Cancer pain treatment pain assessment method and system based on intelligent data processing

Abstract

The invention discloses a cancer pain treatment pain assessment method and system based on intelligent data processing, and relates to the field of treatment assessment, and the method comprises the steps: 1, obtaining basic pathological data, medication records and real-time vital signs of a patient in real time through a hospital information system, extracting symptom features, and carrying out the calculation of the symptom features; identifying the self-contained pain text of the patient, extracting the pain event characteristics, and matching the symptom with the pain event characteristics to form a pain correlation factor matrix; 2, constructing a pain prediction model based on a time sequence convolutional network, and generating a multi-dimensional data acquisition trigger instruction; a mode of combining a time sequence convolutional network and a multi-head self-attention mechanism is adopted, a high-sensitivity acquisition window of a pain event is automatically identified, discontinuous event interference is inhibited through a time decay factor, and a high-weight time period is reserved by utilizing sparsification, so that a data acquisition time period with a high contribution value is automatically obtained; and the recognition precision of the key pain time period is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of treatment evaluation, and in particular to a cancer pain treatment pain evaluation method and system based on intelligent data processing. Background Art

[0002] Cancer pain is the sensation caused by the transmission of information that the painful area needs to be repaired or regulated to the nerve center. It is one of the main causes of pain in patients with advanced cancer. With the development of medical technology, cancer pain management has gradually developed in the direction of precision. Traditional empirical treatment methods can no longer meet the needs of patients. It is urgent to achieve personalized and dynamic pain assessment and intervention through data-driven means. Early assessment and intervention of cancer pain are crucial to the patient's quality of life. The pain perception and response of cancer patients vary from person to person and are affected by physiological, psychological, social and other factors. Different patients have significant differences in sensitivity, tolerance and expression of pain, which makes the assessment and management of cancer pain complicated. Cancer patients often face persistent, multi-dimensional pain, such as somatic pain and neuralgia. Traditional subjective scales rely on patients' immediate recall and are easily affected by emotions and cognitive biases, making it difficult to capture the dynamic changes of pain. Patients vary greatly in their tolerance and side effect responses to analgesics. Different patients have different sensitivity to pain in the early stages of cancer pain, which leads to incorrect assessments of patients' cancer pain symptoms, resulting in early or delayed intervention measures that do not match the ability to treat cancer pain, affecting the timely relief of cancer pain or excessive relief leading to premature addiction or increased treatment tolerance. There is a lack of definition of the most worthy data collection period, which leads to insufficient contribution of the pain assessment data obtained, resulting in errors in treatment measures. Summary of the invention

[0003] 1. Technical issues to be resolved In view of the above-mentioned shortcomings of the prior art, the present invention provides a cancer pain treatment pain assessment method and system based on intelligent data processing, which can effectively solve the problems of the prior art.

[0004] (II) Technical solution To achieve the above objectives, the present invention is implemented through the following technical solutions: The present invention discloses a cancer pain treatment pain assessment method based on intelligent data processing, comprising the following steps: Step 1: Obtain the patient's basic pathological data, medication records and real-time vital signs through the hospital information system in real time, extract the symptoms, identify the patient's self-reported pain text, extract the pain event characteristics, match the symptoms and pain event characteristics, and form a pain correlation factor matrix; Step 2: Build a pain prediction model based on a temporal convolutional network, input the pain correlation factor matrix, identify the key inducing time period of pain events, output the pain contribution coefficient of each time period, and automatically mark it as a high-sensitivity acquisition window when the coefficient exceeds the preset threshold, and synchronously generate a multi-dimensional data acquisition trigger instruction; Step 3: Based on the trigger instruction, facial expression, limb activity, breathing and vocalization parameters are collected in different marked time periods of the current cycle to build a comprehensive time series event library; Step 4: Based on the comprehensive time-series event database, calculate the coefficient of variation of each event indicator according to the preset period, adjust the weight distribution of events in different time periods, and output the dynamic pain index based on the integration of each event indicator and its real-time weight; Step 5: Match the treatment intervention measures in the preset strategy library based on the dynamic pain index of the current cycle; Step 6: According to the fluctuation status of the pain index of the current patient in different cycles, optimize the priority of the treatment intervention measures corresponding to the score and output the final intervention list.

[0005] Furthermore, the working logic of the pain prediction model in step 2 is: Inputting the pain-related factor matrix into a temporal convolutional network to extract the correlation features between symptoms and pain events at different temporal sequences; Adaptively assign weights to temporal correlation features, calculate the significance scores of pain event-inducing periods through a learnable parameter matrix, and screen out key periods that are highly correlated with pain intensity fluctuations; Map the time series correlation features to the 0-1 interval to generate the pain contribution coefficient for each time period; When the pain contribution coefficient exceeds the preset threshold, the high-sensitivity acquisition window marking is triggered, and a number of data acquisition instructions bound to the pain event type and the inducing period are generated. The data acquisition instructions control the frequency and duration of the collection of facial expressions and limb activity parameters.

[0006] Furthermore, the calculation of the saliency score adopts a multi-head self-attention mechanism, including: Map the temporal convolutional features into query vector, key vector and value vector respectively; The matching degree between the query vector and the key vector is calculated by cosine similarity, and the interference of non-continuous events is suppressed by combining the time decay factor; The attention weights are sparsely processed, and the first K high-weight periods are retained as key inducing periods, where the K value is dynamically adjusted according to the historical distribution of the patient's pain events.

[0007] Furthermore, the initial value of the preset threshold is set to 0.65, and feedback adjustment is performed based on the verification result of the true correlation between the actual collected data obtained in the marked window in the historical period and the pain index. When the correlation is lower than the preset confidence, the threshold is lowered according to the gradient descent strategy, and when the correlation is higher than the preset confidence, the threshold is increased.

[0008] Furthermore, the calculation formula of the dynamic pain index in step 4 is: ; In the formula, Represents the current period t The dynamic pain index, Represents the total number of preset indicators, including facial expressions, limb movements, breathing and vocalization parameters, Represents the first i The standardized measurement value of each indicator is mapped to the (0,1) interval through normalization. Representative i The indicators in the same period of history t The standard deviation of Representative i The indicators in the same period of history t The mean of Representative i Items in the period t The real-time weight.

[0009] Furthermore, the real-time weight The calculation formula is: ; In the formula, Represents the preset basic weight, Represents the preset sensitivity factor (0< ≤0.5), represents the coefficient of variation.

[0010] Furthermore, the optimization process of the priority of the treatment intervention measures in step 5 is: Primary screening based on the current dynamic pain index; Obtain historical efficacy data for each treatment intervention and calculate actual pain relief rate and risk index; Based on the actual pain relief rate and risk index, and adjusting the order of options according to the patient's current individual characteristics, a prioritized list of treatment options is generated.

[0011] A cancer pain treatment pain assessment system based on intelligent data processing, comprising: The data capture module is used to connect to the hospital information system in real time, integrate the patient's pathology, medication, vital signs and self-report text data, establish a pain association matrix through natural language processing technology and rule engine, and mark the spatiotemporal association characteristics of symptoms and pain; The time period labeling module is used to dynamically capture the key pain-inducing time periods based on the temporal convolutional network and multi-head self-attention mechanism, output the pain contribution coefficient of each time period, trigger the high-sensitivity window labeling when exceeding the threshold, and generate several data collection instructions; The data acquisition module is used to collect the patient's facial micro-expressions, limb movements, breathing rhythm and vocalization parameters according to the data acquisition instructions, and store them in the comprehensive database after standardization; The pain calculation module is used to combine the coefficient of variation of each indicator with the time period weight distribution, integrate multi-source data and calculate the dynamic pain index through the gradient boosting tree; The strategy matching module is used to match the preset treatment strategy library according to the dynamic pain index, screen candidate intervention measures that meet the individual characteristics of the patient, and output a preliminary treatment plan; The strategy optimization module is used to build a multi-objective optimization model based on historical efficacy and risk data, calculate the pain relief rate and medication risk, and generate a final intervention list based on the calculation results.

[0012] Furthermore, the strategy optimization module is interactively connected to a data calibration module via a wireless network. The data calibration module verifies the correlation between the collected data and the dynamic pain index offline, dynamically adjusts the high-sensitivity window judgment threshold, and optimizes the threshold through a gradient descent algorithm.

[0013] Furthermore, the data capture module is interactively connected to the time period marking module through a wireless network, the time period marking module is interactively connected to the data acquisition module through a wireless network, the data acquisition module is interactively connected to the pain calculation module through a wireless network, the pain calculation module is interactively connected to the strategy matching module through a wireless network, and the strategy matching module is interactively connected to the strategy optimization module through a wireless network.

[0014] (III) Beneficial effects Compared with the known prior art, the technical solution provided by the present invention has the following beneficial effects: By combining a temporal convolutional network with a multi-head self-attention mechanism, the highly sensitive collection windows of pain events are automatically identified, the interference of non-continuous events is suppressed by the time attenuation factor, and high-weight time periods are retained using sparsification, thereby automatically obtaining data collection periods with high contribution value. This solves the problem that traditional methods are difficult to capture the dynamic changes of pain, and significantly improves the recognition accuracy of key pain periods.

[0015] By automatically collecting facial expressions, limb movements, breathing and vocalization parameters when the high-sensitivity window is triggered, a comprehensive time-series event library is constructed. By calculating the coefficient of variation of each indicator and dynamically adjusting the weight distribution, the bias impact of a single physiological indicator or subjective report is effectively reduced, thereby enhancing the objectivity of the evaluation.

[0016] By generating personalized treatment options based on historical efficacy data and real-time patient characteristics through a dynamic priority sorting algorithm, high-risk patients can automatically avoid highly addictive regimens and medication strategies can be adjusted in advance for those with rapid tolerance, thereby helping patients cope with cancer pain throughout the entire cycle. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 A schematic diagram of a process of pain assessment method for cancer pain treatment based on intelligent data processing in the present invention; Figure 2 This is a schematic diagram of the framework of a cancer pain treatment pain assessment system based on intelligent data processing in the present invention.

[0019] The numbers in the figure represent: 1. Data capture module; 2. Time period labeling module; 3. Data collection module; 4. Pain calculation module; 5. Strategy matching module; 6. Strategy optimization module; 7. Data calibration module. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0021] The present invention will be further described below in conjunction with the embodiments.

[0022] ① Embodiment 1: This embodiment is a cancer pain treatment pain assessment method based on intelligent data processing, such as Figure 1 As shown, the following steps are included: Step 1: Obtain the patient's basic pathological data, medication records and real-time vital signs through the hospital information system in real time, extract the symptoms, identify the patient's self-reported pain text, extract the pain event characteristics, match the symptoms with the pain event characteristics, and form a pain-related factor matrix; basic pathological data include tumor stage, metastasis site and pain history, medication records include opioid use cycle and dosage, real-time vital signs include heart rate and blood pressure; the pain-related factor matrix mainly includes the following data: patient basic pathological data: tumor stage, metastasis site, pain history; medication records: opioid use cycle and dosage; real-time vital signs: heart rate, blood pressure and blood oxygen; patient self-reported pain text and pain event characteristics; Step 2: Construct a pain prediction model based on a temporal convolutional network. The construction process of the pain prediction model is as follows: based on the historical patient pain correlation factor matrix and the corresponding annotated pain event time period data, use the temporal convolutional network for training, input the pain correlation factor matrix into the temporal convolutional network, extract the temporal correlation features through the dilated causal convolution layer, use the residual connection to optimize the gradient propagation, and use the Sigmoid function in the output layer to generate the pain contribution coefficient of each time period; Input the pain correlation factor matrix into the trained pain prediction model, identify the key inducing time period of pain events, and output the pain contribution coefficient of each time period. The pain contribution coefficient is in the range of 0-1. When the coefficient exceeds the preset threshold of 0.65, it is automatically marked as a high-sensitivity acquisition window, and a multi-dimensional data acquisition trigger instruction is generated synchronously; The working logic of the pain prediction model is: The pain-related factor matrix is ​​input into the temporal convolutional network to extract the correlation features between symptoms and pain events at different temporal sequences. Adaptively assign weights to temporal correlation features, and calculate the significance score of the pain event induction period through a learnable parameter matrix. The learnable parameter matrix is: a pain correlation factor matrix of one or more pain events; screen out key periods that are highly correlated with pain intensity fluctuations; calculate the significance score using a multi-head self-attention mechanism, including: mapping the temporal convolution features of the original output of the temporal convolutional network into query vectors, key vectors, and value vectors; calculate the matching degree of the query vector and the key vector through cosine similarity, and combine the time decay factor to suppress the interference of non-continuous events; perform sparse processing on the attention weights, retain the first K high-weight periods as key induction periods, where the K value is dynamically adjusted according to the historical distribution of the patient's pain events; The features extracted by the temporal convolutional network are input into the fully connected layer, and then the activation function is applied to generate coefficients, mapping the temporal correlation features to the 0-1 interval to generate the pain contribution coefficient for each period; When the pain contribution coefficient exceeds the preset threshold, the high-sensitivity collection window annotation is triggered, and a number of data collection instructions bound to the pain event type and the inducing period are generated. The data collection instructions control the frequency and duration of the collection of facial expressions and limb activity parameters; the initial value of the preset threshold is set to 0.65, and feedback adjustment is performed based on the verification results of the actual correlation between the actual collected data obtained in the annotation window and the pain index in the historical period. When the correlation is lower than the preset confidence, the threshold is lowered according to the gradient descent strategy, and when the correlation is higher than the preset confidence, the threshold is increased; Step 3: Based on the trigger instruction, facial expression, limb activity, breathing and vocalization parameters are collected in different marked time periods of the current cycle to build a comprehensive time series event library; Step 4: Based on the comprehensive time-series event database, calculate the coefficient of variation of each event indicator according to the preset period, adjust the weight distribution of events in different time periods, and output the dynamic pain index based on the integration of each event indicator and its real-time weight; Step 5: Match the treatment intervention measures in the preset strategy library based on the dynamic pain index of the current cycle; the optimization process of the treatment intervention priority is: Primary screening based on the current dynamic pain index; Obtain historical efficacy data of each therapeutic intervention measure, and calculate the actual pain relief rate and risk index; the calculation process is as follows: for each therapeutic intervention measure, statistically analyze the efficacy data of the target patient group in the same pain index range in historical applications, calculate the proportion of effective relief cases, and obtain the actual pain relief rate of a certain measure; perform weighted calculation based on the incidence rate and severity grade of adverse reactions of the measure in historical data, and the matching results of the individual contraindication characteristics of the current patient, and obtain the risk index of a certain measure; Generate a prioritized list of treatment options based on actual pain relief rate and risk index, and adjust the order of options according to the patient's current individual characteristics; Step 6: According to the fluctuation status of the pain index of the current patient in different cycles, optimize the priority of the treatment intervention measures corresponding to the score and output the final intervention list.

[0023] Compared with existing technologies, the hybrid architecture of the temporal convolutional network and the multi-head self-attention mechanism can identify the temporal correlation characteristics of pain-inducing factors, introduce time decay factors and dynamic sparsification processing, and improve the prediction accuracy in capturing nonlinear pain fluctuations compared with traditional ARIMA and other time series models, effectively solving the problem of irregular cancer pain attack cycles. A three-dimensional behavior observation framework including facial micro-expression recognition, limb movement analysis and speech acoustic parameters is constructed. The coefficient of variation of each item indicator is calculated according to the preset period based on the comprehensive time series item library, and the weight distribution is adjusted in combination with the preset item contribution of different time periods for fusion output. It can dynamically reflect the changes in the patient's pain state in different periods. Compared with the method of calculating the pain index with fixed weights or a single indicator in the existing technology, it can better fit the patient's actual pain situation and improve the accuracy and real-time performance of pain assessment; coefficient of variation of each item indicator: the facial expression intensity value, limb activity sequence and speech acoustic parameters collected by periodic time period are calculated for each item indicator in the historical same period. The ratio of the standard deviation to the mean is the coefficient of variation of the corresponding indicator, which quantifies the degree of discreteness of its fluctuation over time; In the process of matching and optimizing therapeutic interventions, a primary screening is first performed based on the current dynamic pain index, and then the historical efficacy data of each therapeutic intervention is obtained to calculate the actual pain relief rate and risk index. The scheme ranking is adjusted according to the patient's current individual characteristics, and a prioritized list of treatment options is generated. Finally, the priority of the therapeutic interventions corresponding to the score can be further optimized according to the fluctuation state of the patient's pain index in different cycles, and the final intervention list is output. This can provide patients with more personalized, accurate and effective treatment plans, which significantly improves the treatment effect and patient satisfaction compared to the fixed mode or single factor intervention selection method in the existing technology.

[0024] ② Example 2: In other aspects, this example also provides another optimization mechanism based on Example 1, specifically a cancer pain treatment pain assessment system based on intelligent data processing, such as Figure 2 As shown, including: Data capture module 1 is used to connect to the hospital information system in real time, integrate the patient's pathology, medication, vital signs and self-report text data, establish a pain association matrix through natural language processing technology and rule engine, and mark the spatiotemporal association characteristics of symptoms and pain; The time period labeling module 2 is used to dynamically capture the key time periods of pain induction based on the temporal convolutional network and multi-head self-attention mechanism, output the pain contribution coefficient of each time period, trigger the high-sensitivity window labeling when exceeding the threshold, and generate several data collection instructions; Data collection module 3, used to collect the patient's facial micro-expressions, limb movements, breathing rhythm and vocalization parameters according to the data collection instructions, and store them in the comprehensive database after standardization; The pain calculation module 4 is used to combine the coefficient of variation of each indicator with the time period weight distribution, fuse multi-source data and calculate the dynamic pain index through the gradient boosting tree; Strategy matching module 5, used to match the preset treatment strategy library according to the dynamic pain index, screen candidate intervention measures that meet the individual characteristics of the patient, and output a preliminary treatment plan; Strategy optimization module 6, used to build a multi-objective optimization model based on historical efficacy and risk data, calculate pain relief rate and medication risk, and generate a final intervention list based on the calculation results; The strategy optimization module 6 is interactively connected to the data calibration module 7 via a wireless network. The data calibration module 7 verifies the correlation between the collected data and the dynamic pain index offline, dynamically adjusts the high-sensitivity window determination threshold, and optimizes the threshold through a gradient descent algorithm. The data capture module 1 is interactively connected to the time period marking module 2 through a wireless network, the time period marking module 2 is interactively connected to the data collection module 3 through a wireless network, the data collection module 3 is interactively connected to the pain calculation module 4 through a wireless network, the pain calculation module 4 is interactively connected to the strategy matching module 5 through a wireless network, and the strategy matching module 5 is interactively connected to the strategy optimization module 6 through a wireless network.

[0025] In the specific implementation of this embodiment, the data capture module 1 is connected to the hospital information system in real time, integrating the patient's pathology, medication, vital signs and self-report text data, and establishing a pain association matrix and marking the spatiotemporal association characteristics of symptoms and pain through natural language processing technology and rule engine. The time period labeling module 2 is based on the temporal convolutional network and multi-head self-attention mechanism to dynamically capture the key time periods of pain induction, output the pain contribution coefficient of each time period, trigger the high-sensitivity window labeling when exceeding the threshold, and generate data collection instructions to the data collection module 3. The data collection module 3 collects the patient's facial micro-expressions, limb activities, respiratory rhythm and voice parameters according to the instructions, and stores them in the comprehensive database after standardization; The pain calculation module 4 combines the coefficient of variation of each indicator with the time period weight distribution, integrates multi-source data and calculates the dynamic pain index through gradient boosting. The strategy matching module 5 matches the preset treatment strategy library according to the dynamic pain index, screens the candidate intervention measures that meet the individual characteristics of the patient and outputs a preliminary treatment plan. The strategy optimization module 6 constructs a multi-objective optimization model based on historical efficacy and risk data, calculates the pain relief rate and medication risk, and sorts to generate the final intervention list. At the same time, the data calibration module 7 verifies the correlation between the collected data and the dynamic pain index offline, dynamically adjusts the high-sensitivity window judgment threshold, and optimizes the threshold through the gradient descent algorithm.

[0026] The calculation formula of the dynamic pain index in the above embodiment is specifically: ; In the formula, Represents the current period t The dynamic pain index, Represents the total number of preset indicators, including facial expressions, limb movements, breathing and vocalization parameters. Representative period t Neidi iThe standardized measurement value of each indicator is mapped to the interval of 0 and 1 through normalization. Representative i The indicators in the same period of history t The standard deviation of Representative i The indicators in the same period of history t The mean of Representative i Items in the period t The real-time weight of The calculation formula is: ; In the formula, Represents the preset basic weight, Represents the preset sensitivity factor, 0< ≤0.5, represents the coefficient of variation.

[0027] The mean and standard deviation of each indicator are calculated through historical data, and the real-time measurement values ​​are Z-score standardized to eliminate dimensional differences. The coefficient of variation reflects the volatility of the indicator. A high coefficient of variation triggers an increase in weight. For example, when a patient's respiratory parameters fluctuate significantly at night, their weight is increased. The basic weight is preset by clinical experience, and the sensitivity factor controls the weight adjustment range to avoid excessive deviation. It is usually 0.2-0.3. The final weight is dynamically adjusted with the coefficient of variation to enhance the contribution of high volatility indicators. By integrating multimodal data through weighted summation, the dynamic pain index can simultaneously reflect the changes in physiological parameters and time-dependent characteristics. The output results are used to match the treatment strategy library to achieve personalized intervention.

[0028] In summary, the present invention comprehensively integrates the patient's basic pathological data, medication records, real-time vital signs and other information to construct a pain-related factor matrix, laying the foundation for accurate evaluation. Through the pain prediction model constructed based on the temporal convolutional network and the multi-head self-attention mechanism, the key time period of pain induction can be dynamically captured and the pain contribution coefficient can be output. The key time period can be accurately identified and the high-sensitivity acquisition window and trigger instructions can be set and adjusted through a reasonable mechanism, thereby improving the pertinence and effectiveness of data collection. In addition, a comprehensive time-series event library is built based on the trigger instructions, and a dynamic pain index is output by combining the coefficient of variation of event indicators and the time period weight distribution, so that the pain assessment is more in line with the actual situation of the patient. When matching treatment intervention measures, not only the dynamic pain index is considered for primary screening, but also the historical efficacy data and individual characteristics of the patient are combined to optimize the priority and generate a prioritized list of treatment options. Finally, the priority can be further optimized according to the fluctuation status of the patient's pain index to output the final intervention list. The entire process realizes the intelligence and precision of the entire chain from data collection to evaluation to matching of intervention measures, which significantly improves the scientificity and effectiveness of cancer pain treatment.

[0029] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A cancer pain treatment pain assessment method based on intelligent data processing, characterized in that: The following steps are involved: Step 1: Obtain the patient's basic pathological data, medication records and real-time vital signs through the hospital information system in real time, extract the symptoms, identify the patient's self-reported pain text, extract the pain event characteristics, match the symptoms and pain event characteristics, and form a pain correlation factor matrix; Step 2: Build a pain prediction model based on a temporal convolutional network, input the pain correlation factor matrix, identify the key inducing time period of pain events, output the pain contribution coefficient of each time period, and automatically mark it as a high-sensitivity acquisition window when the coefficient exceeds the preset threshold, and synchronously generate a multi-dimensional data acquisition trigger instruction; Step 3: Based on the trigger instruction, facial expression, limb activity, breathing and vocalization parameters are collected in different marked time periods of the current cycle to build a comprehensive time series event library; Step 4: Based on the comprehensive time-series event database, calculate the coefficient of variation of each event indicator according to the preset period, adjust the weight distribution of events in different time periods, and output the dynamic pain index based on the integration of each event indicator and its real-time weight; Step 5: Match the treatment intervention measures in the preset strategy library based on the dynamic pain index of the current cycle; Step 6: According to the fluctuation status of the pain index of the current patient in different cycles, optimize the priority of the treatment intervention measures corresponding to the score and output the final intervention list.

2. The cancer pain treatment pain assessment method based on intelligent data processing according to claim 1, characterized in that: The working logic of the pain prediction model in step 2 is: Inputting the pain-related factor matrix into a temporal convolutional network to extract the correlation features between symptoms and pain events at different temporal sequences; Adaptively assign weights to temporal correlation features, calculate the significance scores of pain event-inducing periods through a learnable parameter matrix, and screen out key periods that are highly correlated with pain intensity fluctuations; Map the time series correlation features to the 0-1 interval to generate the pain contribution coefficient for each time period; When the pain contribution coefficient exceeds the preset threshold, the high-sensitivity acquisition window marking is triggered, and a number of data acquisition instructions bound to the pain event type and the inducing period are generated. The data acquisition instructions control the frequency and duration of the collection of facial expressions and limb activity parameters.

3. The cancer pain treatment pain assessment method based on intelligent data processing according to claim 2, characterized in that: The calculation of the saliency score adopts a multi-head self-attention mechanism, including: Map the temporal convolutional features into query vector, key vector and value vector respectively; The matching degree between the query vector and the key vector is calculated by cosine similarity, and the interference of non-continuous events is suppressed by combining the time decay factor; The attention weights are sparsely processed, and the first K high-weight periods are retained as key inducing periods, where the K value is dynamically adjusted according to the historical distribution of the patient's pain events.

4. The cancer pain treatment pain assessment method based on intelligent data processing according to claim 2, characterized in that: The initial value of the preset threshold is set to 0.65, and feedback adjustment is performed based on the verification result of the true correlation between the actual collected data obtained in the marked window in the historical period and the pain index. When the correlation is lower than the preset confidence, the threshold is lowered according to the gradient descent strategy. When the correlation is higher than the preset confidence, the threshold is increased.

5. The cancer pain treatment pain assessment method based on intelligent data processing according to claim 1, characterized in that: The calculation formula of the dynamic pain index in step 4 is: ； In the formula, Represents the current period t The dynamic pain index, Represents the total number of preset indicators, including facial expressions, limb movements, breathing and vocalization parameters, Represents the first i The standardized measurement value of each indicator is mapped to the (0,1) interval through normalization. Representative i The indicators in the same period of history t The standard deviation of Representative i The indicators in the same period of history t The mean of Representative i Items in the period t The real-time weight.

6. The cancer pain treatment pain assessment method based on intelligent data processing according to claim 5, characterized in that: The real-time weight The calculation formula is: ； In the formula, Represents the preset basic weight, represents the preset sensitivity factor, represents the coefficient of variation.

7. The cancer pain treatment pain assessment method based on intelligent data processing according to claim 1, characterized in that: The optimization process of the priority of the treatment intervention measures in step 5 is: Primary screening based on the current dynamic pain index; Obtain historical efficacy data for each treatment intervention and calculate actual pain relief rate and risk index; Based on the actual pain relief rate and risk index, and adjusting the order of options according to the patient's current individual characteristics, a prioritized list of treatment options is generated.

8. A cancer pain treatment pain assessment system based on intelligent data processing, the system is an implementation system of a cancer pain treatment pain assessment method based on intelligent data processing according to any one of claims 1 to 7, characterized in that: include: The data capture module (1) is used to connect to the hospital information system in real time, integrate the patient's pathology, medication, vital signs and self-report text data, establish a pain association matrix through natural language processing technology and rule engine, and mark the spatiotemporal association characteristics of symptoms and pain; The time period labeling module (2) is used to dynamically capture the key time periods of pain induction based on a temporal convolutional network and a multi-head self-attention mechanism, output the pain contribution coefficient of each time period, trigger high-sensitivity window labeling when exceeding the threshold, and generate a number of data collection instructions; A data collection module (3) is used to collect the patient's facial micro-expressions, limb movements, breathing rhythm and vocalization parameters according to the data collection instructions, and store them in a comprehensive database after standardization; The pain calculation module (4) is used to combine the coefficient of variation of each indicator with the time period weight distribution, fuse multi-source data and calculate the dynamic pain index through the gradient boosting tree; A strategy matching module (5) is used to match a preset treatment strategy library according to the dynamic pain index, screen candidate intervention measures that meet the individual characteristics of the patient, and output a preliminary treatment plan; The strategy optimization module (6) is used to build a multi-objective optimization model based on historical efficacy and risk data, calculate the pain relief rate and medication risk, and generate a final intervention list based on the calculation results.

9. The cancer pain treatment pain assessment system based on intelligent data processing according to claim 8, characterized in that: The strategy optimization module (6) is interactively connected to the data calibration module (7) via a wireless network. The data calibration module (7) dynamically adjusts the high-sensitivity window determination threshold by offline verification of the correlation between the collected data and the dynamic pain index, and optimizes the threshold by a gradient descent algorithm.

10. The cancer pain treatment pain assessment system based on intelligent data processing according to claim 8, characterized in that: The data capture module (1) is interactively connected to the time period marking module (2) via a wireless network, the time period marking module (2) is interactively connected to the data acquisition module (3) via a wireless network, the data acquisition module (3) is interactively connected to the pain calculation module (4) via a wireless network, the pain calculation module (4) is interactively connected to the strategy matching module (5) via a wireless network, and the strategy matching module (5) is interactively connected to the strategy optimization module (6) via a wireless network.

Images