Method and system for multidimensional monitoring of recovery state of anesthetic surgery

Abstract

The application provides a method and system for monitoring the recovery state of an anesthetic operation in multiple dimensions, the method comprising: acquiring time series data of vital signs of a patient within a preset period, and acquiring demographic characteristics of the patient; constructing a feature vector sequence corresponding to the preset period based on the time series data of vital signs and the demographic characteristics of the patient; constructing a time series prediction model; inputting the feature vector sequence into the time series prediction model to output predicted vital signs and a predicted probability of recovery of consciousness of the patient within a future prediction window, calculating a recovery index according to the predicted vital signs and the predicted probability of recovery of consciousness, and outputting early warning information of different levels based on the relationship between the recovery index and a preset threshold. The application constructs a feature vector by extracting characteristic information of electrocardiograms, blood pressure and blood oxygen, predicts vital signs and the probability of recovery of consciousness by using an LSTM model, calculates a recovery index and gives a graded early warning. The application realizes prospective evaluation of the recovery state during the anesthetic recovery period.

Description

Technical Field

[0001] This invention relates to the field of nursing technology, and more specifically, to a method and system for multidimensional monitoring of the recovery status during anesthesia and surgery. Background Technology

[0002] The post-anesthesia care unit (PACU) is a critical location for patients recovering after general anesthesia, and its assessment of recovery status directly impacts patient safety and turnover efficiency. Currently, clinical practice primarily relies on nurses periodically observing and manually recording patients' vital signs (heart rate, blood pressure, blood oxygen saturation) and level of consciousness (e.g., RASS score), combined with experience to judge the recovery progress. However, this assessment method has significant shortcomings: firstly, manual assessment is subjective and intermittent, making it difficult to capture subtle changes during the recovery process in a timely manner; secondly, existing monitoring methods mostly rely on single-indicator threshold alarms (e.g., heart rate below or above a fixed value), lacking the ability to comprehensively analyze the dynamic evolution of multiple parameters, and even more so, failing to predict the patient's recovery trend over a future period. Therefore, there is an urgent need for an intelligent method that can integrate multi-source vital sign data, automatically extract recovery characteristics, and prospectively assess recovery risks. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for multi-dimensional monitoring of the recovery status during anesthesia surgery, so as to improve the above-mentioned problems.

[0004] To achieve the above objectives, this application provides the following technical solution: On the one hand, embodiments of this application provide a method for multi-dimensional monitoring of the recovery status during anesthesia surgery, the method comprising: Acquire the patient's vital signs time-series data and demographic characteristics within a preset time period. The vital signs time-series data includes electrocardiogram time-series data, blood pressure time-series data, and blood oxygen saturation time-series data. The patient's demographic characteristics include age, sex, and body mass index. The cutoff time for the preset time period is the current time. Based on vital sign time-series data and patient demographic characteristics, feature vector sequences corresponding to preset time periods are constructed; a time-series prediction model is then built. The feature vector sequence is input into the time-series prediction model, which outputs the patient's predicted vital signs and the probability of regaining consciousness within the future prediction window. The resuscitation index is calculated based on the predicted vital signs and the probability of regaining consciousness, and different levels of early warning information are output based on the relationship between the resuscitation index and a preset threshold.

[0005] Secondly, this application provides a system for multi-dimensional monitoring of the recovery status during anesthesia surgery, the system comprising: The acquisition module is used to acquire the patient's vital signs time series data and the patient's demographic characteristics within a preset time period. The vital signs time series data includes electrocardiogram time series data, blood pressure time series data, and blood oxygen saturation time series data; the patient's demographic characteristics include age, gender, and body mass index. The cutoff time of the preset time period is the current time. The module is used to construct feature vector sequences corresponding to preset time periods based on vital sign time-series data and patient demographic characteristics; and to construct time-series prediction models. The prediction module is used to input the feature vector sequence into the time-series prediction model, output the patient's predicted vital signs and the probability of regaining consciousness within the future prediction window, calculate the resuscitation index based on the predicted vital signs and the probability of regaining consciousness, and output different levels of early warning information based on the relationship between the resuscitation index and a preset threshold.

[0006] Thirdly, this application provides a device for multi-dimensional monitoring of the recovery status during anesthesia surgery, the device including a memory and a processor. The memory is used to store a computer program; the processor is used to execute the computer program to implement the steps of the above-described method for multi-dimensional monitoring of the recovery status during anesthesia surgery.

[0007] Fourthly, this application provides a readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for multi-dimensional monitoring of the recovery status of anesthesia surgery.

[0008] The beneficial effects of this invention are as follows: 1. This invention extracts multidimensional dynamic features by performing change point detection and segmentation on three time-series data: electrocardiogram, blood pressure, and blood oxygen saturation. This allows for a comprehensive depiction of the physiological evolution of patients during anesthesia recovery, overcoming the one-sidedness of single-indicator assessment.

[0009] 2. This invention constructs a time-series prediction model based on long short-term memory networks. By training with historical patient data, an intelligent model is obtained that can predict future trends in vital signs and the probability of consciousness recovery. Based on this, a comprehensive recovery index is calculated, enabling medical staff to take intervention measures in advance before the patient's condition deteriorates.

[0010] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic diagram of the method for multi-dimensional monitoring of the recovery status during anesthesia surgery as described in this embodiment of the invention; Figure 2 This is a schematic diagram of the system structure for multi-dimensional monitoring of the recovery status during anesthesia surgery as described in this embodiment of the invention; Figure 3 This is a schematic diagram of the device structure for multi-dimensional monitoring of the recovery status during anesthesia surgery as described in this embodiment of the invention. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0014] It should be noted that similar reference numerals or letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0015] Example 1 like Figure 1 As shown in the figure, this embodiment provides a method for multi-dimensional monitoring of the recovery status of anesthesia surgery, which includes steps S1, S2 and S3.

[0016] Step S1: Obtain the patient's vital signs time series data and the patient's demographic characteristics within a preset time period. The vital signs time series data includes electrocardiogram time series data, blood pressure time series data, and blood oxygen saturation time series data. The patient's demographic characteristics include age, gender, and body mass index. The cutoff time of the preset time period is the current time. In this step, vital signs time-series data are acquired as soon as the patient is admitted to the PACU post-anesthesia recovery room; Step S2: Based on the time series data of vital signs and the demographic characteristics of patients, construct the feature vector sequence corresponding to the preset time period; construct the time series prediction model; In this step, based on the time series data of vital signs and the demographic characteristics of the patient, a feature vector sequence corresponding to a preset time period is constructed, including step S21; Step S21: The ECG time-series data, blood pressure time-series data, and blood oxygen saturation time-series data are continuously segmented according to a preset fixed time length to obtain a series of M windows that are connected end to end and do not overlap. Each window contains the corresponding ECG sub-time-series data, blood pressure sub-time-series data, and blood oxygen saturation sub-time-series data. The feature information of the ECG sub-time-series data, blood pressure sub-time-series data, and blood oxygen saturation sub-time-series data under each window is calculated. The feature information corresponding to the three is concatenated with the patient's demographic characteristics to obtain the feature vector under each window. The feature vectors of all windows are stacked in chronological order to form the feature vector sequence corresponding to the preset time period.

[0017] The time-series data of electrocardiogram (ECG), blood pressure, and blood oxygen saturation are continuously segmented according to a preset fixed time length, resulting in a series of consecutive, non-overlapping windows. Each window contains the corresponding ECG sub-time-series data, blood pressure sub-time-series data, and blood oxygen saturation sub-time-series data. This can be understood as follows: Assuming the original electrocardiogram, blood pressure, and blood oxygen saturation signals are all recorded starting from 0 seconds, with a preset fixed recording time of 30 seconds, then: The first window: from 0 seconds to 30 seconds, containing the ECG sub-time series data, blood pressure sub-time series data, and blood oxygen saturation sub-time series data within this time period.

[0018] The second window: from 30 seconds to 60 seconds, contains three sub-time series data within this time period.

[0019] The third window: from 60 seconds to 90 seconds, and so on.

[0020] In this step, the feature information of the electrocardiogram sub-time series data, blood pressure sub-time series data and blood oxygen saturation sub-time series data under each window is calculated, and the feature information corresponding to the three is concatenated with the patient's demographic features to obtain the feature vector under each window. The specific implementation steps include steps S211-S214. Step S211: Detect change points in the ECG sub-time series data, identify change points, and divide the ECG sub-time series data into multiple segments based on the change points; perform R-wave detection on each segment to obtain the RR interval sequence of successive heartbeats, calculate the square of the difference between adjacent RR intervals, sum all the squares to obtain the sum of squares; divide the sum of squares by the total number of heartbeats in the segment minus 1 to obtain the first value, take the square root of the first value to obtain the feature index of each segment, calculate the mean and standard deviation of all feature indices and use them as the first feature information of the ECG sub-time series data; Step S212: Perform change point detection on the blood pressure sub-time series data, identify change points, and divide the blood pressure sub-time series data into multiple segments based on the change points; calculate the mean and standard deviation of all systolic blood pressure values ​​in each segment, divide the standard deviation by the mean to obtain the systolic blood pressure variation coefficient, and use the systolic blood pressure variation coefficient of each segment as the feature index of each segment; calculate the mean and standard deviation of all feature indices and use them as the second feature information of the blood pressure sub-time series data; Step S213: Detect change points in the blood oxygen saturation sub-time series data, identify change points, and divide the blood oxygen saturation sub-time series data into multiple segments based on the change points. Subtract the minimum blood oxygen saturation from the maximum blood oxygen saturation in each segment to obtain the feature index of each segment. Calculate the mean and standard deviation of all feature indexes and use them as the third feature information of the blood oxygen saturation sub-time series data. Step S214: Concatenate the first feature information, the second feature information, and the third feature information with age, gender, and body mass index to obtain the feature vector.

[0021] In step S2, the specific implementation steps for constructing the time series prediction model include steps S22 and S23; Step S22: Obtain complete ECG time-series data, complete blood pressure time-series data, and complete blood oxygen saturation time-series data from admission to PACU to discharge of multiple historical patients, as well as obtain the demographic characteristics of historical patients; continuously segment the complete ECG time-series data, complete blood pressure time-series data, and complete blood oxygen saturation time-series data according to a preset fixed time length to obtain a series of multiple windows that are connected end to end and do not overlap, each window containing the corresponding historical ECG sub-time-series data, historical blood pressure sub-time-series data, and historical blood oxygen saturation sub-time-series data; calculate the feature information of the historical ECG sub-time-series data, historical blood pressure sub-time-series data, and historical blood oxygen saturation sub-time-series data under each window, and concatenate the corresponding feature information of the three with the demographic characteristics of historical patients to obtain the historical feature vector under each window; Step S23: Stack the historical feature vectors of M consecutive windows in chronological order to form a sample sequence; construct the label corresponding to each sample sequence, wherein the end time of the time interval covered by each sample sequence is denoted as the first moment, the prediction window is [first moment, first moment + preset time length], calculate the arithmetic mean of the heart rate sampling points in the prediction window as the true heart rate, calculate the arithmetic mean of the systolic blood pressure sampling points as the true systolic blood pressure, calculate the arithmetic mean of the blood oxygen saturation sampling points as the true blood oxygen saturation; obtain the true label of consciousness recovery at the end time of the prediction window; use the true heart rate, true systolic blood pressure, true blood oxygen saturation, and true label of consciousness recovery as the label of each sample sequence; construct a long short-term memory network model, train the long short-term memory network model based on multiple sample sequences and their labels, and obtain the time-series prediction model.

[0022] In this step, the true label of consciousness recovery at the end of the prediction window can be understood as: obtaining the actual RASS score (human score) at the end of the prediction window. If the RASS score is greater than or equal to -1, the true label of consciousness recovery is 1, indicating that consciousness has been recovered; otherwise, the true label of consciousness recovery is 0, indicating that consciousness has not been recovered.

[0023] In this step, a long short-term memory network model is constructed. The long short-term memory network model is trained based on multiple sample sequences and their labels to obtain the time series prediction model. The specific implementation steps include step S231. Step S231: Construct a Long Short-Term Memory (LSTM) network model, which includes an input layer, an LSTM hidden layer, a Dropout layer, and a fully connected output layer. Using the feature vector sequence as input and the corresponding label as output, the LSM network model is trained using a loss function. The loss function is a weighted sum of the mean squared error of heart rate prediction, the mean squared error of systolic blood pressure prediction, the mean squared error of blood oxygen saturation prediction, and the cross-entropy loss of consciousness recovery prediction. After training, a time-series prediction model is obtained.

[0024] In this step, a Long Short-Term Memory (LSTM) network is used. The training process for this network is as follows: First, an LSTM network model is constructed, with the following structure: an input layer, an LSTM hidden layer, a Dropout layer, and a fully connected output layer. The input layer receives the feature vector sequence; the LSTM hidden layer contains 64 memory units to capture long-range dependencies in the time series; the Dropout layer has a dropout rate of 0.3 to prevent overfitting; and the fully connected output layer contains four nodes, outputting predicted heart rate, predicted systolic blood pressure, predicted blood oxygen saturation (these three nodes use linear activation functions), and predicted probability of regaining consciousness (this node uses a sigmoid activation function, with an output value between 0 and 1).

[0025] Secondly, training samples are constructed. Each training sample includes a sequence of feature vectors and its label. During training, the sequence of feature vectors is used as input and the corresponding label is used as output. Then, a loss function is defined, which is a weighted sum of the loss functions for each prediction task. Specifically: Heart rate prediction loss is calculated using mean squared error, which is the square of the difference between the predicted heart rate and the actual heart rate; the same applies to systolic blood pressure prediction loss and blood oxygen saturation prediction loss.

[0026] The loss prediction for consciousness recovery uses cross-entropy. When the true label is 1, the loss value is calculated as follows: calculate the natural logarithm of the predicted probability and then take the opposite of the logarithm. When the true label is 0, the loss value is calculated as follows: first subtract the predicted probability from 1 to get the difference, then calculate the natural logarithm of the difference, and finally take the opposite of the logarithm. Then, the four losses are summed with weights of 0.3:0.3:0.2:0.2 to obtain the loss function; Finally, the Adam optimizer was used to train the model with a learning rate of 0.001 and a batch size of 64. During training, the training samples were divided into training, validation, and test sets. After each training epoch, the total loss on the validation set was calculated. When the loss on the validation set no longer decreased for 10 consecutive epochs, training was stopped, and the model parameters with the minimum validation loss were saved. This model is the trained time series prediction model.

[0027] Step S3: Input the feature vector sequence into the time series prediction model, output the patient's predicted vital signs and predicted probability of consciousness recovery within the future prediction window, calculate the resuscitation index based on the predicted vital signs and predicted probability of consciousness recovery, and output different levels of early warning information based on the relationship between the resuscitation index and the preset threshold.

[0028] In this step, after the feature vector sequence is input into the time series prediction model, the time series prediction model outputs the predicted heart rate, predicted systolic blood pressure, predicted blood oxygen saturation, and predicted probability of recovery of consciousness for the time period [current time, current time + preset time length]. Recovery Index = 100 × (1 - predicted probability of consciousness recovery) × (|predicted heart rate - baseline heart rate| / baseline heart rate + |predicted systolic blood pressure - baseline systolic blood pressure| / baseline systolic blood pressure + |baseline oxygen saturation - predicted oxygen saturation| / baseline oxygen saturation) / 3; The baseline heart rate, baseline systolic blood pressure, and baseline oxygen saturation can be individually set based on factors such as the patient's age and preoperative underlying diseases. For example, for adult patients without underlying diseases, the baseline heart rate can be set to 70 beats per minute, the baseline systolic blood pressure to 120 mmHg, and the baseline oxygen saturation to 98%.

[0029] Based on the relationship between the resuscitation index and a preset threshold, different levels of early warning information are output. The following can be used (the first and second thresholds can be customized as needed): If the resuscitation index is less than the first threshold, a green warning is output, indicating that the patient's resuscitation process is expected to be normal. If the first threshold is less than or equal to the recovery index and less than the second threshold, a yellow warning will be issued, indicating that the patient's recovery process is expected to be delayed and that enhanced monitoring is recommended. If the recovery index is greater than or equal to the second threshold, a red alert will be issued, indicating that the patient's recovery may be interrupted or worsen, and it is recommended to prepare intervention measures in advance.

[0030] Example 2 like Figure 2 As shown in the figure, this embodiment provides a system for multi-dimensional monitoring of the recovery status of anesthesia surgery. The system includes an acquisition module 1, a construction module 2, and a prediction module 3.

[0031] The acquisition module 1 is used to acquire the patient's vital signs time series data and the patient's demographic characteristics within a preset time period. The vital signs time series data includes electrocardiogram time series data, blood pressure time series data, and blood oxygen saturation time series data; the patient's demographic characteristics include age, gender, and body mass index. The cutoff time of the preset time period is the current time. Module 2 is used to construct feature vector sequences corresponding to preset time periods based on vital sign time series data and patient demographic characteristics; and to construct time series prediction models. Prediction module 3 is used to input the feature vector sequence into the time series prediction model, output the patient's predicted vital signs and the probability of regaining consciousness within the future prediction window, calculate the resuscitation index based on the predicted vital signs and the probability of regaining consciousness, and output different levels of early warning information based on the relationship between the resuscitation index and the preset threshold.

[0032] In one specific embodiment of this disclosure, the construction module 2 further includes a segmentation unit 21.

[0033] Segmentation unit 21 is used to continuously segment the electrocardiogram time series data, blood pressure time series data, and blood oxygen saturation time series data according to a preset fixed time length, resulting in a series of M windows that are connected end to end and do not overlap. Each window contains the corresponding electrocardiogram sub-time series data, blood pressure sub-time series data, and blood oxygen saturation sub-time series data. The feature information of the electrocardiogram sub-time series data, blood pressure sub-time series data, and blood oxygen saturation sub-time series data under each window is calculated. The feature information corresponding to the three is concatenated with the patient's demographic characteristics to obtain the feature vector under each window. The feature vectors of all windows are stacked in chronological order to form the feature vector sequence corresponding to the preset time period.

[0034] In one specific embodiment of this disclosure, the segmentation unit 21 further includes a first calculation unit 211, a second calculation unit 212, a third calculation unit 213, and a splicing unit 214.

[0035] The first calculation unit 211 is used to perform change point detection on the electrocardiogram sub-time series data, identify change points, and divide the electrocardiogram sub-time series data into multiple segments based on the change points; to perform R wave detection on each segment to obtain the RR interval sequence of successive heartbeats, calculate the square of the difference between adjacent RR intervals, sum all the squares to obtain the sum of squares; divide the sum of squares by the total number of heartbeats in the segment minus 1 to obtain the first value, take the square root of the first value to obtain the feature index of each segment, calculate the mean and standard deviation of all feature indices and use them as the first feature information of the electrocardiogram sub-time series data; The second calculation unit 212 is used to perform change point detection on the blood pressure sub-time series data, identify change points, and divide the blood pressure sub-time series data into multiple segments based on the change points; calculate the mean and standard deviation of all systolic blood pressure values ​​in each segment, divide the standard deviation by the mean to obtain the systolic blood pressure variation coefficient, and use the systolic blood pressure variation coefficient of each segment as the feature index of each segment; calculate the mean and standard deviation of all feature indices and use them as the second feature information of the blood pressure sub-time series data; The third calculation unit 213 is used to perform change point detection on the blood oxygen saturation sub-time series data, identify the change points, and divide the blood oxygen saturation sub-time series data into multiple segments based on the change points. The maximum blood oxygen saturation in each segment is subtracted from the minimum blood oxygen saturation to obtain the feature index of each segment. The mean and standard deviation of all feature indexes are calculated and used as the third feature information of the blood oxygen saturation sub-time series data. The splicing unit 214 is used to splice the first feature information, the second feature information, the third feature information with age, gender, and body mass index to obtain a feature vector.

[0036] In one specific embodiment of this disclosure, the construction module 2 further includes an acquisition unit 22 and a construction unit 23.

[0037] The acquisition unit 22 is used to acquire complete electrocardiogram time-series data, complete blood pressure time-series data, and complete blood oxygen saturation time-series data of multiple historical patients from admission to discharge from PACU, as well as to acquire the demographic characteristics of historical patients; the complete electrocardiogram time-series data, complete blood pressure time-series data, and complete blood oxygen saturation time-series data are continuously segmented according to a preset fixed time length to obtain a series of multiple windows that are connected end to end and do not overlap. Each window contains the corresponding historical electrocardiogram sub-time-series data, historical blood pressure sub-time-series data, and historical blood oxygen saturation sub-time-series data; the feature information of the historical electrocardiogram sub-time-series data, historical blood pressure sub-time-series data, and historical blood oxygen saturation sub-time-series data under each window is calculated, and the corresponding feature information of the three is concatenated with the demographic characteristics of historical patients to obtain the historical feature vector under each window; Unit 23 is used to stack the historical feature vectors of M consecutive windows in chronological order as a sample sequence; construct a label corresponding to each sample sequence, wherein the end time of the time interval covered by each sample sequence is denoted as the first moment, the prediction window is [first moment, first moment + preset time length], the arithmetic mean of the heart rate sampling points in the prediction window is calculated as the true heart rate, the arithmetic mean of the systolic blood pressure sampling points is calculated as the true systolic blood pressure, and the arithmetic mean of the blood oxygen saturation sampling points is calculated as the true blood oxygen saturation; obtain the true label of consciousness recovery at the end time of the prediction window; use the true heart rate, true systolic blood pressure, true blood oxygen saturation, and true label of consciousness recovery as the label of each sample sequence; construct a long short-term memory network model, train the long short-term memory network model based on multiple sample sequences and their labels, and obtain the time-series prediction model.

[0038] In one specific embodiment of this disclosure, the construction unit 23 further includes a training unit 231.

[0039] Training unit 231 is used to construct a long short-term memory network model, which includes an input layer, an LSTM hidden layer, a Dropout layer, and a fully connected output layer. The long short-term memory network model is trained with the feature vector sequence as input and the corresponding label as output. The loss function is a weighted sum of the mean squared error of heart rate prediction, the mean squared error of systolic blood pressure prediction, the mean squared error of blood oxygen saturation prediction, and the cross-entropy loss of consciousness recovery prediction. After training, a time-series prediction model is obtained.

[0040] It should be noted that the specific methods by which each module performs operations in the system described in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0041] Example 3 Corresponding to the above method embodiments, this disclosure also provides a device for multi-dimensional monitoring of the recovery status of anesthesia and surgery. The device for multi-dimensional monitoring of the recovery status of anesthesia and surgery described below can be referred to in correspondence with the method for multi-dimensional monitoring of the recovery status of anesthesia and surgery described above.

[0042] Figure 3 This is a block diagram illustrating a device 300 for multi-dimensional monitoring of the recovery status during anesthesia surgery, according to an exemplary embodiment. Figure 3 As shown, the device 300 for multi-dimensional monitoring of anesthesia recovery status may include: a processor 301 and a memory 302. The device 300 may also include one or more of a multimedia component 303, an I / O interface 304, and a communication component 305.

[0043] The processor 301 controls the overall operation of the device 300 for multi-dimensional monitoring of anesthesia recovery status to complete all or part of the steps in the method for multi-dimensional monitoring of anesthesia recovery status. The memory 302 stores various types of data to support the operation of the device 300, including instructions for any application or method operating on the device 300, as well as application-related data such as contact data, sent and received messages, images, audio, video, etc. The memory 302 can be implemented using any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 303 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 302 or transmitted via the communication component 305. The audio component also includes at least one speaker for outputting audio signals. I / O interface 304 provides an interface between processor 301 and other interface modules, such as keyboards, mice, and buttons. These buttons can be virtual or physical. Communication component 305 is used for wired or wireless communication between the multi-dimensional monitoring device 300 for anesthesia recovery and other devices. Wireless communication includes Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 305 may include a Wi-Fi module, a Bluetooth module, or an NFC module.

[0044] In an exemplary embodiment, the device 300 for multidimensional monitoring of the recovery status of anesthesia surgery may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described method for multidimensional monitoring of the recovery status of anesthesia surgery.

[0045] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the method for multidimensional monitoring of the recovery status of anesthesia surgery described above. For example, the computer-readable storage medium may be the memory 302 including the program instructions described above, which may be executed by the processor 301 of the device 300 for multidimensional monitoring of the recovery status of anesthesia surgery to complete the method for multidimensional monitoring of the recovery status of anesthesia surgery described above.

[0046] Example 4 Corresponding to the above method embodiments, this disclosure also provides a readable storage medium. The readable storage medium described below can be referred to in conjunction with the method for multi-dimensional monitoring of the recovery status of anesthesia surgery described above.

[0047] A readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method described in the above-described method embodiment for multi-dimensional monitoring of the recovery status of anesthesia surgery.

[0048] Specifically, the readable storage medium can be a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, or any other readable storage medium capable of storing program code.

[0049] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for multi-dimensional monitoring of recovery status during anesthesia surgery, characterized in that, include: Acquire the patient's vital signs time-series data and demographic characteristics within a preset time period. The vital signs time-series data includes electrocardiogram time-series data, blood pressure time-series data, and blood oxygen saturation time-series data. The patient's demographic characteristics include age, sex, and body mass index. The cutoff time for the preset time period is the current time. Based on vital sign time-series data and patient demographic characteristics, feature vector sequences corresponding to preset time periods are constructed; a time-series prediction model is then built. The feature vector sequence is input into the time-series prediction model, which outputs the patient's predicted vital signs and the probability of regaining consciousness within the future prediction window. The resuscitation index is calculated based on the predicted vital signs and the probability of regaining consciousness, and different levels of early warning information are output based on the relationship between the resuscitation index and a preset threshold.

2. The method for multi-dimensional monitoring of anesthesia recovery status according to claim 1, characterized in that, Based on time-series data of vital signs and the demographic characteristics of patients, a feature vector sequence corresponding to a preset time period is constructed, including: The time-series data of electrocardiogram (ECG), blood pressure, and blood oxygen saturation are continuously segmented according to a preset fixed time length to obtain a series of M consecutive and non-overlapping windows. Each window contains the corresponding ECG, blood pressure, and blood oxygen saturation sub-time-series data. The feature information of the ECG, blood pressure, and blood oxygen saturation sub-time-series data in each window is calculated. The corresponding feature information is concatenated with the patient's demographic characteristics to obtain the feature vector for each window. The feature vectors of all windows are stacked in chronological order to form the feature vector sequence corresponding to the preset time period.

3. The method for multi-dimensional monitoring of anesthesia recovery status according to claim 2, characterized in that, Calculate the feature information of the ECG sub-time series data, blood pressure sub-time series data, and blood oxygen saturation sub-time series data under each window. Concatenate the corresponding feature information of the three data with the patient's demographic features to obtain the feature vector under each window, including: Change point detection is performed on the ECG sub-time series data to identify change points, and the ECG sub-time series data is divided into multiple segments based on the change points. R wave detection is performed on each segment to obtain the RR interval sequence of successive heartbeats. The square of the difference between adjacent RR intervals is calculated, and all squares are summed to obtain the sum of squares. The sum of squares is divided by the total number of heartbeats in the segment minus 1 to obtain the first value. The square root of the first value is used to obtain the feature index of each segment. The mean and standard deviation of all feature indices are calculated and used as the first feature information of the ECG sub-time series data. Change point detection is performed on the blood pressure sub-time series data to identify change points, and the blood pressure sub-time series data is divided into multiple segments based on the change points; the mean and standard deviation of all systolic blood pressure values ​​in each segment are calculated, and the standard deviation is divided by the mean to obtain the systolic blood pressure coefficient of variation. The systolic blood pressure coefficient of variation of each segment is used as the feature index of each segment; the mean and standard deviation of all feature indices are calculated and used as the second feature information of the blood pressure sub-time series data. Change point detection is performed on the blood oxygen saturation sub-time series data to identify change points. Based on the change points, the blood oxygen saturation sub-time series data is divided into multiple segments. The maximum blood oxygen saturation in each segment is subtracted from the minimum blood oxygen saturation to obtain the feature index of each segment. The mean and standard deviation of all feature indexes are calculated and used as the third feature information of the blood oxygen saturation sub-time series data. The first feature information, the second feature information, and the third feature information are concatenated with age, gender, and body mass index to obtain the feature vector.

4. The method for multi-dimensional monitoring of anesthesia recovery status according to claim 1, characterized in that, Constructing a time series prediction model includes: Complete time-series ECG, blood pressure, and oxygen saturation data from admission to discharge of multiple historical patients were acquired, along with their demographic characteristics. The complete ECG, blood pressure, and oxygen saturation data were continuously segmented according to a preset fixed time length, resulting in a series of consecutive, non-overlapping windows. Each window contained corresponding historical ECG, blood pressure, and oxygen saturation sub-time-series data. Feature information for each window's historical ECG, blood pressure, and oxygen saturation sub-time-series data was calculated, and this feature information was concatenated with the historical patients' demographic characteristics to obtain the historical feature vector for each window. The historical feature vectors of M consecutive windows are stacked in chronological order to form a sample sequence. A label is constructed for each sample sequence, where the end time of the time interval covered by each sample sequence is denoted as the first moment, and the prediction window is [first moment, first moment + preset time length]. The arithmetic mean of the heart rate sampling points within the prediction window is calculated as the true heart rate, the arithmetic mean of the systolic blood pressure sampling points is calculated as the true systolic blood pressure, and the arithmetic mean of the blood oxygen saturation sampling points is calculated as the true blood oxygen saturation. The true label of consciousness recovery at the end time of the prediction window is obtained. The true heart rate, true systolic blood pressure, true blood oxygen saturation, and true label of consciousness recovery are used as the label for each sample sequence. A Long Short-Term Memory (LSTM) network model is constructed and trained based on multiple sample sequences and their labels to obtain a time-series prediction model.

5. The method for multi-dimensional monitoring of anesthesia recovery status according to claim 1, characterized in that, A long short-term memory (LSTM) network model is constructed and trained based on multiple sample sequences and their labels to obtain a time-series prediction model, including: A Long Short-Term Memory (LSTM) network model is constructed, which includes an input layer, an LSTM hidden layer, a Dropout layer, and a fully connected output layer. The LSM network model is trained using a loss function, which is a weighted sum of the mean squared error of heart rate prediction, the mean squared error of systolic blood pressure prediction, the mean squared error of blood oxygen saturation prediction, and the cross-entropy loss of consciousness recovery prediction. After training, a time-series prediction model is obtained.

6. A system for multi-dimensional monitoring of the recovery status during anesthesia surgery, characterized in that, include: The acquisition module is used to acquire the patient's vital signs time series data and the patient's demographic characteristics within a preset time period. The vital signs time series data includes electrocardiogram time series data, blood pressure time series data, and blood oxygen saturation time series data; the patient's demographic characteristics include age, gender, and body mass index. The cutoff time of the preset time period is the current time. The module is used to construct feature vector sequences corresponding to preset time periods based on vital sign time-series data and patient demographic characteristics; and to construct time-series prediction models. The prediction module is used to input the feature vector sequence into the time-series prediction model, output the patient's predicted vital signs and the probability of regaining consciousness within the future prediction window, calculate the resuscitation index based on the predicted vital signs and the probability of regaining consciousness, and output different levels of early warning information based on the relationship between the resuscitation index and a preset threshold.

7. The system for multi-dimensional monitoring of anesthesia recovery status according to claim 6, characterized in that, Build modules, including: The segmentation unit is used to continuously segment the time-series data of electrocardiogram (ECG), blood pressure, and blood oxygen saturation according to a preset fixed time length, resulting in a series of M consecutive and non-overlapping windows. Each window contains the corresponding ECG sub-time-series data, blood pressure sub-time-series data, and blood oxygen saturation sub-time-series data. The feature information of the ECG, blood pressure, and blood oxygen saturation sub-time-series data under each window is calculated, and the corresponding feature information is concatenated with the patient's demographic characteristics to obtain the feature vector under each window. The feature vectors of all windows are stacked in chronological order to form the feature vector sequence corresponding to the preset time period.

8. The method for multi-dimensional monitoring of anesthesia recovery status according to claim 7, characterized in that, Segmentation unit, including: The first calculation unit is used to detect change points in the ECG sub-time series data, identify change points, and divide the ECG sub-time series data into multiple segments based on the change points; to perform R-wave detection on each segment to obtain the RR interval sequence of successive heartbeats, calculate the square of the difference between adjacent RR intervals, sum all the squares to obtain the sum of squares; divide the sum of squares by the total number of heartbeats in the segment minus 1 to obtain the first value, take the square root of the first value to obtain the feature index of each segment, calculate the mean and standard deviation of all feature indices and use them as the first feature information of the ECG sub-time series data; The second calculation unit is used to detect change points in the blood pressure sub-time series data, identify change points, and divide the blood pressure sub-time series data into multiple segments based on the change points; calculate the mean and standard deviation of all systolic blood pressure values ​​in each segment, divide the standard deviation by the mean to obtain the systolic blood pressure coefficient of variation, and use the systolic blood pressure coefficient of variation of each segment as the feature index of each segment; calculate the mean and standard deviation of all feature indices and use them as the second feature information of the blood pressure sub-time series data; The third calculation unit is used to detect change points in the blood oxygen saturation sub-time series data, identify change points, and divide the blood oxygen saturation sub-time series data into multiple segments based on the change points. The maximum blood oxygen saturation in each segment is subtracted from the minimum blood oxygen saturation to obtain the feature index of each segment. The mean and standard deviation of all feature indexes are calculated and used as the third feature information of the blood oxygen saturation sub-time series data. The concatenation unit is used to concatenate the first feature information, the second feature information, the third feature information with age, gender, and body mass index to obtain a feature vector.

9. The system for multi-dimensional monitoring of anesthesia recovery status according to claim 6, characterized in that, Build modules, including: The acquisition unit is used to acquire complete ECG time-series data, complete blood pressure time-series data, and complete blood oxygen saturation time-series data of multiple historical patients from admission to discharge, as well as the demographic characteristics of historical patients. The complete ECG time-series data, complete blood pressure time-series data, and complete blood oxygen saturation time-series data are continuously segmented according to a preset fixed time length, resulting in a series of consecutive and non-overlapping windows. Each window contains corresponding historical ECG sub-time-series data, historical blood pressure sub-time-series data, and historical blood oxygen saturation sub-time-series data. The feature information of the historical ECG sub-time-series data, historical blood pressure sub-time-series data, and historical blood oxygen saturation sub-time-series data under each window is calculated. The corresponding feature information of the three is concatenated with the demographic characteristics of historical patients to obtain the historical feature vector under each window. The system constructs a unit to stack historical feature vectors from M consecutive windows in chronological order as a sample sequence. It then constructs labels for each sample sequence, where the end time of the time interval covered by each sample sequence is designated as the first moment, and the prediction window is defined as [first moment, first moment + preset time length]. The arithmetic mean of heart rate sampling points within the prediction window is calculated as the true heart rate, the arithmetic mean of systolic blood pressure sampling points is calculated as the true systolic blood pressure, and the arithmetic mean of blood oxygen saturation sampling points is calculated as the true blood oxygen saturation. The system obtains the true label of consciousness recovery at the end of the prediction window. The true heart rate, true systolic blood pressure, true blood oxygen saturation, and the true label of consciousness recovery are used as labels for each sample sequence. Finally, a long short-term memory (LSTM) network model is constructed and trained based on multiple sample sequences and their labels to obtain a time-series prediction model.

10. The system for multi-dimensional monitoring of anesthesia recovery status according to claim 9, characterized in that, Building blocks, including: The training unit is used to construct a Long Short-Term Memory (LSTM) network model, which includes an input layer, an LSTM hidden layer, a Dropout layer, and a fully connected output layer. The LSM network model is trained using a loss function, which is a weighted sum of the mean squared error of heart rate prediction, the mean squared error of systolic blood pressure prediction, the mean squared error of blood oxygen saturation prediction, and the cross-entropy loss of consciousness recovery prediction. After training, a time-series prediction model is obtained.

Images