged state based on input data, and can determine the patient 101's brain autoregulation state based on the determined confidence values. For example, if the brain autoregulation model 124 determines that the confidence value associated with the patient 101 being in an intact state is higher than the confidence value associated with the patient 101 being in a damaged state, then the processing circuitry system 110 uses the model 124 to determine that the patient 101's brain autoregulation state is intact. Conversely, if the brain autoregulation model 124 determines that the confidence value associated with the patient 101 being in a damaged state is higher than the confidence value associated with the patient 101 being in an intact state, then the processing circuitry system 110 uses the model 124 to determine that the patient 101's brain autoregulation state is damaged.

[0111] In some examples, the brain autoregulation model 124 determines a confidence difference threshold. If the brain autoregulation model 124 determines that the difference between the confidence value associated with the patient 101 being in a damaged state and the confidence value associated with the patient 101 being in an intact state is less than or equal to the confidence difference threshold, then the processing circuitry system 110 can use the brain autoregulation model 124 to determine that the patient 101's brain autoregulation state is unknown.

[0112] In some examples, the brain autoregulation model 124 can average the currently determined confidence value associated with the patient 101 being in a damaged state with a previous confidence value associated with the patient 101 being in a damaged state at the same blood pressure as the patient 101's current blood pressure to determine an average confidence value associated with the patient 101 being in a damaged state. Similarly, the brain autoregulation model 124 can average the currently determined confidence value associated with the patient 101 being in an intact state with a previous confidence value associated with the patient 101 being in an intact state at the same blood pressure as the patient 101's current blood pressure to determine an average confidence value associated with the patient 101 being in an intact state. The brain autoregulation model 124 can determine the patient 101's brain autoregulation state by comparing the average confidence value associated with the patient 101 being in an intact state and the average confidence value associated with the patient 101 being in a damaged state, such as by using the techniques described above.

[0113] In some examples, the brain autoregulation model 124 may, in response to determining confidence values ​​associated with the patient 101 being in an intact state and in a damaged state, record these confidence values ​​in a confidence matrix stored in memory 120. The values ​​in the confidence matrix can be used posteriorly to calculate the brain autoregulation state of the patient 101.

[0114] In response to determining the autoregulation state of patient 101, the processing circuitry system 110 may also return to collecting patient data about patient 101 over a certain period of time (702). In some examples, the processing circuitry system 110 may wait for one second to pass before returning to collecting patient data about patient 101 over a certain period of time.

[0115] The method further includes, in response to determining the autoregulation state of the patient 101, the processing circuitry system 110 outputting an indication of the autoregulation state of the patient 101 (such as for display on a display 132) (712). For example, the processing circuitry system 110 may output a graphical user interface (such as... Figure 6 The graphical user interface 600 can provide the currently determined brain autoregulation state of patient 101 and additional information associated with patient 101, such as patient 101's blood pressure, patient 101's regional brain oxygen saturation, etc.

[0116] Figure 8 This is a flowchart illustrating an example method for monitoring a patient's brain autoregulation state. Although regarding the processing circuitry system 110 of the brain autoregulation monitoring system 100 ( Figure 1 )right Figure 8 As described, in other examples, different processing circuitry systems may be performed individually or in combination with processing circuitry system 110. Figure 8 Any part of the technology.

[0117] like Figure 8 As shown, the processing circuitry 110 of the brain autoregulation monitoring system 100 can receive a blood pressure signal indicating the patient 101's blood pressure over a certain time period and an oxygen saturation signal indicating the patient 101's regional brain oxygen saturation over that time period (802). The processing circuitry 110 can use the neural network algorithm of the brain autoregulation model 124 to determine the patient 101's brain autoregulation state, at least in part, based on the patient 101's blood pressure and regional brain oxygen saturation over that time period (804). The processing circuitry 110 can send a signal indicating the patient 101's brain autoregulation state to an output device (806).

[0118] In some examples, the neural network algorithm of the brain autoregulation model 124 is used to determine the brain autoregulation state of patient 101 based at least in part on one or more of the following: the gradient of patient 101's blood pressure during the time period, the gradient of regional brain oxygen saturation of patient 101 during the time period, and the brain oxygenation index of patient 101 during the time period.

[0119] In some examples, the processing circuitry 110 is further configured to use a neural network algorithm of the brain autoregulation model 124 to determine the brain autoregulation state of patient 101 based at least in part on a bypass marker indicating that patient 101 is undergoing cardiopulmonary bypass surgery during that time period.

[0120] In some examples, the neural network algorithm is trained via machine learning on training data 214 to classify the autoregulatory state of patient 101’s brain into one of the following states: damaged, intact, or unknown.

[0121] In some examples, training data 214 includes two or more of the following: blood pressure of one or more patients over time, regional cerebral oxygen saturation values ​​of one or more patients over time, gradients of blood pressure of one or more patients in each time period of multiple time periods, gradients of regional cerebral oxygen saturation of one or more patients in each time period of multiple time periods, cerebral oxygenation index (COx) of blood pressure and regional cerebral oxygen saturation of one or more patients in each time period of multiple time periods, one or more bypass markers indicating whether one or more patients are undergoing cardiopulmonary bypass surgery during each time period of multiple time periods, morphological features of one or more of the following: blood pressure or regional cerebral oxygen saturation during each time period of multiple time periods, systolic blood pressure of one or more patients over time, diastolic blood pressure of one or more patients over time, or demographic data associated with one or more patients.

[0122] In some examples, the blood pressure of one or more patients over time includes, for each of these time periods, the blood pressure during the corresponding time period minus the mean blood pressure over time, and the regional cerebral oxygen saturation of one or more patients over time includes, for each of these time periods, the regional cerebral oxygen saturation during the corresponding time period minus the mean regional cerebral oxygen saturation over time.

[0123] In some examples, the processing circuitry 110 may be further configured to use a neural network algorithm of the brain autoregulation model 124 to determine the brain autoregulation state of patient 101 by at least the following operations: determining a first confidence score associated with the brain autoregulation state of patient 101 being intact, determining a second confidence score associated with the brain autoregulation state of patient 101 being damaged, and classifying the brain autoregulation state of patient 101 into one of the following states: damaged, intact, or unknown, based at least in part on comparing the first confidence score and the second confidence score.

[0124] In some examples, in order to classify the autoregulation state of patient 101, the processing circuitry system 101 is further configured to determine whether the difference between a first confidence score and a second confidence score is within a confidence threshold, and in response to determining that the difference between the first confidence score and the second confidence score is less than or equal to the confidence threshold, classify the patient's autoregulation state as unknown.

[0125] In some examples, the processing circuitry 101 is further configured to: determine an average first confidence score associated with the patient 101’s brain autoregulation state being intact as the average of a first confidence score associated with the patient 101’s brain autoregulation state being intact at the patient’s current blood pressure and a first set of previously determined confidence scores at the current blood pressure; determine an average second confidence score associated with the patient 101’s brain autoregulation state being impaired as the average of a second confidence score associated with the patient 101’s brain autoregulation state being impaired at the patient’s current blood pressure and a second set of previously determined confidence scores at the current blood pressure; and classify the patient 101’s brain autoregulation state by comparing the average first confidence score and the average second confidence score to classify the patient 101’s brain autoregulation state as one of the following states: impaired, intact, or unknown.

[0126] The techniques described in this disclosure include those belonging to device 100, processing circuitry system 110, control circuitry system 122, sensing circuitry systems 140, 142, or various components, and can be implemented at least in part in hardware, software, firmware, or any combination thereof. For example, different aspects of these techniques can be implemented in one or more processors, including one or more microprocessors, DSPs, ASICs, FPGAs, or any other equivalent integrated or discrete logic circuits, and any combination of such components, embodied in programmers such as clinician or patient programmers, medical devices, or other devices. For example, the processing circuitry system, control circuitry system, and sensing circuitry system, as well as other processors and controllers described herein, can be implemented at least in part as or include one or more executable applications, application modules, libraries, classes, methods, objects, routines, subroutines, firmware, and / or embedded code, etc.

[0127] In one or more examples, the functions described in this disclosure may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, these functions may be stored as one or more instructions or code on a computer-readable medium and executed by a hardware-based processing unit. The computer-readable medium may be an article of manufacture including a non-transitory computer-readable storage medium encoded with instructions. Instructions embedded in or encoded in an article of manufacture including an encoded non-transitory computer-readable storage medium, such as instructions included in or encoded in a non-transitory computer-readable storage medium, when executed by one or more processors, may enable one or more programmable processors or other processors to implement one or more of the techniques described herein. Examples of non-transitory computer-readable storage media may include RAM, ROM, programmable ROM (PROM), erasable programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, hard disk, optical disk ROM (CD-ROM), floppy disk, magnetic tape cassette, magnetic media, optical media, or any other computer-readable storage device or tangible computer-readable medium.

[0128] In some examples, computer-readable storage media include non-transitory media. The term "non-transitory" can indicate that the storage medium is not implemented with a carrier wave or a propagated signal. In some examples, non-transitory storage media can store data that may change over time (e.g., in RAM or cache).

[0129] The functionalities described herein can be provided within dedicated hardware and / or software modules. Describing different features as modules or units is intended to highlight functional differences and does not imply that such modules or units must be implemented by different hardware or software components. Rather, the functionality associated with one or more modules or units can be performed by different hardware or software components, or integrated within common or different hardware or software components. Furthermore, these techniques can be fully implemented within one or more circuit or logic elements.

[0130] The following terms include example topics described in this article.

[0131] Clause 1: A method comprising: receiving, by a processing circuit system, a blood pressure signal indicating a patient's blood pressure over a certain time period and an oxygen saturation signal indicating the patient's regional brain oxygen saturation over the time period; determining, by the processing circuit system, a brain autoregulation state of the patient based at least in part on the patient's blood pressure over the time period and the patient's regional brain oxygen saturation over the time period using a neural network algorithm of a brain autoregulation model; and sending, by the processing circuit system, a signal indicating the patient's brain autoregulation state to an output device.

[0132] Clause 2: The method as described in Clause 1, wherein determining the patient's brain autoregulation state using the neural network algorithm of the brain autoregulation model further comprises: determining the brain autoregulation state by the processing circuit system based at least in part on one or more of the following: the gradient of the patient's blood pressure during the time period; the gradient of the patient's regional brain oxygen saturation during the time period; or the patient's brain oxygenation index during the time period.

[0133] Clause 3: The method as described in Clause 1 or 2, wherein determining the patient's brain autoregulation state using a neural network algorithm of the brain autoregulation model further comprises: determining the brain autoregulation state by the processing circuitry system based at least in part on a bypass marker indicating that the patient is undergoing cardiopulmonary bypass surgery during the time period.

[0134] Clause 4: The method of any one of Clauses 1 to 3, wherein the neural network algorithm is trained via machine learning on training data to classify the patient’s brain autoregulation state into one of the following states: damaged, intact, or unknown.

[0135] Clause 5: The method as described in Clause 4, wherein the training data includes two or more of the following: blood pressure of one or more patients over time; regional cerebral oxygen saturation values ​​of the one or more patients over time; gradients of blood pressure of the one or more patients in each of a plurality of time periods; gradients of regional cerebral oxygen saturation values ​​of the one or more patients in each of the plurality of time periods; a brain oxygenation index (COx) determined based on blood pressure and regional cerebral oxygen saturation of the one or more patients in each of the plurality of time periods; one or more bypass markers indicating whether the one or more patients are undergoing cardiopulmonary bypass surgery during each of the time periods; morphological characteristics of one or more of blood pressure or regional cerebral oxygen saturation during each of the time periods; systolic blood pressure of the one or more patients over time; diastolic blood pressure of the one or more patients over time; or demographic data associated with the one or more patients.

[0136] Clause 6: The method as described in Clause 5, wherein: the blood pressure of the one or more patients over time includes, for each time period, the blood pressure during the corresponding time period minus the mean of the blood pressure over time; and the regional cerebral oxygen saturation value of the one or more patients over time includes, for each time period, the regional cerebral oxygen saturation value during the corresponding time period minus the mean of the regional cerebral oxygen saturation value over time.

[0137] Clause 7: The method of any one of Clauses 4 to 6, wherein determining the patient's brain autoregulation state using a neural network algorithm of the brain autoregulation model further comprises: determining a first confidence score associated with the patient's brain autoregulation state being intact by the processing circuitry system; determining a second confidence score associated with the patient's brain autoregulation state being impaired by the processing circuitry system; and classifying the patient's brain autoregulation state into one of the following states: impaired, intact, or unknown by the processing circuitry system based at least in part on a comparison of the first confidence score and the second confidence score.

[0138] Clause 8: The method as described in Clause 7, wherein classifying the patient's brain autoregulation state further comprises: determining by the processing circuitry whether the difference between the first confidence score and the second confidence score is within a confidence threshold; and, in response to determining that the difference between the first confidence score and the second confidence score is less than or equal to the confidence threshold, classifying the patient's brain autoregulation state as unknown by the processing circuitry.

[0139] Clause 9: The method as described in Clause 7 or 8 further comprises: determining by the processing circuitry a mean first confidence score associated with the patient's intact brain autoregulation state as the average of a first confidence score associated with the patient's intact brain autoregulation state at the patient's current blood pressure and a first set of previously determined confidence scores at the current blood pressure; determining a mean second confidence score associated with the patient's impaired brain autoregulation state as the average of a second confidence score associated with the patient's impaired brain autoregulation state at the patient's current blood pressure and a second set of previously determined confidence scores at the current blood pressure; and wherein classifying the patient's brain autoregulation state further comprises: comparing the mean first confidence score and the mean second confidence score by the processing circuitry to classify the patient's brain autoregulation state as one of the following states: impaired, intact, or unknown.

[0140] Clause 10: A system comprising: a blood pressure sensing device; an oxygen saturation sensing device; and a processing circuit system configured to: receive from the blood pressure sensing device a blood pressure signal indicating a patient's blood pressure over a certain time period, and from the oxygen saturation sensing device an oxygen saturation signal indicating the patient's regional brain oxygen saturation over the time period; determine the patient's brain autoregulation state based at least in part on the patient's blood pressure and the patient's regional brain oxygen saturation over the time period using a neural network algorithm of a brain autoregulation model; and send a signal indicating the patient's brain autoregulation state to an output device.

[0141] Clause 11: The system as described in Clause 10, wherein the determination of the patient's brain autoregulation state using the neural network algorithm of the brain autoregulation model is further based at least in part on one or more of the following: the gradient of the patient's blood pressure during the time period; the gradient of the patient's regional brain oxygen saturation during the time period; or the patient's brain oxygenation index during the time period.

[0142] Clause 12: A system as described in Clause 10 or 11, wherein the processing circuitry is further configured to use a neural network algorithm of the brain autoregulation model to determine the patient’s brain autoregulation state based at least in part on a bypass marker indicating that the patient is undergoing cardiopulmonary bypass surgery during the said time period.

[0143] Clause 13: The system as described in any one of Clauses 10 to 12, wherein the neural network algorithm is trained via machine learning on training data to classify the patient’s brain autoregulation state into one of the following states: damaged, intact, or unknown.

[0144] Clause 14: A system as described in Clause 13, wherein the training data includes two or more of the following: blood pressure of one or more patients over time; regional cerebral oxygen saturation values ​​of the one or more patients over time; gradients of blood pressure of the one or more patients in each of a plurality of time periods; gradients of regional cerebral oxygen saturation of the one or more patients in each of the plurality of time periods; cerebral oxygenation index (COx) of blood pressure and regional cerebral oxygen saturation of the one or more patients in each of the time periods; one or more bypass markers indicating whether the one or more patients are undergoing cardiopulmonary bypass surgery during each of the time periods; morphological features of one or more of the following: blood pressure or regional cerebral oxygen saturation during each of the time periods; systolic blood pressure of the one or more patients over time; diastolic blood pressure of the one or more patients over time; or demographic data associated with the one or more patients.

[0145] Clause 15: The system as described in Clause 14, wherein: the blood pressure of the one or more patients over time includes, for each time period, the blood pressure during the corresponding time period minus the mean of the blood pressure over time; and the regional cerebral oxygen saturation of the one or more patients over time includes, for each time period, the regional cerebral oxygen saturation during the corresponding time period minus the mean of the regional cerebral oxygen saturation over time.

[0146] Clause 16: The system of any one of Clauses 13 to 15, wherein the processing circuitry is configured to use a neural network algorithm of the brain autoregulation model to determine the patient’s brain autoregulation state by at least the following operations: determining a first confidence score associated with the patient’s brain autoregulation state being intact; determining a second confidence score associated with the patient’s brain autoregulation state being impaired; and classifying the patient’s brain autoregulation state into one of the following states: impaired, intact, or unknown, at least in part based on a comparison of the first confidence score with the second confidence score.

[0147] Clause 17: The system as described in Clause 16, wherein, in order to classify the patient's brain autoregulation state, the processing circuitry is further configured to: determine whether the difference between the first confidence score and the second confidence score is within a confidence threshold; and, in response to determining that the difference between the first confidence score and the second confidence score is less than or equal to the confidence threshold, classify the patient's brain autoregulation state as unknown.

[0148] Clause 18: A system as described in Clause 16 or 17, wherein the processing circuitry is further configured to: determine an average first confidence score associated with the patient's intact brain autoregulation state as the average of a first confidence score associated with the patient's intact brain autoregulation state at the patient's current blood pressure and a first set of previously determined confidence scores at the current blood pressure; determine an average second confidence score associated with the patient's impaired brain autoregulation state as the average of a second confidence score associated with the patient's impaired brain autoregulation state at the patient's current blood pressure and a second set of previously determined confidence scores at the current blood pressure; and classify the patient's brain autoregulation state by comparing the average first confidence score with the average second confidence score to classify the patient's brain autoregulation state as one of the following states: impaired, intact, or unknown.

[0149] Clause 19: A non-transitory computer-readable and storable medium comprising: receiving a blood pressure signal indicating a patient's blood pressure over a time period and an oxygen saturation signal indicating a patient's regional brain oxygen saturation over the time period; determining a patient's brain autoregulation state based at least in part on the patient's blood pressure over the time period and the patient's regional brain oxygen saturation over the time period using a neural network algorithm of a brain autoregulation model; and transmitting a signal indicating the patient's brain autoregulation state to an output device.

[0150] Clause 20: A non-transitory computer-readable and storable medium as described in Clause 19, wherein the instruction causing the processing circuitry to determine the patient's brain autoregulation state using a neural network algorithm of the brain autoregulation model further causes the processing circuitry to: determine a first confidence score associated with the patient's brain autoregulation state being intact; determine a second confidence score associated with the patient's brain autoregulation state being impaired; and classify the patient's brain autoregulation state into one of the following states: impaired, intact, or unknown, at least in part based on a comparison of the first confidence score with the second confidence score.

[0151] Various examples of this disclosure have been described. Any combination of the described systems, operations, or functions is contemplated. These and other examples fall within the scope of the following claims.

Claims

1. A system comprising: Blood pressure sensing device; Oxygen saturation sensing device; as well as Processing circuitry system, the processing circuitry system being configured to: The blood pressure signal, which indicates the patient's blood pressure over a certain period of time, is received from the blood pressure sensing device, and the oxygen saturation signal, which indicates the patient's regional brain oxygen saturation over the same period of time, is received from the oxygen saturation sensing device. A neural network algorithm using a brain autoregulation model determines the patient's brain autoregulation state based at least in part on the patient's blood pressure and regional brain oxygen saturation during the time period, wherein the neural network algorithm is trained via machine learning on training data to classify the patient's brain autoregulation state into one of the following states: damaged, intact, or unknown; and The output device sends a signal indicating the patient's brain autoregulation state.

2. The system as claimed in claim 1, wherein, The neural network algorithm of the brain autoregulation model is used to determine the patient's brain autoregulation state, which is further based at least in part on one or more of the following: The gradient of the patient's blood pressure during the time period; The gradient of regional cerebral oxygen saturation in the patient during the stated time period; or The patient's brain oxygenation index during the stated time period.

3. The system as described in any one of claims 1 or 2, wherein, The processing circuitry is further configured to use a neural network algorithm of the brain autoregulation model to determine the patient’s brain autoregulation state, at least in part, based on bypass markers indicating that the patient was undergoing cardiopulmonary bypass surgery during the time period.

4. The system as described in any one of claims 1 to 3, wherein, The training data includes two or more of the following: Blood pressure changes over time in one or more patients; The regional brain oxygen saturation values ​​of one or more patients over time; The gradient of blood pressure for one or more patients in each of multiple time periods; The gradient of regional cerebral oxygen saturation in the one or more patients during each of the plurality of time periods; Blood pressure and regional cerebral oxygen saturation (COx) of the one or more patients during each time period of the time period; One or more bypass markers indicating whether the one or more patients are undergoing cardiopulmonary bypass surgery during each time period of the time period; One or more of the following morphological characteristics: blood pressure or brain oxygen saturation in the said region during each time period of the said time period; The systolic blood pressure of the one or more patients over time; The diastolic blood pressure of the one or more patients over time; or Demographic data associated with the one or more patients.

5. The system of claim 4, wherein: The blood pressure of the one or more patients over time includes, for each time period, the blood pressure during the corresponding time period minus the mean of the blood pressure over time. and The regional cerebral oxygen saturation of the one or more patients over time includes, for each time period, the regional cerebral oxygen saturation during the corresponding time period minus the mean of the regional cerebral oxygen saturation over time.

6. The system as described in any one of claims 1 to 5, wherein, The processing circuitry is configured to use a neural network algorithm of the brain autoregulation model to determine the patient's brain autoregulation state by at least the following operations: Determine the first confidence score associated with the patient's intact brain autoregulation state; Determine the second confidence score associated with impaired brain autoregulation in the patient; and The patient's brain autoregulation state is classified into one of the following states: damaged, intact, or unknown, based at least in part on comparing the first confidence score with the second confidence score.

7. The system of claim 6, wherein, In order to classify the patient's brain autoregulation state, the processing circuitry is further configured as follows: Determine whether the difference between the first confidence score and the second confidence score is within the confidence threshold; and In response to determining that the difference between the first confidence score and the second confidence score is less than or equal to a confidence threshold, the patient's brain autoregulation state is classified as unknown.

8. The system as claimed in any one of claims 6 or 7, wherein, The processing circuit system is further configured as follows: The mean first confidence score associated with the patient’s intact brain autoregulation state is determined as the average of the first confidence score associated with the patient’s intact brain autoregulation state at the patient’s current blood pressure and the first set of previously determined confidence scores at the current blood pressure. The average second confidence score associated with the patient's impaired brain autoregulation was determined as the average of the second confidence score associated with the patient's impaired brain autoregulation at the patient's current blood pressure and the second set of previously determined confidence scores at the current blood pressure; and The patient's brain autoregulation state is classified by comparing the average first confidence score with the average second confidence score, and the patient's brain autoregulation state is classified into one of the following states: damaged, intact, or unknown.

9. A non-transitory computer-readable and storable medium, the medium comprising instructions that, when executed, cause a processing circuit system to: Receive a blood pressure signal indicating the patient's blood pressure over a certain period of time and an oxygen saturation signal indicating the patient's regional brain oxygen saturation over the same period of time; A neural network algorithm using a brain autoregulation model determines the patient's brain autoregulation state based at least in part on the patient's blood pressure and regional brain oxygen saturation during the time period, wherein... The neural network algorithm is trained via machine learning on training data to classify the patient's brain autoregulation state into one of the following states: damaged, intact, or unknown; and The output device sends a signal indicating the patient's brain autoregulation state.

Images