Autoregulation monitoring using deep learning
In some examples, a system is configured to determine, using a neural network algorithm of a cerebral autoregulation model, a cerebral autoregulation status of the patient based at least in part on a blood pressure of the patient over a period of time and regional cerebral oxygen saturation of the patient over the period of time.
1 . A system, comprising:
an output device to display an autoregulation status of a patient;
processing circuitry communicatively coupled to the output device; and
memory storing instructions that, when executed by the processing circuitry, cause the processing circuitry to perform operations, comprising:
receive a regional oxygen saturation signal indicative of a regional oxygen saturation of the patient during a period of time;
receive a blood pressure signal indicative of a blood pressure of the patient during the period of time;
determine, via an autoregulation model, the autoregulation status of the patient based on the regional oxygen saturation signal and the blood pressure signal, wherein the autoregulation model is trained via machine learning over training data to classify the autoregulation status of the patient as one of impaired, intact, or unknown based on comparing two or more confidence scores associated with the autoregulation status of the patient; and
instruct the output device to present a graphical user interface (GUI) to display the autoregulation status of the patient.
2 . The system of claim 1 , wherein the autoregulation model is trained via machine learning over the training data to classify the autoregulation status of the patient based on comparing a first confidence score of the two or more confidence scores associated with the autoregulation status of the patient being intact with a second confidence score of the two or more confidence scores associated with the autoregulation status of the patient being impaired.
3 . The system of claim 1 , wherein the instructions, when executed by the processing circuitry, cause the processing circuitry to provide an age of the patient, a diet of the patient, a lifestyle of the patient, or any combination thereof to the autoregulation model to determine the autoregulation status.
4 . The system of claim 1 , wherein the instructions, when executed by the processing circuitry, cause the processing circuitry to perform further operations, comprising:
determine a respective gradient of the regional oxygen saturation of the patient over the period of time;
determine a respective gradient of the blood pressure of the patient over the period of time; and
provide the respective gradient of the regional oxygen saturation and the respective gradient of the blood pressure to the autoregulation model to determine the autoregulation status.
5 . The system of claim 1 , wherein the instructions, when executed by the processing circuitry, cause the processing circuitry to instruct the output device to update the GUI to display at least one of the blood pressure of the patient, the regional oxygen saturation of the patient, a pulse rate of the patient, or a respiration rate of the patient.
6 . The system of claim 1 , wherein the instructions, when executed by the processing circuitry, cause the processing circuitry to perform further operations, comprising:
instruct the output device to present the GUI to display a first zone representing an intact autoregulation zone for the patient, a second zone representing an impaired autoregulation zone for the patient, and the blood pressure overlaid onto the first zone, the second zone, or both.
7 . The system of claim 6 , wherein the instructions, when executed by the processing circuitry, cause the processing circuitry to perform further operations, comprising:
determine a threshold autoregulation value;
determine a first region above the threshold autoregulation value being the first zone; and
determine a second region below the threshold autoregulation value being the second zone.
8 . The system of claim 1 , wherein the period of time is 30 seconds, 60 seconds, 90 seconds, or 120 seconds.
9 . A method comprising:
receiving, via processing circuitry, a blood pressure signal indicative of a blood pressure of a patient over a period of time;
receiving, via the processing circuitry, an oxygen saturation signal indicative of a regional oxygen saturation of the patient over the period of time;
inputting, via the processing circuitry, the blood pressure of the patient over the period of time and the regional oxygen saturation of the patient over the period of time into an autoregulation model to determine an autoregulation status of the patient, wherein the autoregulation model is trained via machine learning over training data to classify the autoregulation status of the patient as one of impaired, intact, or unknown based on comparing two or more confidence scores associated with the autoregulation status of the patient; and
instructing, via the processing circuitry, an output device to display the autoregulation status of the patient.
10 . The method of claim 9 , further comprising:
inputting, via the processing circuitry, the blood pressure of the patient over the period of time and the regional oxygen saturation of the patient over the period of time into the autoregulation model to:
determine a first confidence score of the two or more confidence scores, wherein the first confidence score is associated with the autoregulation status of the patient being intact;
determine a second confidence score of the two or more confidence scores, wherein the second confidence score is associated with the autoregulation status of the patient being impaired; and
classify the autoregulation status of the patient based on comparing the first confidence score and the second confidence score to determine the autoregulation status of the patient.
11 . The method of claim 9 , wherein instructing, via the processing circuitry, the output device to display the autoregulation status comprises:
instructing the output device to display a graph comprising a first region associated with an intact autoregulation zone for the patient, a second region associated with an impaired autoregulation zone for the patient, and the blood pressure overlaid onto at least one of the intact autoregulation zone or the impaired autoregulation zone.
12 . The method of claim 9 , comprising inputting, via the processing circuitry, a bypass flag into the autoregulation model to determine the autoregulation status of the patient during a cardiopulmonary bypass procedure.
13 . The method of claim 9 , comprising:
providing, via the processing circuitry, the training data comprising one or more of the blood pressure signal, the oxygen saturation signal, or the autoregulation status to the autoregulation model.
14 . The method of claim 9 , comprising:
providing, via the processing circuitry, the training data comprising one or more of a plurality of mean arterial pressures over additional periods of time, a plurality of regional oxygen saturation values over the additional periods of time, and a respective plurality of truth labels that correspond to the plurality of mean arterial pressures and the plurality of regional oxygen saturation values to train the autoregulation model, wherein each truth label of the plurality of truth labels indicates the autoregulation status as either an intact autoregulation status or an impaired autoregulation status.
15 . The method of claim 9 , comprising:
in response to receiving at least one of the blood pressure signal or the oxygen saturation signal, cleaning, via the processing circuitry, the at least one of the blood pressure signal, the oxygen saturation signal, or both.