A patient post-operative pain risk prediction system

Abstract

The application relates to the technical field of postoperative pain risk prediction, in particular to a patient postoperative pain risk prediction system. The system collects multi-modal physiological signals of a patient in an anesthesia recovery period and obtains a standardized physiological signal sequence through preprocessing; extracts an autonomic nerve feature subset and a motion behavior feature subset according to the standardized physiological signal sequence; constructs an autonomic nerve feature vector after normalizing each feature value in the autonomic nerve feature subset, and constructs a motion behavior feature vector after normalizing each feature value in the motion behavior feature subset; calculates a coupling synchronization degree between the autonomic nerve feature vector and the motion behavior feature vector; obtains a restlessness type discrimination result through restlessness type discrimination analysis of the coupling synchronization degree; outputs early warning information when the restlessness type discrimination result is pain restlessness; and realizes accurate identification and early warning of pain restlessness of the patient in the anesthesia recovery period.

Description

Technical Field

[0001] This invention relates to the field of postoperative pain risk prediction technology, specifically a postoperative pain risk prediction system for patients. Background Technology

[0002] Postoperative pain risk prediction is a crucial component of perioperative medical management, playing a vital role in ensuring patient safety and promoting postoperative recovery. After surgery, patients are typically transferred to the postanesthesia monitoring and treatment room for monitoring and management during the anesthesia recovery period, which generally lasts 0.3 to 2 hours or even longer. During the anesthesia recovery period, patients are in a transitional phase from anesthesia to gradually regaining consciousness, and are prone to a series of pathophysiological changes and internal environment disturbances, among which agitation is one of the most common clinical manifestations. Agitation during anesthesia recovery is mainly divided into two types: one is agitation caused by non-pain factors such as anesthetic drug metabolism and central nervous system dysfunction; the other is painful agitation triggered by surgical trauma. Although these two types of agitation are highly similar in their external manifestations, such as involuntary limb movements, increased heart rate, and elevated blood pressure, their pathological mechanisms and clinical management strategies are quite different: painful agitation requires timely administration of analgesics to relieve patient suffering, while agitation during anesthesia recovery is mainly relieved through sedation or by waiting for the natural metabolism of anesthetic drugs.

[0003] However, patients emerging from anesthesia are in a state of impaired consciousness, and their cognitive function and language expression abilities have not yet recovered. They are unable to accurately describe their feelings to medical staff through traditional pain self-rating scales or verbal communication, which poses a significant challenge to the differential diagnosis of agitation types. Currently, clinical practice mainly relies on the experience and behavioral observation of medical staff to distinguish between the two types of agitation. However, due to the significant overlap in physiological indicators and behavioral manifestations between the two, even experienced medical staff find it difficult to make an accurate judgment. Existing postoperative pain assessment techniques mostly focus on the quantitative analysis of pain intensity levels, attempting to assess the degree of pain through physiological parameters or behavioral characteristics, but neglecting a more fundamental and crucial preliminary question—whether the patient's current agitation is actually caused by pain. This deviation in the technical approach has led to a dilemma in clinical practice: if painful agitation is misdiagnosed as anesthesia recovery agitation, the patient will suffer from avoidable severe pain, which will not only affect the postoperative recovery process but may also trigger serious complications such as cardiovascular stress. Conversely, if anesthesia recovery agitation is misdiagnosed as painful agitation and analgesics are administered, the patient will face unnecessary medication risks such as respiratory depression and delayed awakening, while also wasting medical resources.

[0004] Therefore, there is an urgent need for a system that can accurately distinguish between anesthesia recovery agitation and pain-related agitation within the anesthesia recovery window, which patients cannot self-report, in order to achieve early identification and warning of pain-related agitation, thereby improving the level of medical safety and patient comfort during anesthesia recovery. Summary of the Invention

[0005] (1) Technical problems to be solved This invention provides a postoperative pain risk prediction system for patients, which addresses the problem of accurately distinguishing between postoperative agitation and painful agitation in patients who are unable to self-report due to impaired consciousness during the anesthesia recovery period.

[0006] (2) Technical solution To achieve the above objectives, the present invention provides a postoperative pain risk prediction system for patients, the system comprising: a standardized physiological signal sequence generation module, a coupling synchronization degree calculation module, and an agitation type discrimination analysis and early warning module, wherein each module is sequentially connected in communication; A standardized physiological signal sequence generation module is used to collect multimodal physiological signals from patients in the anesthesia recovery period and preprocess the multimodal physiological signals to obtain a standardized physiological signal sequence.

[0007] The coupling synchronization degree calculation module is used to extract agitation feature set based on the standardized physiological signal sequence, separate autonomic neural feature subset and motor behavior feature subset from the agitation feature set; construct autonomic neural feature vector after normalizing each feature value in the autonomic neural feature subset, construct motor behavior feature vector after normalizing each feature value in the motor behavior feature subset; and calculate the coupling synchronization degree between the autonomic neural feature vector and the motor behavior feature vector.

[0008] The agitation type discrimination analysis and early warning module is used to obtain agitation type discrimination results by analyzing the coupling synchronization degree. The agitation type discrimination results include agitation during anesthesia recovery and painful agitation. When the agitation type discrimination result is painful agitation, an early warning message is output.

[0009] Furthermore, the coupling synchronization calculation module includes the following steps: The time-frequency domain transformations of the autonomic nervous system feature vector and the motor behavior feature vector are performed respectively to obtain the time-frequency feature matrix of the autonomic nervous system and the time-frequency feature matrix of the motor behavior.

[0010] The phase-locking values ​​of the autonomic nervous system time-frequency feature matrix and the motor behavior time-frequency feature matrix in each frequency band are calculated, and the phase-locking values ​​in each frequency band are weighted and fused to obtain the cross-modal phase synchronization index. The mutual information between the autonomic neural feature vector and the motor behavior feature vector is calculated to obtain the feature mutual information value. .

[0011] According to the cross-modal phase synchronization index and feature mutual information value Calculate the coupling synchronization degree The coupling synchronization degree The calculation formula is: .

[0012] in, and These are the cross-modal phase synchronization index and the fusion index of feature mutual information values, respectively. This is the coupling suppression coefficient; This is the coupling scale parameter, used to normalize the exponential term.

[0013] Furthermore, the disturbance type discrimination analysis and early warning module includes: The agitation discrimination calculation module is used to construct the prototype vectors of agitation during anesthesia recovery and painful agitation; it concatenates the autonomic nerve feature vector with the motor behavior feature vector to obtain the current agitation feature vector, calculates the feature space distance between the current agitation feature vector and the prototype vector of agitation during anesthesia recovery to obtain the first prototype distance, and calculates the feature space distance between the current agitation feature vector and the prototype vector of painful agitation to obtain the second prototype distance.

[0014] The agitation discrimination calculation value is calculated based on the coupling synchronization degree, the first prototype distance, and the second prototype distance, and the agitation type discrimination result is determined based on the agitation discrimination calculation value.

[0015] Furthermore, the noise discrimination calculation module includes: The coupling modulation coefficient calculation module is used to calculate the coupling modulation coefficient by means of the coupling synchronization degree through an autonomic nervous system-motor dual-channel dynamic coupling modulation mechanism. The coupling modulation coefficient The calculation formula is: .

[0016] in, Indicates the degree of coupling synchronization. This represents the preset coupling synchronization baseline value. This represents the mapping slope parameter. It is a time-varying coupling attenuation factor. This is the cross-modal consistency correction factor.

[0017] The noise discrimination calculation value calculation submodule is used to calculate the noise discrimination value based on the distance of the first prototype. Second prototype distance Calculate the prototype distance difference value The prototype distance difference value According to the coupling modulation coefficient Distance difference with prototype The noise discrimination value is calculated.

[0018] Furthermore, the coupling modulation coefficient calculation module includes the following steps: The time-varying coupling attenuation factor The calculation formula is: .

[0019] in, The attenuation intensity coefficient, For reference coupling synchronization degree, This refers to the physiological rhythm cycle during the anesthesia recovery period. This is the initial phase; The cumulative time after the patient enters the anesthesia recovery period; the time-varying coupling attenuation factor is used to capture the periodic fluctuation characteristics of the autonomic nervous system during the anesthesia recovery period.

[0020] Furthermore, the coupling modulation coefficient calculation module also includes the following steps: The cross-modal consistency correction factor The calculation formula is: .

[0021] in, The cosine similarity between the autonomic nervous system feature vector and the motor behavior feature vector is given. The upper limit of cosine similarity is defined as follows: The cross-modal consistency correction factor is used to enhance the modulation effect when the autonomic neural feature vector and the motor behavior feature vector exhibit directional consistency.

[0022] Furthermore, the noisy discrimination calculation value calculation submodule includes: A nonlinear tensor coupling discriminant function is constructed based on the coupling modulation coefficients and the prototype distance difference value. The nonlinear tensor coupling discriminant function is as follows: .

[0023] in, This represents the calculated value for the agitation discrimination. Indicates the boundary compensation parameters; The adaptive boundary adjustment function is calculated using the following formula: .

[0024] in, For boundary expansion coefficient, This represents the upper limit of coupling synchronization. The reference distance parameter is used; the adaptive boundary adjustment function expands the discrimination boundary when the coupling synchronization degree is low, which is used to improve the robustness of noise type discrimination.

[0025] Furthermore, the noise discrimination calculation module also includes the following steps: When the calculated agitation value is greater than the preset positive discrimination threshold, the agitation type discrimination result is determined to be painful agitation.

[0026] When the calculated agitation value is less than the preset negative discrimination threshold, the agitation type discrimination result is determined to be agitation during anesthesia recovery.

[0027] When the agitation discrimination calculation value is between the preset positive discrimination threshold and the preset negative discrimination threshold, if the coupling synchronization degree is greater than the preset coupling judgment threshold, the agitation type discrimination result is determined to be painful agitation; otherwise, the agitation type discrimination result is determined to be agitation during anesthesia recovery.

[0028] Furthermore, the noisy discrimination calculation value calculation submodule also includes the following steps: Acquire historical agitation samples with a coupling synchronization degree lower than a preset coupling judgment threshold, and calculate the average discrimination distance by taking the mean of the absolute values ​​of the prototype distance differences of each historical agitation sample. ; Based on the average discrimination distance , Discrimination sensitivity adjustment coefficient and historical sample distribution dispersion Calculated boundary compensation parameters The boundary compensation parameters The calculation formula is: .

[0029] in, ∈(0,1], This is the dispersion weighting coefficient.

[0030] (3) Beneficial effects Compared with the prior art, the beneficial effects of the present invention are: 1. By collecting multimodal physiological signals from patients during anesthesia recovery, constructing autonomic nervous system feature vectors and motor behavior feature vectors respectively, and calculating the coupling synchronization degree between the two, the essential differences in neuromotor response patterns between anesthesia recovery agitation and pain-related agitation can be effectively captured. Pain-related agitation is characterized by a highly coordinated response between the autonomic nervous system and motor behavior, while anesthesia recovery agitation exhibits relatively independent disordered fluctuation characteristics. Based on this physiological mechanism difference, the two types of agitation can be accurately distinguished.

[0031] 2. It can promptly identify painful agitation and output early warning information during the critical window period when patients are in a state of impaired consciousness and cannot report it themselves. This enables medical staff to provide timely and effective analgesic intervention for patients who are truly in pain, while avoiding unnecessary analgesic medication for agitated patients during anesthesia recovery. This ensures the timeliness of treatment for patients in pain and reduces the medical risks such as respiratory depression and delayed awakening faced by non-pain patients due to misuse of analgesics, thereby improving the overall medical safety level and patient comfort during anesthesia recovery. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the module composition of a postoperative pain risk prediction system for patients according to Embodiment 1 of the present invention. Detailed Implementation

[0033] 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 embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] Example 1: As Figure 1 As shown, this embodiment provides a postoperative pain risk prediction system for patients. The system includes: a standardized physiological signal sequence generation module, a coupling synchronization degree calculation module, and an agitation type discrimination analysis and early warning module, with each module being connected to the other in sequence. A standardized physiological signal sequence generation module is used to collect multimodal physiological signals from patients in the anesthesia recovery period and preprocess the multimodal physiological signals to obtain a standardized physiological signal sequence.

[0035] For example, the anesthesiology department of a tertiary hospital used this system to monitor the anesthesia recovery period of a patient after laparoscopic cholecystectomy. The patient was a 52-year-old male weighing 68 kg. The surgery lasted 75 minutes and was performed under general anesthesia. After the surgery, the patient was transferred to the post-anesthesia monitoring room, 8 minutes after the anesthetic infusion was stopped.

[0036] The nurse connected an electrocardiogram (ECG) monitor, pulse oximeter, electromyography (EMG) sensor, and triaxial accelerometer to the patient. The ECG monitor acquired the patient's electrocardiogram (ECG) signals and arterial blood pressure waveforms, with a sampling frequency set to 500 data points per second. The pulse oximeter acquired photoplethysmography (PPG) pulse wave signals, with a sampling frequency of 100 data points per second. The EMG sensor was attached to the surface of the patient's bilateral deltoid and quadriceps muscles, acquiring muscle electrical activity signals, with a sampling frequency of 1000 data points per second. The triaxial accelerometer was fixed to the patient's right wrist and right ankle, acquiring limb movement signals, with a sampling frequency of 50 data points per second.

[0037] Twenty-three minutes into the anesthesia recovery period, the patient began to exhibit agitation, with involuntary movements of the limbs. The heart rate increased from 72 beats per minute in the early postoperative period to 98 beats per minute, and the systolic blood pressure rose from 112 mmHg to 138 mmHg. The system then initiated the agitation analysis procedure.

[0038] The acquired raw multimodal physiological signals typically contain various noise and interference components, thus requiring preprocessing to obtain a standardized physiological signal sequence. Electrocardiogram (ECG) signals are bandpass filtered from 0.05 Hz to 150 Hz to remove baseline drift and high-frequency noise interference. Electromyography (EMG) signals are bandpass filtered from 20 Hz to 450 Hz and then subjected to a 50 Hz power frequency notch filter. Accelerometer signals are low-pass filtered to remove high-frequency jitter noise. All signals are aligned according to a unified time reference to form a standardized physiological signal sequence, and the analysis window is set to 60 consecutive seconds of data following the onset of agitation.

[0039] The coupling synchronization degree calculation module is used to extract an agitation feature set based on the standardized physiological signal sequence, separate an autonomic neural feature subset and a motor behavior feature subset from the agitation feature set; construct an autonomic neural feature vector after normalizing each feature value in the autonomic neural feature subset, construct a motor behavior feature vector after normalizing each feature value in the motor behavior feature subset; and calculate the coupling synchronization degree between the autonomic neural feature vector and the motor behavior feature vector. For example, the agitation feature set extracted from the standardized physiological signal sequence by the system needs to be further separated into two subsets, corresponding to the activity features of the autonomic nervous system and the activity features of somatic motor behavior, respectively. Regarding autonomic nervous system features, the system calculates heart rate variability indices: the root mean square of the difference between adjacent heartbeats is 42.3 milliseconds, and the ratio of low-frequency power to high-frequency power is 3.87. These two indices reflect the balance between the sympathetic and parasympathetic nervous systems. The system also calculates the pulse wave conduction time variation coefficient as 0.089, skin conductance amplitude as 4.2 microsiemens, and blood pressure variability index as 0.156. Regarding motor behavior features, the system analyzes electromyographic signals to obtain muscle activation intensity: the root mean square amplitude of the deltoid muscle is 78.5 microvolts, and the root mean square amplitude of the quadriceps femoris muscle is 92.3 microvolts. Accelerometer signal analysis shows a limb movement intensity index of 0.847, a principal component of movement frequency of 2.3 Hz, and a movement asymmetry index of 0.234, indicating a certain difference in the amplitude of movement between the left and right limbs.

[0040] The system performs min-max normalization on five autonomic nervous system-related features, mapping them to the interval between 0 and 1. The root mean square value of heart rate variability (RMS) of 42.3 ms corresponds to a normalized value of 0.72; the low-frequency to high-frequency power ratio of 3.87 corresponds to a normalized value of 0.81; the coefficient of variation of pulse wave conduction time of 0.089 corresponds to a normalized value of 0.63; the amplitude of skin conductance response of 4.2 microSiemens corresponds to a normalized value of 0.58; and the blood pressure variability index of 0.156 corresponds to a normalized value of 0.69. Based on this, an autonomic nervous system feature vector is constructed, with its five components being 0.72, 0.81, 0.63, 0.58, and 0.69, respectively.

[0041] The five feature values ​​related to motor behavior were also normalized. The root mean square amplitude of the deltoid muscle (RMS amplitude) of 78.5 μV corresponds to a normalized value of 0.68, the RMS amplitude of the quadriceps muscle (RMS amplitude) of 92.3 μV corresponds to a normalized value of 0.74, the limb movement intensity index of 0.847 corresponds to a normalized value of 0.79, the principal component of movement frequency of 2.3 Hz corresponds to a normalized value of 0.65, and the movement asymmetry index of 0.234 corresponds to a normalized value of 0.43. Thus, a motor behavior feature vector was constructed, with its five components being 0.68, 0.74, 0.79, 0.65, and 0.43, respectively.

[0042] The coupling synchronization calculation module includes the following steps: The time-frequency domain transformations of the autonomic nervous system feature vector and the motor behavior feature vector are performed respectively to obtain the time-frequency feature matrix of the autonomic nervous system and the time-frequency feature matrix of the motor behavior.

[0043] The phase-locking values ​​of the autonomic nervous system time-frequency feature matrix and the motor behavior time-frequency feature matrix in each frequency band are calculated, and the phase-locking values ​​in each frequency band are weighted and fused to obtain the cross-modal phase synchronization index. The mutual information between the autonomic neural feature vector and the motor behavior feature vector is calculated to obtain the feature mutual information value. .

[0044] According to the cross-modal phase synchronization index and feature mutual information value Calculate the coupling synchronization degree The coupling synchronization degree The calculation formula is: .

[0045] in, and These are the cross-modal phase synchronization index and the fusion index of feature mutual information values, respectively. This is the coupling suppression coefficient; This is the coupling scale parameter, used to normalize the exponential term.

[0046] For example, coupling synchronization is a numerical metric used to quantify the degree of correlation and synchronous change between two feature vectors. Short-time Fourier transforms are performed on the autonomic neural feature vector and the motor behavior feature vector, respectively, converting the time-domain signals into time-frequency domain representations, resulting in the autonomic neural time-frequency feature matrix and the motor behavior time-frequency feature matrix. The system calculates the phase-locking value of the two time-frequency feature matrices in each frequency band. In the low-frequency band (0.04 to 0.15 Hz), the phase-locking value is 0.73, and the weighting coefficient is set to 0.4. In the high-frequency band (0.15 to 0.4 Hz), the phase-locking value is 0.81, and the weighting coefficient is set to 0.35. In the ultra-high-frequency band (0.4 to 1.0 Hz), the phase-locking value is 0.67, and the weighting coefficient is set to 0.25. The phase-locked values ​​of the three frequency bands are weighted and summed: 0.73 multiplied by 0.4 gives 0.292, 0.81 multiplied by 0.35 gives 0.284, and 0.67 multiplied by 0.25 gives 0.168. The sum of the three terms gives a cross-modal phase synchronization index of 0.744.

[0047] The mutual information of features is calculated by constructing joint and marginal probability distributions for the two vectors and then using kernel density estimation to calculate the mutual information, resulting in a mutual information value of 0.682. The fusion exponents are set to 0.6 and 0.4, the coupling suppression coefficient is set to 0.15, and the coupling scale parameter is set to 0.5. First, the 0.6 power of the cross-modal phase synchronization exponent is calculated: 0.744 raised to the power of 0.6 equals 0.832. Then, the 0.4 power of the mutual information value is calculated: 0.682 raised to the power of 0.4 equals 0.862. Multiplying these two terms yields 0.717. The numerator of the exponent term is calculated: 0.744 multiplied by 0.682 equals 0.507, divided by 0.5 equals 1.014, negative value is taken, and the natural exponent is calculated to obtain 0.363, multiplied by 0.15 equals 0.054, and added by 1 equals 1.054. The final coupling synchronization degree is 0.717 divided by 1.054, which gives 0.680.

[0048] The agitation type discrimination analysis and early warning module is used to obtain agitation type discrimination results from the coupling synchronization degree through agitation type discrimination analysis. The agitation type discrimination results include agitation during anesthesia emergence and painful agitation. When the agitation type discrimination result is painful agitation, an early warning message is output. The early warning message can be presented in various ways, such as popping up an alert window on the display screen of the bedside monitor or central monitoring station, issuing a prompt sound and flashing light through an audible and visual alarm device, or pushing a notification message to the mobile terminals of medical staff. The early warning message may include key information such as the patient's bed number, current time, coupling synchronization degree value, and discrimination confidence level, which facilitates medical staff to quickly locate and respond to the situation.

[0049] The disturbance type discrimination analysis and early warning module includes: The agitation discrimination calculation module is used to construct the prototype vectors of agitation during anesthesia recovery and painful agitation; it concatenates the autonomic nerve feature vector with the motor behavior feature vector to obtain the current agitation feature vector, calculates the feature space distance between the current agitation feature vector and the prototype vector of agitation during anesthesia recovery to obtain the first prototype distance, and calculates the feature space distance between the current agitation feature vector and the prototype vector of painful agitation to obtain the second prototype distance.

[0050] For example, the system enters the agitation type discrimination analysis stage. The system pre-stores prototype vectors for agitation during anesthesia recovery and for painful agitation. These two prototype vectors are category centers calculated based on clinically diagnosed agitation samples from the hospital's historical case database. The ten components of the prototype vector for agitation during anesthesia recovery are 0.65, 0.73, 0.55, 0.48, 0.61, 0.52, 0.58, 0.63, 0.71, and 0.38, respectively. The ten components of the prototype vector for painful agitation are 0.71, 0.79, 0.61, 0.56, 0.68, 0.67, 0.73, 0.78, 0.64, and 0.45, respectively.

[0051] The system concatenates the current patient's autonomic nervous system feature vector with the motor behavior feature vector to form a ten-dimensional current agitation feature vector, with components of 0.72, 0.81, 0.63, 0.58, 0.69, 0.68, 0.74, 0.79, 0.65, and 0.43, respectively.

[0052] The system calculates the Euclidean distance between the current agitation feature vector and the prototype agitation vector during anesthesia recovery. The squares of the differences are calculated component by component: the square of the first component difference (0.07) is 0.0049, the square of the second component difference (0.08) is 0.0064, the square of the third component difference (0.08) is 0.0064, the square of the fourth component difference (0.10) is 0.0100, the square of the fifth component difference (0.08) is 0.0064, the square of the sixth component difference (0.16) is 0.0256, the square of the seventh component difference (0.16) is 0.0256, the square of the eighth component difference (0.16) is 0.0256, the square of the ninth component difference (-0.06) is 0.0036, and the square of the tenth component difference (0.05) is 0.0025. The sum of the ten squares is 0.117, and the square root of this sum gives the first prototype distance as 0.342.

[0053] The system calculates the Euclidean distance between the current agitation feature vector and the painful agitation prototype vector. The square of the difference is calculated component by component: the square of the difference of the first component (0.01) is 0.0001, the square of the difference of the second component (0.02) is 0.0004, the square of the difference of the third component (0.02) is 0.0004, the square of the difference of the fourth component (0.02) is 0.0004, the square of the difference of the fifth component (0.01) is 0.0001, the square of the difference of the sixth component (0.01) is 0.0001, the square of the difference of the seventh component (0.01) is 0.0001, the square of the difference of the eighth component (0.01) is 0.0001, the square of the difference of the ninth component (0.01) is 0.0001, and the square of the difference of the tenth component (-0.02) is 0.0004. The sum of the ten squares is 0.0022, and the square root of this sum gives the second prototype distance as 0.047.

[0054] The agitation discrimination calculation value is calculated based on the coupling synchronization degree, the first prototype distance, and the second prototype distance, and the agitation type discrimination result is determined based on the agitation discrimination calculation value.

[0055] The noise discrimination calculation module also includes the following steps: When the calculated agitation value is greater than the preset positive discrimination threshold, the agitation type discrimination result is determined to be painful agitation.

[0056] When the calculated agitation value is less than the preset negative discrimination threshold, the agitation type discrimination result is determined to be agitation during anesthesia recovery.

[0057] When the calculated agitation threshold is between a preset positive and a preset negative threshold, if the coupling synchronization degree is greater than a preset coupling judgment threshold, the agitation type is determined to be painful agitation; otherwise, it is determined to be agitation during anesthesia recovery. A three-interval discrimination rule is used, dividing the calculated agitation threshold into three intervals: clearly defined as painful agitation, clearly defined as agitation during anesthesia recovery, and a boundary region requiring auxiliary discrimination using coupling synchronization degree. When the calculated agitation threshold is between a preset positive and a preset negative threshold, it indicates that the comprehensive discrimination information is insufficient to make a clear judgment, and the patient is currently in the boundary region between two agitation types. In cases where the calculated agitation threshold is in the boundary region, directly reverting to the core indicator of coupling synchronization degree ensures the physiological rationality of the discrimination result. The setting of the coupling determination threshold can refer to the dividing point of the coupling synchronization degree distribution of the two types of agitation in the training samples, for example, selecting the critical value that minimizes the sum of the misclassification rates of the two types of samples.

[0058] The noise discrimination calculation module includes: The coupling modulation coefficient calculation module is used to calculate the coupling modulation coefficient by means of the coupling synchronization degree through an autonomic nervous system-motor dual-channel dynamic coupling modulation mechanism. The coupling modulation coefficient The calculation formula is: .

[0059] in, Indicates the degree of coupling synchronization. This represents the preset coupling synchronization baseline value. This represents the mapping slope parameter. It is a time-varying coupling attenuation factor. This is a cross-modal consistency correction factor; The time-varying coupling attenuation factor The calculation formula is: .

[0060] in, The attenuation intensity coefficient, For reference coupling synchronization degree, This refers to the physiological rhythm cycle during the anesthesia recovery period. This is the initial phase; The cumulative time after the patient enters the anesthesia recovery period; the time-varying coupling attenuation factor is used to capture the periodic fluctuation characteristics of the autonomic nervous system during the anesthesia recovery period.

[0061] The cross-modal consistency correction factor The calculation formula is: .

[0062] in, The cosine similarity between the autonomic nervous system feature vector and the motor behavior feature vector is given. The upper limit of cosine similarity is defined as follows: The cross-modal consistency correction factor is used to enhance the modulation effect when the autonomic neural feature vector and the motor behavior feature vector exhibit directional consistency.

[0063] The noise discrimination calculation value calculation submodule is used to calculate the noise discrimination value based on the distance of the first prototype. Second prototype distance Calculate the prototype distance difference value The prototype distance difference value According to the coupling modulation coefficient Distance difference with prototype The noise discrimination value is calculated.

[0064] For example, the coupling synchronization baseline value is set to 0.5, and the mapping slope parameter is set to 8. First, the Sigmoid function part is calculated: the coupling synchronization degree 0.680 minus the baseline value 0.5 gives 0.180, multiply by the mapping slope 8 to get 1.44, take the negative value and calculate the natural exponent to get 0.237, add 1 to get 1.237, and take the reciprocal to get 0.808.

[0065] The cumulative time for the patient to enter the anesthesia recovery period was 23 minutes, equivalent to 1380 seconds. The attenuation intensity coefficient was set to 0.15, the reference coupling synchronization was set to 0.6, the physiological rhythm cycle of the anesthesia recovery period was set to 600 seconds, and the initial phase was set to 0. The exponential attenuation term was calculated as follows: the coupling synchronization of 0.680 divided by the reference value of 0.6 yielded 1.133, which, after taking the negative value, yielded a natural exponent of 0.322. The sine square term was calculated as follows: 1380 divided by 600 yielded 2.3, which, multiplied by twice pi, yielded 14.45. The sine of this sine was -0.588, which, when squared, yielded 0.346. The attenuation intensity of 0.15 multiplied by 0.322 multiplied by 0.346 yielded 0.017. Subtracting this value from 1 yielded a time-varying coupling attenuation factor of 0.983.

[0066] The cosine similarity between the autonomic neural feature vector and the motor behavior feature vector is calculated. The dot product of the two vectors is: 0.72 x 0.68 + 0.81 x 0.74 + 0.63 x 0.79 + 0.58 x 0.65 + 0.69 x 0.43, resulting in 0.490 + 0.599 + 0.498 + 0.377 + 0.297, which equals 2.261. The magnitude of the autonomic neural feature vector is: the sum of squares of its components 0.518 + 0.656 + 0.397 + 0.336 + 0.476 equals 2.383, and the square root is 1.544. The magnitude of the motor behavior feature vector is: the sum of squares of its components 0.462 + 0.548 + 0.624 + 0.423 + 0.185 equals 2.242, and the square root is 1.497. The cosine similarity is 2.261 divided by the product of 1.544 and 1.497, which is 2.311, resulting in 0.978.

[0067] The upper limit for cosine similarity is set to 1.0. The numerator is calculated as follows: 0.978 multiplied by the natural logarithm of 1.978 (0.682) equals 0.667; the hyperbolic tangent is then calculated to obtain 0.582. The denominator is calculated as follows: 1.0 multiplied by the natural logarithm of 2.0 (0.693) equals 0.693; the hyperbolic tangent is then calculated to obtain 0.601. The cross-modal consistency correction factor is 0.582 divided by 0.601, resulting in 0.968.

[0068] The final coupling modulation coefficient is 0.808 multiplied by 0.983 multiplied by 0.968, resulting in 0.769. Subtracting the second prototype distance of 0.047 from the first prototype distance of 0.342 yields a prototype distance difference of 0.295.

[0069] The noisiness discrimination calculation value calculation submodule includes: A nonlinear tensor coupling discriminant function is constructed based on the coupling modulation coefficients and the prototype distance difference value. The nonlinear tensor coupling discriminant function is as follows: .

[0070] in, This represents the calculated value for the agitation discrimination. Indicates the boundary compensation parameters; The noise discrimination calculation value calculation submodule also includes the following steps: Acquire historical agitation samples with a coupling synchronization degree lower than a preset coupling judgment threshold, and calculate the average discrimination distance by taking the mean of the absolute values ​​of the prototype distance differences of each historical agitation sample. ; Based on the average discrimination distance , Discrimination sensitivity adjustment coefficient and historical sample distribution dispersion Calculated boundary compensation parameters The boundary compensation parameters The calculation formula is: .

[0071] in, ∈(0,1], This is the dispersion weighting coefficient.

[0072] The adaptive boundary adjustment function is calculated using the following formula: .

[0073] in, For boundary expansion coefficient, This represents the upper limit of coupling synchronization. The reference distance parameter is used; the adaptive boundary adjustment function expands the discrimination boundary when the coupling synchronization degree is low, which is used to improve the robustness of noise type discrimination.

[0074] For example, 127 historical agitation samples with a coupling synchronization degree below the coupling judgment threshold of 0.45 were extracted from the historical sample database. The mean of the absolute values ​​of the prototype distance differences of each sample was calculated, resulting in an average discriminant distance of 0.183. The dispersion of the historical sample distribution was 0.067. The discriminant sensitivity adjustment coefficient was set to 0.8, and the dispersion weighting coefficient was set to 1.2. The discriminant sensitivity adjustment coefficient ranged from (0,1], and its value could be optimized through cross-validation on the validation dataset or set according to the specific clinical requirements for discriminant sensitivity and specificity. The terms in parentheses were calculated as follows: 0.067 divided by 0.183 yielded 0.366, multiplied by 1.2 yielded 0.439, added by 1 yielded 1.439, and the square root yielded 1.200. The boundary compensation parameter was 0.8 multiplied by 0.183 multiplied by 1.200, resulting in 0.176.

[0075] The boundary expansion coefficient is set to 0.3, the upper limit of coupling synchronization is set to 1.0, and the reference distance parameter is set to 0.25. The term within parentheses is calculated as follows: 1 minus the quotient of 0.680 divided by 1.0 (0.680) yields 0.320, squared to get 0.102, multiplied by 0.3 to get 0.031, and added to 1 to get 1.031. The prototype distance difference value of 0.295 is positive, so the sign function is set to 1. The exponent term is calculated as follows: 0.295 divided by 0.25 yields 1.180, which is negative, and the natural exponent is calculated to get 0.307. The adaptive boundary adjustment function value is 1 multiplied by 1.031 multiplied by 0.307, resulting in 0.317.

[0076] The system calculates the noise discrimination value based on the nonlinear tensor coupling discriminant function. The first term is the coupling modulation coefficient 0.769 multiplied by the prototype distance difference value 0.295, resulting in 0.227. The second term is 1 minus 0.769, resulting in 0.231, multiplied by the boundary compensation parameter 0.176, and multiplied by the adaptive boundary adjustment function value 0.317, resulting in 0.013. The noise discrimination value is the sum of the two terms, which equals 0.240.

[0077] The system's preset positive discrimination threshold is 0.15, and the negative discrimination threshold is -0.15. The current agitation calculation value of 0.240 is greater than the positive discrimination threshold of 0.15, so the system directly classifies the agitation type as painful agitation. The system then sends an alert to the monitoring workstation, indicating that the patient's current agitation is painful and recommending analgesia. Upon receiving the alert, the anesthesiologist can promptly assess the patient and provide appropriate analgesic intervention, such as intravenous administration of opioid analgesics or adjustment of the patient-controlled analgesia (PCA) pump parameters. For example, approximately 4 minutes after administration, the patient's agitation significantly decreased, the heart rate dropped to 82 beats per minute, and the blood pressure decreased to 124 mmHg and 76 mmHg, respectively. After 15 minutes of continued observation, the patient gradually regained consciousness, was able to cooperate with commands, and complained of pain at the incision site, which was consistent with the system's painful agitation classification result. When the agitation type is determined to be agitation during the anesthesia recovery period, the system will not output a pain warning. Medical staff can take sedation measures or continue to observe and wait for the anesthetic drugs to be metabolized naturally, depending on the clinical situation, to avoid the risks of respiratory depression and delayed awakening caused by unnecessary use of analgesics.

[0078] Finally, it should be noted that although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A postoperative pain risk prediction system for patients, characterized in that, The system includes: a standardized physiological signal sequence generation module, a coupling synchronization calculation module, and agitation type discrimination analysis and early warning module, with each module communicating with the others in sequence. A standardized physiological signal sequence generation module is used to collect multimodal physiological signals from patients in the anesthesia recovery period and preprocess the multimodal physiological signals to obtain a standardized physiological signal sequence. The coupling synchronization degree calculation module is used to extract an agitation feature set based on the standardized physiological signal sequence, separate an autonomic neural feature subset and a motor behavior feature subset from the agitation feature set; construct an autonomic neural feature vector after normalizing each feature value in the autonomic neural feature subset, construct a motor behavior feature vector after normalizing each feature value in the motor behavior feature subset; and calculate the coupling synchronization degree between the autonomic neural feature vector and the motor behavior feature vector. The agitation type discrimination analysis and early warning module is used to obtain agitation type discrimination results by analyzing the coupling synchronization degree. The agitation type discrimination results include agitation during anesthesia recovery and painful agitation. When the agitation type discrimination result is painful agitation, an early warning message is output.

2. The postoperative pain risk prediction system for patients according to claim 1, characterized in that, The coupling synchronization calculation module includes the following steps: The time-frequency domain transformations of the autonomic neural feature vector and the motor behavior feature vector are performed respectively to obtain the time-frequency feature matrix of the autonomic neural system and the time-frequency feature matrix of the motor behavior. The phase-locking values ​​of the autonomic nervous system time-frequency feature matrix and the motor behavior time-frequency feature matrix in each frequency band are calculated, and the phase-locking values ​​in each frequency band are weighted and fused to obtain the cross-modal phase synchronization index. The mutual information value is obtained by calculating the mutual information between the autonomic neural feature vector and the motor behavior feature vector. ; According to the cross-modal phase synchronization index and feature mutual information value Calculate the coupling synchronization degree The coupling synchronization degree The calculation formula is: ； in, and These are the cross-modal phase synchronization index and the fusion index of feature mutual information values, respectively. This is the coupling suppression coefficient; This is the coupling scale parameter, used to normalize the exponential term.

3. The postoperative pain risk prediction system for patients according to claim 2, characterized in that, The disturbance type discrimination analysis and early warning module includes: The agitation discrimination calculation module is used to construct the prototype vector of agitation during anesthesia recovery and the prototype vector of painful agitation; the autonomic nerve feature vector and the motor behavior feature vector are concatenated to obtain the current agitation feature vector; the feature space distance between the current agitation feature vector and the prototype vector of agitation during anesthesia recovery is calculated to obtain the first prototype distance; the feature space distance between the current agitation feature vector and the prototype vector of painful agitation is calculated to obtain the second prototype distance. The agitation discrimination calculation value is calculated based on the coupling synchronization degree, the first prototype distance, and the second prototype distance, and the agitation type discrimination result is determined based on the agitation discrimination calculation value.

4. The postoperative pain risk prediction system for patients according to claim 3, characterized in that, The agitation discrimination calculation module includes: The coupling modulation coefficient calculation module is used to calculate the coupling modulation coefficient by means of the coupling synchronization degree through an autonomic nervous system-motor dual-channel dynamic coupling modulation mechanism. The coupling modulation coefficient The calculation formula is: ； in, Indicates the degree of coupling synchronization. This represents the preset coupling synchronization baseline value. This represents the mapping slope parameter. The time-varying coupling attenuation factor, This is a cross-modal consistency correction factor; The noise discrimination calculation value calculation submodule is used to calculate the noise discrimination value based on the distance of the first prototype. Second prototype distance Calculate the prototype distance difference value The prototype distance difference value According to the coupling modulation coefficient Distance difference with prototype The noise discrimination value is calculated.

5. The postoperative pain risk prediction system for patients according to claim 4, characterized in that, The coupling modulation coefficient calculation module includes the following steps: The time-varying coupling attenuation factor The calculation formula is: ； in, The attenuation intensity coefficient, For reference coupling synchronization degree, This refers to the physiological rhythm cycle during the anesthesia recovery period. This is the initial phase; The cumulative time after the patient enters the anesthesia recovery period; the time-varying coupling attenuation factor is used to capture the periodic fluctuation characteristics of the autonomic nervous system during the anesthesia recovery period.

6. The postoperative pain risk prediction system for patients according to claim 4, characterized in that, The coupling modulation coefficient calculation module further includes the following steps: The cross-modal consistency correction factor The calculation formula is: ； in, The cosine similarity between the autonomic nervous system feature vector and the motor behavior feature vector is given. The upper limit of cosine similarity is defined as follows: The cross-modal consistency correction factor is used to enhance the modulation effect when the autonomic neural feature vector and the motor behavior feature vector exhibit directional consistency.

7. The postoperative pain risk prediction system for patients according to claim 4, characterized in that, The noisiness discrimination calculation value calculation submodule includes: A nonlinear tensor coupling discriminant function is constructed based on the coupling modulation coefficients and the prototype distance difference value. The nonlinear tensor coupling discriminant function is as follows: ； in, This represents the calculated value for the agitation discrimination. Indicates the boundary compensation parameters; The adaptive boundary adjustment function is calculated using the following formula: ； in, For boundary expansion coefficient, This represents the upper limit of coupling synchronization. The reference distance parameter is used; the adaptive boundary adjustment function expands the discrimination boundary when the coupling synchronization degree is low, which is used to improve the robustness of noise type discrimination.

8. The postoperative pain risk prediction system for patients according to claim 7, characterized in that, The noise discrimination calculation module also includes the following steps: When the calculated agitation value is greater than the preset positive discrimination threshold, the agitation type discrimination result is determined to be painful agitation; When the calculated agitation discrimination value is less than the preset negative discrimination threshold, the agitation type discrimination result is determined to be agitation during anesthesia recovery. When the agitation discrimination calculation value is between the preset positive discrimination threshold and the preset negative discrimination threshold, if the coupling synchronization degree is greater than the preset coupling judgment threshold, the agitation type discrimination result is determined to be painful agitation; otherwise, the agitation type discrimination result is determined to be agitation during anesthesia recovery.

9. A postoperative pain risk prediction system for patients according to claim 7, characterized in that, The noise discrimination calculation value calculation submodule also includes the following steps: Acquire historical agitation samples with a coupling synchronization degree lower than a preset coupling judgment threshold, and calculate the average discrimination distance by taking the mean of the absolute values ​​of the prototype distance differences of each historical agitation sample. ; Based on the average discrimination distance , Discrimination sensitivity adjustment coefficient and historical sample distribution dispersion Calculated boundary compensation parameters The boundary compensation parameters The calculation formula is: ； in, ∈(0,1], This is the dispersion weighting coefficient.

Images