Tumor radiotherapy rehabilitation nursing intervention system based on artificial intelligence

Abstract

The invention discloses a tumor radiotherapy rehabilitation nursing intervention system based on artificial intelligence, and relates to the technical field of tumor rehabilitation nursing. By integrating multi-dimensional data and combining clinical standards to output quantitative evaluation levels and key weak dimensions, rehabilitation weakness of a patient is accurately positioned, a targeted target is provided for the nursing intervention strategy generation module, a personalized scheme is generated, and pertinence and performability of intervention are ensured; policy execution data are captured in real time and a scheme is dynamically adjusted, so that the intervention process always adapts to the rehabilitation state change of a patient, and meanwhile, active management and control on high-risk groups are enhanced by periodically collecting core indexes and linking risk upgrading and priority intervention, so that the tumor radiotherapy rehabilitation nursing is changed from passive response to active pre-judgment, and the rehabilitation nursing efficiency is improved. And the overall quality of rehabilitation nursing after radiotherapy and the prognosis effect of a patient are optimized.

Description

Technical Field

[0001] This invention relates to the field of tumor rehabilitation nursing technology, and in particular to an artificial intelligence-based tumor radiotherapy rehabilitation nursing intervention system. Background Technology

[0002] In clinical practice of tumor radiotherapy, existing rehabilitation nursing relies heavily on the experience of medical staff for assessment. This assessment is often one-dimensional and fails to comprehensively capture subtle changes in brain structure during rehabilitation. Furthermore, intervention plans are often generic templates. Chinese patent CN113643795A discloses an artificial intelligence-based rehabilitation nursing system for oral and maxillofacial head and neck tumors. This system includes a patient smart terminal, an AI processing system, and a hospital nursing smart terminal, all interconnected. The patient smart terminal integrates a touch input display module, a data acquisition module, an upload module, and an alarm communication module. The touch input display module is used to input and store basic patient information, medical records, and assessment questionnaire information. The data acquisition module collects images, videos, and audio information during the patient's recovery process. The upload module, connected to the touch input display module and the data acquisition module, uploads the stored patient information, medical records, assessment questionnaire information, and images, videos, and audio information to the AI ​​processing system. This invention can automatically identify the patient's training achievement level, assisting in further rehabilitation decisions.

[0003] While the aforementioned patents can achieve the collection of rehabilitation data and automatic identification of training achievement levels in patients with oral and maxillofacial head and neck tumors, the following problems still exist: 1. It is unable to capture subtle rehabilitation changes after tumor radiotherapy, making it difficult to meet the specific rehabilitation needs of radiotherapy patients. It also fails to adjust the weight of indicators based on clinical priorities such as radiotherapy target organ sensitivity and complication risk level, making it impossible to accurately identify key weak dimensions of rehabilitation. Without establishing a dynamic adjustment mechanism and risk warning mechanism based on real-time data, it is impossible to identify potential risks that may be induced after radiotherapy in advance. It only relies on general rehabilitation training standards, and there is still a problem of delayed intervention. It is impossible to provide radiotherapy patients with precise intervention support that is tailored to their rehabilitation characteristics. Summary of the Invention

[0004] The purpose of this invention is to provide an artificial intelligence-based tumor radiotherapy rehabilitation nursing intervention system. By integrating multi-dimensional data and standardizing its processing, the system accurately outputs rehabilitation levels and key weak dimensions, generates personalized nursing strategies, monitors and dynamically adjusts them in real time, and achieves precise management of the entire process of tumor radiotherapy rehabilitation nursing, meeting the full-cycle rehabilitation needs of patients and solving the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: An AI-based intervention system for cancer radiotherapy rehabilitation and nursing care includes: The user profile acquisition module is used to acquire the personal profiles of cancer radiotherapy patients and extract the corresponding rehabilitation data from the personal profiles. The radiotherapy rehabilitation status assessment module is used to quantitatively assess the rehabilitation status of patients after radiotherapy based on the acquired rehabilitation data and in combination with preset clinical rehabilitation standards, and output the rehabilitation status assessment results, including rehabilitation assessment level and key weak dimensions of rehabilitation. The nursing intervention strategy generation module is used to generate nursing intervention strategies for the patient based on the rehabilitation status assessment results, combined with intervention optimization goals and constraints. The intervention effect feedback module is used to monitor the actual implementation effect of nursing intervention strategies in real time, obtain real-time feedback data during the implementation of nursing intervention strategies, evaluate the intervention effectiveness of nursing intervention strategies, and dynamically adjust nursing intervention strategies based on the intervention effectiveness evaluation results.

[0006] Furthermore, the process by which the user profile acquisition module extracts the rehabilitation data corresponding to the patient includes: The extracted rehabilitation data is read, the data type of the rehabilitation data is determined, and the corresponding similar clinical samples are matched in the preset rehabilitation case database according to the data type of the rehabilitation data. Read the similar clinical samples, determine the distribution characteristics of the rehabilitation data of the similar clinical samples, and generate data verification constraints based on the distribution characteristics and data types of the rehabilitation data; The extracted rehabilitation data is validated based on constraints, and data that does not meet the constraints is marked based on the validation results. The reasons for the abnormality of the data to be verified are analyzed. Data features of the verified rehabilitation data are extracted and weighted based on a preset clinical priority system, wherein the clinical priority system is set according to key influencing factors of tumor radiotherapy. The weighted feature data is integrated based on the time dimension to construct a three-dimensional rehabilitation data matrix. Redundancy processing is then performed on the three-dimensional rehabilitation data matrix to output standardized rehabilitation data, which serves as input data for the radiotherapy rehabilitation status assessment module.

[0007] Furthermore, the radiotherapy rehabilitation status assessment module includes: The rehabilitation status assessment model construction unit is used to construct a rehabilitation status assessment model based on preset clinical rehabilitation standards and combined with target data features in standardized rehabilitation data. The rehabilitation status assessment unit is used to input standardized rehabilitation data into the rehabilitation status assessment model, obtain the patient's comprehensive rehabilitation score, and map the rehabilitation assessment level based on a preset score range. The weak dimension identification unit is used to extract the influence weight of each data feature in the standardized rehabilitation data on the comprehensive rehabilitation score, determine the contribution of each dimension feature, and select the dimension feature with the greatest impact on the rehabilitation level as the key weak dimension of rehabilitation. The risk warning unit is used to mark patients whose rehabilitation assessment level is lower than the preset level or whose characteristic indicators in a certain dimension exceed the preset clinical safety range as high-risk and trigger the priority intervention mechanism.

[0008] Furthermore, the specific process for identifying key weaknesses in rehabilitation includes: The target data features in standardized rehabilitation data are classified according to clinical dimensions, namely physiological indicators, symptom control, psychological state, and daily living ability. Calculate the influence weight of each target data feature on the comprehensive rehabilitation score under each dimension, and adjust the influence weight of each target data feature in conjunction with the clinical priority system; The corrected impact weights are arranged from largest to smallest. Target data features are selected based on the impact weight queue, and the clinical dimensions corresponding to target data features that exceed the preset clinical reference range are taken as key weak dimensions for rehabilitation. Retrieve rehabilitation data of similar patients in the preset rehabilitation case database whose radiotherapy protocol similarity is greater than a preset similarity threshold, compare the patient's weakness dimension with the difference of similar patients, and if the difference is greater than a preset difference threshold, then confirm that the rehabilitation key weakness dimension is the final rehabilitation key weakness dimension.

[0009] Furthermore, the adjustment of the influence weights of each target data feature also includes: The marginal contribution of each target feature in the standardized rehabilitation data to the comprehensive rehabilitation score in the rehabilitation status assessment model is quantified, and the impact of changes in the feature values ​​of the target features on the score results is determined. Multiple assessment samples from the same patient are aggregated and the mean is calculated as the original influence weight of the target feature, while the direction of influence of the target feature is marked. The sensitivity coefficient of the target organ at the radiotherapy site, the correlation with underlying diseases, and the risk level of complications are obtained according to the clinical priority system, and the priority coefficient is obtained based on the sensitivity coefficient of the target organ at the radiotherapy site, the correlation with underlying diseases, and the risk level of complications. Based on the priority coefficient and the original influence weight of the target feature, and combined with the direction calibration coefficient corresponding to the influence direction of the target feature, the original influence weight is adjusted to obtain the corrected influence weight.

[0010] Furthermore, the priority intervention mechanism also includes: risk classification for patients marked as high-risk, integrating the patient's corresponding standardized rehabilitation data and key weak dimensions data according to the risk level, and simultaneously marking the priority of the patient's personal file processing status.

[0011] Furthermore, the nursing intervention strategy generation module includes: The objective function construction unit is used to construct a multi-objective optimization objective function based on the key weak dimensions and risk levels of rehabilitation, combined with a preset clinical priority system, and to determine the constraints of the multi-objective optimization objective function. The optimization strategy generation unit is used to retrieve effective intervention plans from rehabilitation data of similar patients, input each plan into the objective function of multi-objective optimization, calculate the fitness of the plan, and verify whether the plan meets the constraints of the objective function. The objective function is iteratively optimized based on multi-objective optimization, and its clinical feasibility is verified. Based on the verification results, the plan with the highest fitness is selected, the optimal nursing intervention strategy is generated, and it is synchronously linked to the patient's personal file, and the execution status of the nursing intervention strategy is marked.

[0012] Furthermore, the intervention effect feedback module also includes acquiring core characteristic indicator data of patients marked as high-risk at preset collection time intervals, comparing the core characteristic indicator data with a preset improvement range, and if the core characteristic indicators collected for a preset number of consecutive times are not lower than the preset improvement range, then the risk level of the patient is increased and the priority intervention mechanism is re-triggered until the core characteristic indicators return to the preset improvement range.

[0013] Furthermore, the AI-based tumor radiotherapy rehabilitation nursing intervention system also includes a dynamic dose-response adaptive adjustment module, used to calculate the cumulative bioequivalent dose of radiotherapy in real time and dynamically calibrate the intervention intensity threshold during the execution of the nursing intervention strategy, performing the following steps: The core characteristic indicators collected in real time from the intervention effect feedback module include the radiotherapy fractionated dose record D, fractionated irradiation time interval ΔT, target volume change rate RV, oxygen enhancement ratio OER, and cell survival curve αβ value. Specifically, the fractionated dose record D is obtained from the dosimetric report output by the radiotherapy equipment, in gray (Gy); the fractionated irradiation time interval ΔT is calculated from the schedule records of the radiotherapy information management system, in hours (h); the target volume change rate RV is calculated by comparing the target volume contours at the current and initial times, and is a dimensionless parameter; the oxygen enhancement ratio OER is measured from the blood oxygen level-dependent signal of real-time functional magnetic resonance imaging, and is a dimensionless parameter; and the cell survival curve αβ value is obtained by matching the tumor pathology type from the radiobiology database, in gray (Gy). Based on the extracted radiotherapy physical parameters and biological response parameters, calculate the cumulative biological equivalent dose BED at the current moment, and the formula is as follows: ; The definitions and acquisition methods of each parameter are as follows: BED represents the cumulative biological equivalent dose up to the current moment, with the unit of Gray (Gy). The higher the value, the higher the tumor control probability and the greater the risk of normal tissue complications; D represents the physical absorbed dose of a single radiotherapy, which is obtained by analyzing the dose output log of the radiotherapy device; αβ represents the linear-quadratic model parameter of the target tissue, which is matched from the pre-set radiobiological parameter library according to the tumor pathological type. For lung cancer, it is 10 Gy, and for prostate cancer, it is 3 Gy; k represents the oxygen enhancement ratio correction coefficient, taking an empirical value of 0.8 to balance the influence of oxygenation status on radiation sensitivity; OER represents the real-time measured tumor oxygen enhancement ratio, which is obtained by calculating the blood oxygenation level-dependent signal of functional magnetic resonance imaging; RV represents the change rate of the target volume relative to the initial volume, which is obtained by performing three-dimensional registration and segmentation calculations on continuously acquired CT images using medical image analysis software. Based on the calculated BED value, perform real-time comparison and decision-making with the pre-set normal tissue tolerance window [Tlow, Tup]: When BED < Tlow, increase the current nursing intervention intensity threshold by 15%, and push a strengthened nutrition support and early rehabilitation training enhancement instruction to the nursing intervention strategy generation module; when BED > Tup, decrease the current nursing intervention intensity threshold by 20%, and simultaneously trigger a suspension of high-intensity physical training and an instruction to increase antioxidant supplementation for protection; when Tlow ≤ BED ≤ Tup, keep the existing intervention intensity unchanged, and record the current BED value as the reference value for the next fractionated dose calibration. Write the calibrated intervention intensity threshold into the dynamic intervention label field of the patient's personal file in real-time, and push it to the intervention effect feedback module as the basis for weight allocation of the core feature indicators in the next acquisition cycle. Through this closed-loop adjustment mechanism, ensure that the nursing intervention intensity is precisely adaptive to the change of the biological equivalent dose until the end of the radiotherapy course or the system detects that BED is within ±5% of the central value of the tolerance window for 7 consecutive natural days, and then automatically lock the current intervention plan.

[0014] Furthermore, the tumor radiotherapy rehabilitation nursing intervention system based on artificial intelligence also includes a personalized dose-symptom association module, which is used to establish a quantitative association model between radiotherapy dose distribution and symptom manifestations and achieve symptom early warning, and perform the following steps: The user profile acquisition module extracts historical radiotherapy dose distribution data for patients, including the target dose homogeneity index HI, the volume of organs at risk of radiation exposure Vx, and the maximum dose Dmax. Simultaneously, the intervention effect feedback module retrieves corresponding symptom records, including radiation dermatitis grading, mucositis scores, and fatigue assessments. The dose homogeneity index HI is obtained by calculating the homogeneity of the target dose distribution and is a dimensionless parameter. The volume of organs at risk of radiation exposure Vx represents the percentage of volume receiving doses of x Gy or higher, obtained through dose-volume histogram analysis. The maximum dose Dmax is obtained by analyzing the peak dose in the dose distribution cloud map, expressed in Gray (Gy). Symptom records are extracted in a structured manner from the electronic medical record system and quantified and graded according to CTCAE standards. Based on the extracted dose parameters and symptom data, a dose-symptom association model is constructed, as shown in the following formula: ; Where: S is the comprehensive symptom prediction score, which is a dimensionless parameter. The higher the value, the greater the risk of symptoms. , , The weighting coefficients are determined based on multiple regression analysis and satisfy the following conditions: HI stands for Target Dose Uniformity Index. The percentage of lung volume irradiated with 20 Gy or more; α is the organ-specific sensitivity coefficient, taken as 0.3 for lung cancer radiotherapy and 0.5 for head and neck radiotherapy; Dmax is the maximum dose to the organ at risk; Dref is the reference dose, taken as the tolerable dose limit of the organ; EUD is the equivalent uniform dose to the organ at risk, calculated using the formula... Calculated, where For relative volume, Where 'a' represents the corresponding dose, and 'a' represents an organ-specific parameter. Based on the calculated comprehensive symptom prediction score S, and the preset symptom risk threshold... Comparison: When When the risk is low, maintain the current care intervention strategy; when When the risk level is medium, targeted symptom monitoring and preventative interventions should be added to the nursing intervention strategy; when... When high risk is detected, a symptom alert is triggered and an enhanced symptom management plan is initiated, including adjusting radiotherapy plan parameters and optimizing supportive drug treatment. The symptom risk assessment results are updated in real time to the symptom warning field of the patient's personal file and linked with the nursing intervention strategy generation module to dynamically adjust the intervention focus and monitoring frequency, forming a dose-symptom closed-loop management until the radiotherapy course ends or three consecutive assessments show that the symptom risk has dropped to a low-risk level.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention comprehensively integrates multi-dimensional data from radiotherapy fundamentals, physiological monitoring, individual patient data, and historical rehabilitation feedback. By combining clinical standard outputs of quantitative assessment levels and key weak dimensions, it accurately identifies patients' rehabilitation shortcomings, providing targeted objectives for the nursing intervention strategy generation module. Based on the radiotherapy rehabilitation status assessment results, combined with intervention goals and individual patient constraints, it generates personalized plans covering diet, exercise, symptom management, and psychological counseling, ensuring the pertinence and feasibility of the intervention. 2. This invention captures strategy execution data in real time and dynamically adjusts the plan to ensure that the intervention process always adapts to changes in the patient's recovery status. At the same time, it implements dynamic monitoring and risk escalation mechanisms for high-risk patients. By regularly collecting core indicators and linking risk escalation and priority intervention, it strengthens proactive management of high-risk groups, realizing a shift from passive response to proactive prediction in tumor radiotherapy rehabilitation nursing, and optimizing the overall quality of post-radiotherapy rehabilitation nursing and patient prognosis. Attached Figure Description

[0016] Figure 1 This is a block diagram of the AI-based tumor radiotherapy rehabilitation nursing intervention system of the present invention. Detailed Implementation

[0017] 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.

[0018] To address the technical challenges of existing cancer radiotherapy rehabilitation nursing models—such as reliance on experience, limited perspectives that fail to comprehensively reflect rehabilitation status, generalized intervention strategies lacking specificity, and delayed feedback on intervention effects that prevents dynamic adjustments based on real-time patient recovery—and thus hindering the fulfillment of precision nursing needs, please refer to [the relevant documentation / reference]. Figure 1 This embodiment provides the following technical solution: An AI-based intervention system for cancer radiotherapy rehabilitation and nursing care includes: The user profile acquisition module is used to acquire the personal profiles of cancer radiotherapy patients and extract the corresponding rehabilitation data from the personal profiles, including basic radiotherapy data, physiological monitoring data, individual patient data, and historical rehabilitation feedback data. The radiotherapy rehabilitation status assessment module is used to quantitatively assess the rehabilitation status of patients after radiotherapy based on the acquired rehabilitation data and in combination with preset clinical rehabilitation standards, and output the rehabilitation status assessment results, including rehabilitation assessment level and key weak dimensions of rehabilitation. The nursing intervention strategy generation module is used to generate nursing intervention strategies for the patient based on the rehabilitation status assessment results, combined with intervention optimization goals and constraints. These strategies include various dimensions such as diet, exercise, symptom management, and psychological counseling, ensuring the safety, relevance, and feasibility of the nursing intervention strategies. The intervention effect feedback module is used to monitor the actual implementation effect of nursing intervention strategies in real time, obtain real-time feedback data during the implementation of nursing intervention strategies, evaluate the intervention effectiveness of nursing intervention strategies, and dynamically adjust nursing intervention strategies based on the intervention effectiveness evaluation results; Data on core characteristic indicators of patients marked as high-risk are collected at preset collection intervals. The core characteristic indicator data are compared with the preset improvement range. If the core characteristic indicators collected for a preset number of consecutive times are not lower than the preset improvement range, the risk level of the patient is increased and the priority intervention mechanism is retried until the core characteristic indicators return to the preset clinical safety range or the preset improvement range.

[0019] In this embodiment, the basic data for radiotherapy includes the radiotherapy site, total radiotherapy dose, fractionated dose, and radiotherapy cycle; In this embodiment, the physiological monitoring data includes routine blood indicators, white blood cell count, platelet count, liver and kidney function indicators, pain score, fatigue level, sleep duration and quality, etc. In this embodiment, individual patient data includes age, gender, underlying diseases, diabetes, cardiovascular diseases, tumor stage, post-radiotherapy complications, skin reactions, gastrointestinal reactions, etc. In this embodiment, historical rehabilitation feedback data includes dietary adherence, exercise performance, and subjective evaluation of symptom relief. In this embodiment, the rehabilitation status of brain structures such as the brain and amygdala after tumor radiotherapy can be accurately quantified and assessed by this system. By integrating multi-dimensional rehabilitation data, the system clearly identifies key weak dimensions that affect their recovery, avoiding rehabilitation delays caused by inaccurate assessments. For the risk of brain functional abnormalities that may induce epilepsy, the system can generate appropriate intervention strategies based on the weak dimensions to reduce the interference of abnormal brain activity on rehabilitation. The dynamic feedback mechanism can track the pace of brain function recovery in real time, making nursing interventions more consistent with the brain structure repair process. Compared with traditional static nursing, it is more in line with the individual brain rehabilitation characteristics and improves the targeting of interventions. In this embodiment, by comprehensively integrating multi-dimensional data on radiotherapy fundamentals, physiological monitoring, individual patient data, and historical rehabilitation feedback, and combining this with clinical standard output of quantitative assessment levels and key weaknesses, the system accurately identifies patients' rehabilitation shortcomings, providing targeted objectives for the nursing intervention strategy generation module. Based on the assessment results, considering intervention goals and individual patient constraints, a personalized plan covering diet, exercise, symptom management, and psychological counseling is generated, ensuring the intervention's relevance and feasibility. Real-time capture of strategy execution data and dynamic adjustment of the plan ensure the intervention process always adapts to changes in the patient's rehabilitation status. Simultaneously, dynamic monitoring and risk escalation mechanisms for high-risk patients are embedded in the overall process. By regularly collecting core indicators and linking risk escalation with priority intervention, proactive management of high-risk groups is strengthened. This effectively solves the comprehensive problems of data fragmentation, subjective assessment, generic intervention, and delayed risk warning in existing technologies. It realizes a shift from passive response to proactive prediction in tumor radiotherapy rehabilitation nursing, with multi-dimensional data and dynamically adapted plan development, improving the accuracy, safety, and effectiveness of rehabilitation nursing. This provides patients with full-cycle, personalized rehabilitation management support, significantly optimizing the overall quality of post-radiotherapy rehabilitation nursing and patient prognosis.

[0020] In this embodiment, the process by which the user profile acquisition module extracts the rehabilitation data corresponding to the patient includes: The extracted rehabilitation data is read, the data type of the rehabilitation data is determined, and the corresponding similar clinical samples are matched in the preset rehabilitation case database according to the data type of the rehabilitation data. Read the similar clinical samples and determine the distribution characteristics of the rehabilitation data of the similar clinical samples, such as the fluctuation range of white blood cell count and the trend of symptom score changes during the same period. Generate data verification constraints based on the distribution characteristics and data types of the rehabilitation data. The extracted rehabilitation data is validated based on constraints, and data that does not meet the constraints is marked as data to be verified based on the validation results. The reasons for the abnormality of the data to be verified are analyzed. For example, if it is due to data acquisition error, abnormal heart rate caused by equipment failure, or incorrect pain score, it is automatically removed and re-acquisition is triggered. If it is due to changes in the patient's actual condition, such as a sudden drop in platelet count caused by sudden complications or a sudden drop in sleep quality caused by psychological stress, it is marked as clinically important data and an abnormality degree label is attached, including mild / moderate / severe. Data features of the validated rehabilitation data are extracted, including time-series features of physiological indicators, symptom correlation features, and intervention response features. These data features are then weighted based on a pre-defined clinical priority system. This clinical priority system is set according to key influencing factors of tumor radiotherapy, including: target organ sensitivity weights corresponding to the radiotherapy site (e.g., oral mucosal reaction features of head and neck radiotherapy have higher weights than other features); weights related to underlying diseases (e.g., blood glucose-related physiological features of diabetic patients have higher weights than those of non-diabetic patients); and weights of complication risk levels (e.g., symptom features of grade III or higher skin reactions have higher weights than grade I reactions). The weighted feature data is structured and integrated based on the time dimension to construct a three-dimensional rehabilitation data matrix, which includes real-time snapshot data, periodic trend data, and historical correlation data. Redundancy processing is performed on the three-dimensional rehabilitation data matrix to retain core features that are strongly correlated with radiotherapy rehabilitation assessment (screened by Pearson correlation coefficient, with an absolute value of correlation coefficient ≥ 0.6). Finally, standardized rehabilitation data after verification, weighting, integration, and feature screening is output as input data for the radiotherapy rehabilitation status assessment module.

[0021] In this embodiment, the real-time snapshot data is the current pain score, the periodic trend data is the white blood cell count change curve over 7 consecutive days, and the historical correlation data is the correlation between dietary compliance and weight change during previous radiotherapy cycles. In this embodiment, similar clinical samples in a preset rehabilitation case database are matched according to data type. Targeted verification constraints are generated based on the distribution characteristics of rehabilitation data, making the verification standards more closely aligned with the patient's specific radiotherapy scenario and improving the accuracy of data verification. By analyzing and classifying abnormal causes, and combining with a preset clinical priority system, features that have a more critical impact on radiotherapy rehabilitation can be captured first. The three-dimensional rehabilitation data matrix achieves structured integration of the time dimension, and strongly correlated core features are selected and retained. This avoids the subsequent input chaos of the evaluation model caused by data fragmentation, and reduces the computational load of the model through redundancy processing, ensuring that the output standardized rehabilitation data can fully reflect the patient's rehabilitation dynamics, thus guaranteeing accuracy and effectiveness from the data source.

[0022] In this embodiment, the radiotherapy rehabilitation status assessment module includes: The rehabilitation status assessment model construction unit is used to construct a rehabilitation status assessment model based on preset clinical rehabilitation standards and combined with target data features in standardized rehabilitation data. The rehabilitation status assessment unit is used to input standardized rehabilitation data into the rehabilitation status assessment model to obtain the patient's comprehensive rehabilitation score, which ranges from 0 to 100 points. The score is then mapped to a rehabilitation assessment level based on a preset score range: 0-20 points correspond to severe lag (Level 1), 21-40 points correspond to moderate lag (Level 2), 41-60 points correspond to basic achievement (Level 3), 61-80 points correspond to good (Level 4), and 81-100 points correspond to excellent (Level 5). The weak dimension identification unit is used to extract the influence weight of each data feature in the standardized rehabilitation data on the comprehensive rehabilitation score, determine the contribution of each dimension feature, and select the dimension feature with the greatest impact on the rehabilitation level as the key weak dimension of rehabilitation, such as low white blood cell count, high VAS pain score, poor exercise compliance and other descriptions directly related to clinical indicators. The risk warning unit is used to mark patients whose rehabilitation assessment level is lower than the preset level or whose characteristic indicators in a certain dimension exceed the preset clinical safety range as high-risk and trigger a priority intervention mechanism: push high-risk warning information to medical and nursing terminals in real time, and simultaneously retrieve the patient's three-dimensional rehabilitation data matrix and key weak dimensions to provide an urgent processing basis for the nursing intervention strategy generation module.

[0023] In this embodiment, the preset clinical safety range includes a white blood cell count < Liver and kidney function indicators exceeded the upper limit of normal by 1.5 times; In this embodiment, a model is constructed by using preset clinical rehabilitation standards as the core and combining target features in standardized rehabilitation data to ensure that the assessment results are highly compatible with clinical rehabilitation needs. By combining the comprehensive rehabilitation score with the interval mapping of rehabilitation assessment levels, the rehabilitation status has a quantifiable and comparable unified standard, which solves the problem of individual differences in assessment results among different medical staff. The influence weight of each data feature on the comprehensive score is extracted, and the key weak dimensions directly related to clinical indicators are accurately located, which effectively improves the accuracy and response efficiency of the radiotherapy rehabilitation nursing intervention system.

[0024] In this embodiment, the specific process for identifying key weaknesses in rehabilitation includes: The target data features in standardized rehabilitation data are classified according to clinical dimensions, namely physiological indicators, symptom control, psychological state, and daily living ability. Calculate the influence weight of each target data feature under each dimension on the comprehensive rehabilitation score, and adjust the influence weight of each target data feature in conjunction with the clinical priority system: if a feature belongs to a high priority category, such as the oral mucosal reaction feature of patients undergoing head and neck radiotherapy or the blood glucose-related feature of diabetic patients, then add a priority coefficient of 1.2-1.5 times on the original influence weight. The corrected impact weights are arranged from largest to smallest. Target data features are selected based on the impact weight queue, and the clinical dimensions corresponding to target data features that exceed the preset clinical reference range are taken as key weak dimensions for rehabilitation. Retrieve rehabilitation data of similar patients in the preset rehabilitation case database whose radiotherapy protocol similarity is greater than a preset similarity threshold, compare the patient's weakness dimension with the difference of similar patients, and if the difference is greater than a preset difference threshold, then confirm that the rehabilitation key weakness dimension is the final rehabilitation key weakness dimension.

[0025] In this embodiment, the correction of the influence weights of each target data feature also includes: The marginal contribution of each target feature in the standardized rehabilitation data to the comprehensive rehabilitation score in the rehabilitation status assessment model is quantified, and the impact of changes in the feature values ​​of the target features on the score results is determined. Multiple assessment samples from the same patient were aggregated and the mean was calculated as the original influence weight of the target feature. At the same time, the direction of the influence of the target feature was marked. If the overall rehabilitation score showed an upward trend when the value of the target feature increased, such as an increase in exercise execution rate or dietary compliance, it was marked as a positive influence. If the overall rehabilitation score showed a downward trend when the value of the target feature increased, such as an increase in pain score or a worsening of fatigue, it was marked as a negative influence. The sensitivity coefficient of the target organ at the radiotherapy site, the correlation with underlying diseases, and the risk level of complications are obtained according to the clinical priority system, and the priority coefficient is obtained based on the sensitivity coefficient of the target organ at the radiotherapy site, the correlation with underlying diseases, and the risk level of complications. Based on the priority coefficient and the original influence weight of the target feature, and combined with the direction calibration coefficient corresponding to the influence direction of the target feature, the original influence weight is adjusted to obtain the corrected influence weight.

[0026] In this embodiment, the sensitivity of the target organ at the radiotherapy site is divided into three levels: high, medium, and low, with priority coefficients of 1.2-1.6, 1.0-1.2, and 0.8-1.0, respectively; the correlation of the underlying disease is divided into three levels: strong, medium, and weak, with priority coefficients of 1.3-1.5, 1.1-1.3, and 0.9-1.1, respectively; the risk level of complications is divided into four levels (I-IV) according to CTCAE, with priority coefficients of 0.9-1.1, 1.1-1.3, 1.3-1.5, and 1.5-1.7, respectively; and the priority coefficient of each target feature is the product of the priority coefficient of its category. In this embodiment, if the direction of influence of the target feature is consistent with the clinical rehabilitation expectation, the direction calibration coefficient is set to 1. If the direction of influence of the target feature deviates from the clinical rehabilitation expectation, the direction calibration coefficient is set to 0.8-0.9 in combination with a preset clinical experience threshold. This value is used to adjust and weaken the interference of abnormal deviation on the weight. In this embodiment, triggering the priority intervention mechanism further includes: risk stratifying patients marked as high-risk, and assigning patients with a rehabilitation assessment level of 1 (severely lagging) or core characteristic indicators exceeding a preset high-multiple threshold of the preset clinical safety range (such as white blood cell count < 1). Patients with a rehabilitation assessment level of 2 (moderate lag) or whose core characteristic indicators exceed the preset low threshold of the preset clinical safety range are classified as emergency risk. The system integrates standardized rehabilitation data and key weak dimensions of rehabilitation data corresponding to patients according to their risk levels. For patients at high risk, it integrates real-time snapshot data, periodic trend data over the past 3 days, characteristic change curves of key weak dimensions, and basic disease emergency plans from the patient's individual data in the three-dimensional rehabilitation data matrix. For patients at regular risk, it integrates key weak dimension data and similar intervention response records from historical related data. It triggers a multi-terminal collaborative response, pushing a high-risk warning pop-up to the responsible nurse's terminal, including the risk level, core abnormal indicators, and weak dimensions. The pop-up has a read confirmation button, and a reminder is given every 5 minutes if not confirmed. It pushes a structured report to the attending physician's terminal, including data evidence and preliminary intervention suggestions, such as suggesting a high-protein diet and suspending high-intensity exercise if the white blood cell count is low. It also simultaneously marks the priority of the patient's personal file processing status and displays it at the top of the nursing intervention strategy generation queue.

[0027] In this embodiment, the target features are divided into four major clinical dimensions: physiological indicators, symptom control, psychological state, and daily living abilities. This covers all influencing factors of radiotherapy rehabilitation, avoiding misjudgment of weaknesses caused by cross-dimensional confusion. By combining the clinical priority system with the superposition coefficient of high-priority features, the marginal contribution of quantitative features to the comprehensive score, and the marking of the direction of influence, and incorporating the sensitivity of the target organ at the radiotherapy site, the correlation with underlying diseases, the risk level of complications, and the direction calibration coefficient for weight correction, the weights are aligned with the individual patient's treatment background, avoiding the problems of misjudgment of negative features and underestimation of high clinical value features, and ensuring the clinical suitability of the weight calculation. By comparing the difference in data of similar patients who have met the similarity standard of radiotherapy plans, normal rehabilitation fluctuations that are common in similar groups are excluded, and only abnormal weaknesses that significantly deviate from the group are identified, further reducing the risk of misjudgment. This avoids the blindness of traditional priority interventions that only know the risks but not the core issues, allowing intervention resources to focus on key rehabilitation shortcomings and connecting with the targeted optimization needs of the nursing intervention strategy generation module, laying the foundation for accurate problem positioning in the formulation of personalized intervention plans.

[0028] In this embodiment, the nursing intervention strategy generation module includes: The objective function construction unit is used to construct a multi-objective optimization objective function based on the key weak dimensions and risk levels of rehabilitation, combined with a preset clinical priority system, and to determine the constraints of the multi-objective optimization objective function. The optimization strategy generation unit is used to retrieve effective intervention plans from rehabilitation data of similar patients, input each plan into the objective function of multi-objective optimization, calculate the fitness of the plan, and verify whether the plan meets the constraints of the objective function. The objective function is iteratively optimized based on multi-objective optimization, and its clinical feasibility is verified. Based on the verification results, the plan with the highest fitness is selected. The optimal nursing intervention strategy is generated according to the dimensions of diet, exercise, symptom management, and psychological counseling. The execution standards for each dimension are marked: daily breakfast intake of 2 eggs + 200ml of milk, 15 minutes of gentle stretching exercise at 15:00 every day, monitoring requirements: pain score is recorded 2 hours after execution, and the trigger conditions are adjusted: if nausea and vomiting occur, greasy food should be stopped. The results are synchronously linked to the patient's personal file, and the execution status of the nursing intervention strategy is marked.

[0029] In this embodiment, individual contraindications, clinical safety constraints, and execution feasibility constraints are used as constraints on the objective function to ensure that the values ​​of the decision variables are within the constraints. In this embodiment, the objective function of multi-objective optimization has the core objectives of maximizing the improvement of key weak dimensions of rehabilitation, maximizing patient intervention compliance, and minimizing risk level adaptation bias. In this embodiment, clinical feasibility verification includes confirming the suitability of the protocol to the patient's high-risk characteristics. For patients with emergency risks, it is necessary to additionally verify whether the protocol is compatible with their underlying disease emergency plan. Based on individual patient data, the parameters of the compliance prediction model are adjusted, and the feasibility of the protocol is reassessed.

[0030] In this embodiment, by identifying key weaknesses and risk levels in rehabilitation, a multi-objective optimization function is constructed based on a pre-defined clinical priority system. The core objectives are to maximize the improvement in weaknesses, maximize patient adherence to interventions, and minimize risk level mismatch. This ensures the strategy directly addresses the patient's core rehabilitation shortcomings while also considering the patient's willingness to implement the plan and risk suitability. Individual contraindications, clinical safety constraints, and feasibility constraints are incorporated into the objective function's constraints to ensure decision variables remain within a safe and feasible range. Effective intervention plans for similar patients are retrieved from a pre-defined rehabilitation case database. These plans are input into the objective function to calculate fitness and verify constraints. Further iterative optimization and clinical feasibility verification ensure that the final high-fitness plan aligns with both individual patient characteristics and clinical treatment guidelines. Finally, an optimal strategy with clearly defined implementation standards and adjustment trigger conditions is generated based on dimensions of diet, exercise, symptom management, and psychological support. This strategy is simultaneously linked to the patient's personal file and execution status is marked, providing clear monitoring data for the subsequent intervention effect feedback module and ensuring the precision and effectiveness of rehabilitation nursing from the execution stage.

[0031] In this embodiment, the AI-based tumor radiotherapy rehabilitation nursing intervention system further includes a dynamic dose-response adaptive adjustment module, which is used to calculate the cumulative bioequivalent dose of radiotherapy in real time and dynamically calibrate the intervention intensity threshold during the execution of the nursing intervention strategy, and perform the following steps: The core characteristic indicators collected in real time from the intervention effect feedback module include the radiotherapy fractionated dose record D, fractionated irradiation time interval ΔT, target volume change rate RV, oxygen enhancement ratio OER, and cell survival curve αβ value. Specifically, the fractionated dose record D is obtained from the dosimetric report output by the radiotherapy equipment, in gray (Gy); the fractionated irradiation time interval ΔT is calculated from the schedule records of the radiotherapy information management system, in hours (h); the target volume change rate RV is calculated by comparing the target volume contours at the current and initial times, and is a dimensionless parameter; the oxygen enhancement ratio OER is measured from the blood oxygen level-dependent signal of real-time functional magnetic resonance imaging, and is a dimensionless parameter; and the cell survival curve αβ value is obtained by matching the tumor pathology type from the radiobiology database, in gray (Gy). Based on the extracted radiotherapy physical parameters and biological response parameters, the cumulative bioequivalent dose (BED) at the current moment is calculated using the following formula: ; The definitions and acquisition methods of each parameter are as follows: BED represents the cumulative biological equivalent dose up to the current moment, with the unit of Gray (Gy). The higher the value, the higher the tumor control probability and the greater the risk of normal tissue complications; D represents the physical absorbed dose of a single radiotherapy, which is obtained by analyzing the dose output log of the radiotherapy equipment; αβ represents the linear-quadratic model parameter of the target tissue, which is matched from a preset radiobiological parameter library according to the tumor pathological type. For lung cancer, it takes 10 Gy, and for prostate cancer, it takes 3 Gy; k represents the oxygen enhancement ratio correction coefficient, taking an empirical value of 0.8, which is used to balance the influence of the oxygenation state on radiation sensitivity; OER represents the tumor oxygen enhancement ratio measured in real time, which is obtained by calculating the blood oxygenation level-dependent signal of functional magnetic resonance imaging; RV represents the change rate of the target volume relative to the initial volume, which is obtained by performing three-dimensional registration and segmentation calculations on continuously acquired CT images through medical image analysis software. Based on the calculated BED value, it is compared and decided with the preset normal tissue tolerance window [Tlow, Tup] in real time: when BED < Tlow, the current nursing intervention intensity threshold is increased by 15%, and a strengthening instruction for nutritional support and early rehabilitation training is pushed to the nursing intervention strategy generation module; when BED > Tup, the current nursing intervention intensity threshold is decreased by 20%, and at the same time, a command to pause high-intensity physical training and increase antioxidant supplementation protection is triggered; when Tlow ≤ BED ≤ Tup, the existing intervention intensity remains unchanged, and the current BED value is recorded as the benchmark reference value for the next fractionated dose calibration. The calibrated intervention intensity threshold is written into the dynamic intervention label field of the patient's personal file in real time and pushed to the intervention effect feedback module as the basis for weight allocation of the core feature indicators in the next acquisition cycle. Through this closed-loop adjustment mechanism, it is ensured that the nursing intervention intensity is precisely adapted to the change of the biological equivalent dose until the end of the radiotherapy course or the system detects that the BED is within ±5% of the central value of the tolerance window for 7 consecutive natural days, and the current intervention plan is automatically locked.

[0032] The dynamic dose-effect adaptive adjustment module is based on a linear quadratic model in radiobiology. It dynamically adjusts the intensity of nursing interventions by calculating the cumulative bioequivalent dose in real time. Specifically, the fractionated dose record (D) is obtained by parsing the dose grid data in the DICOM-RT dosimetry report of the radiotherapy equipment, extracting the MU value of each control point and the machine output factor for calculation. The fractionated irradiation time interval (ΔT) is obtained from the work log of the radiotherapy information management system, accurately recording the start and end timestamps of each irradiation and calculating the difference. The target volume change rate (RV) is calculated by performing three-dimensional registration on continuously acquired CT images using medical image analysis software, employing a deformation registration algorithm to calculate the ratio of the current target volume to the initial planned volume. The oxygen enhancement ratio (OER) is measured using blood oxygen level-dependent signals from functional magnetic resonance imaging (fMRI), specifically by performing a BOLD-fMRI scan before radiotherapy to calculate the relative change rate of the T2 value in the tumor region. The cell survival curve (αβ) value is matched from a pre-set radiobiological parameter library according to the tumor pathology type. Based on these parameters, the bioequivalent dose formula is applied. Calculations were performed, with the oxygen enhancement ratio correction coefficient k set to 0.8 to balance the dynamic changes in oxygenation status. The calculated BED value was compared with the preset normal tissue tolerance window. When the BED was lower than the lower limit of the tolerance window, the intervention intensity was increased by 15%, and when the BED exceeded the upper limit of the tolerance window, the intervention intensity was decreased by 20%, ensuring that the nursing intervention intensity was accurately matched with the radiobiological effect.

[0033] To better illustrate this embodiment, the following specific implementation examples are provided: In clinical application, the radiotherapy process of a non-small cell lung cancer patient was monitored. The fractionated dose D=2.0Gy was obtained from the DICOM-RT file of the Varian medical linear accelerator. The fractionation interval ΔT=24h was extracted from the MOSAIQ radiotherapy information system. Deformation registration was performed on weekly CT scan images using the RayStation treatment planning system, and the target volume change rate RV=0.85 was calculated. A BOLD-fMRI scan was performed before radiotherapy, and the tumor area OER=1.8 was measured. Based on the adenocarcinoma type in the pathology report, αβ=10Gy was matched. The BED for a single irradiation was calculated using the formula: D×(1+D / (αβ))=2.0×(1+2.0 / 10) = 2.4 Gy, (1 + k × (OER-1) × (1-RV)) = (1 + 0.8 × (1.8-1) × (1-0.85)) = 1.096, single BED = 2.4 × 1.096 = 2.6304 Gy; considering the cumulative effect of 30 irradiations, the total BED = 30 × 2.6304 = 78.912 Gy. Compared with the preset lung tissue tolerance window [60 Gy, 80 Gy], since the BED is slightly higher than the upper limit of the tolerance window, the system will lower the current nursing intervention intensity threshold by 20% and simultaneously trigger the instruction to suspend high-intensity physical training and increase antioxidant supplementation protection; at the same time, the calibrated intervention intensity threshold will be written into the dynamic intervention tag of the patient's personal file for use in the next collection cycle.

[0034] In this embodiment, the AI-based tumor radiotherapy rehabilitation nursing intervention system further includes a personalized dose-symptom correlation module, used to establish a quantitative correlation model between radiotherapy dose distribution and symptom manifestations and to achieve symptom early warning, performing the following steps: The user profile acquisition module extracts historical radiotherapy dose distribution data for patients, including the target dose homogeneity index HI, the volume of organs at risk of radiation exposure Vx, and the maximum dose Dmax. Simultaneously, the intervention effect feedback module retrieves corresponding symptom records, including radiation dermatitis grading, mucositis scores, and fatigue assessments. The dose homogeneity index HI is obtained by calculating the homogeneity of the target dose distribution and is a dimensionless parameter. The volume of organs at risk of radiation exposure Vx represents the percentage of volume receiving doses of x Gy or higher, obtained through dose-volume histogram analysis. The maximum dose Dmax is obtained by analyzing the peak dose in the dose distribution cloud map, expressed in Gray (Gy). Symptom records are extracted in a structured manner from the electronic medical record system and quantified and graded according to CTCAE standards. Based on the extracted dose parameters and symptom data, a dose-symptom association model is constructed, as shown in the following formula: ; Where: S is the comprehensive symptom prediction score, which is a dimensionless parameter. The higher the value, the greater the risk of symptoms. , , The weighting coefficients are determined based on multiple regression analysis and satisfy the following conditions: HI stands for Target Dose Uniformity Index. The percentage of lung volume irradiated with 20 Gy or more; α is the organ-specific sensitivity coefficient, taken as 0.3 for lung cancer radiotherapy and 0.5 for head and neck radiotherapy; Dmax is the maximum dose to the organ at risk; Dref is the reference dose, taken as the tolerable dose limit of the organ; EUD is the equivalent uniform dose to the organ at risk, calculated using the formula... Calculated, where For relative volume, Where 'a' represents the corresponding dose, and 'a' represents an organ-specific parameter. Based on the calculated comprehensive symptom prediction score S, and the preset symptom risk threshold... Comparison: When When the risk is low, maintain the current care intervention strategy; when When the risk level is medium, targeted symptom monitoring and preventative interventions should be added to the nursing intervention strategy; when... When high risk is detected, a symptom alert is triggered and an enhanced symptom management plan is initiated, including adjusting radiotherapy plan parameters and optimizing supportive drug treatment. The symptom risk assessment results are updated in real time to the symptom warning field of the patient's personal file and linked with the nursing intervention strategy generation module to dynamically adjust the intervention focus and monitoring frequency, forming a dose-symptom closed-loop management until the radiotherapy course ends or three consecutive assessments show that the symptom risk has dropped to a low-risk level.

[0035] The personalized dose-symptom association module works by establishing a quantitative relationship between dosimetric parameters and clinical symptoms to predict symptom risk. Specifically, the target dose uniformity index (HI) is calculated by analyzing dose distribution data from the treatment planning system using the formula HI = (D5 - D95) / Dprescription, where D5 and D95 ​​represent the dose received by 5% and 95% of the target volume, respectively. The irradiated volume of organs at risk (Vx) is obtained by analyzing dose-volume histogram data, specifically extracting the relative volume percentage corresponding to a given dose level. The maximum dose (Dmax) is obtained by analyzing hotspot areas in the dose distribution cloud map and taking the maximum dose of all voxels. Symptom records are extracted from a structured electronic medical record system, quantifying symptoms such as radiation dermatitis and mucositis into levels 1-5 according to the CTCAE 5.0 standard. Based on these parameters, a dose-symptom association model is constructed. The weighting coefficient , , The optimal weight combination was determined through multiple linear regression analysis and fitted to historical data using the least squares method; the organ-specific sensitivity coefficient α was determined based on the radiation site; the reference dose Dref was taken from the organ tolerance dose limit recommended by QUANTEC; and the equivalent uniform dose EUD was determined using the formula. Calculate the relative volume. Extracted from the dose-volume histogram; the calculated comprehensive symptom prediction score S is compared with a preset threshold to achieve symptom risk classification and early warning.

[0036] To better illustrate this embodiment, the following specific implementation examples are provided: In symptom monitoring of patients with head and neck tumors, dose distribution data were exported from the Pinnacle treatment planning system, and the HI was calculated to be 0.15; parotid gland tissue was extracted from the dose-volume histogram. Analysis of the dose-hysteresis plot yielded a spinal cord Dmax of 42.5 Gy; according to CTCAE standards, grade 2 mucositis and grade 1 dermatitis were recorded; weighting coefficients were obtained based on regression analysis of historical data. , , With a head and neck radiotherapy sensitivity coefficient α=0.5 and a spinal cord reference dose Dref=45Gy, the calculated EUD=36.8Gy was used to calculate the comprehensive symptom prediction score. First item =0.4×0.15×(1+0.5×ln(0.65))≈0.4×0.15×(1+0.5×(-0.4308))=0.4×0.15×0.7846=0.047076, the second term =0.3×42.5 / 45=0.3×0.944=0.2832, the third term =0.3×36.8 / 45=0.3×0.818=0.2454, the total S=0.047076+0.2832+0.2454=0.575676≈0.576; compared with the preset threshold In comparison, S values ​​between 0.4 and 0.6 indicate a medium risk level. Targeted symptom monitoring and preventative interventions should be added to the nursing intervention strategy, and the assessment results should be updated to the symptom warning field in the patient's personal file.

[0037] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An artificial intelligence-based tumor radiotherapy rehabilitation nursing intervention system, characterized in that, The method comprises the following steps: a user profile acquisition module is used to acquire the personal profile of a tumor radiotherapy patient and extract the rehabilitation data corresponding to the patient from the personal profile; a radiotherapy rehabilitation state evaluation module is used to quantitatively evaluate the rehabilitation state of the patient after radiotherapy based on the acquired rehabilitation data and in combination with a preset clinical rehabilitation standard, and output a rehabilitation state evaluation result, including a rehabilitation evaluation grade and a rehabilitation key weak dimension; a nursing intervention strategy generation module is used to generate a nursing intervention strategy for the patient based on the rehabilitation state evaluation result, in combination with an intervention optimization target and a constraint condition; an intervention effect feedback module is used to monitor the actual execution effect of the nursing intervention strategy in real time, acquire real-time feedback data in the execution process of the nursing intervention strategy, evaluate the intervention effectiveness of the nursing intervention strategy, and dynamically adjust the nursing intervention strategy according to the intervention effectiveness evaluation result.

2. The artificial intelligence-based tumor radiotherapy rehabilitation nursing intervention system according to claim 1, wherein The process in which the user profile acquisition module extracts the rehabilitation data corresponding to the patient comprises the following steps: the extracted rehabilitation data is read, the data type of the rehabilitation data is determined, and corresponding similar clinical samples are matched in a preset rehabilitation case database according to the data type of the rehabilitation data; the similar clinical samples are read, the rehabilitation data distribution characteristics of the similar clinical samples are determined, and constraint conditions for data verification are generated based on the rehabilitation data distribution characteristics and the data type; the extracted rehabilitation data is subjected to data verification based on the constraint conditions, and the data to be verified that does not meet the constraint conditions is marked based on the verification result, and the abnormal reasons of the data to be verified are analyzed; the data characteristics of the rehabilitation data after verification are extracted, and the data characteristics are subjected to weighted processing based on a preset clinical priority system, wherein the clinical priority system is set according to key influencing factors of tumor radiotherapy; the characteristic data after weighted processing is integrated based on the time dimension, a three-dimensional rehabilitation data matrix is constructed, and the three-dimensional rehabilitation data matrix is subjected to redundancy processing, and standardized rehabilitation data is output as input data of the radiotherapy rehabilitation state evaluation module. 3.The AI-based tumor radiotherapy rehabilitation nursing intervention system of claim 2, wherein, The radiotherapy rehabilitation state evaluation module comprises: a rehabilitation state evaluation model construction unit is used to construct a rehabilitation state evaluation model based on a preset clinical rehabilitation standard and in combination with target data characteristics in the standardized rehabilitation data; a rehabilitation state evaluation unit is used to input the standardized rehabilitation data into the rehabilitation state evaluation model, acquire a comprehensive rehabilitation score of the patient, and map a rehabilitation evaluation grade based on a preset score interval; a weak dimension identification unit is used to extract the influence weight of each data characteristic in the standardized rehabilitation data on the comprehensive rehabilitation score, determine the contribution degree of each dimension characteristic, and select the dimension characteristic that has the greatest influence on the rehabilitation grade as a rehabilitation key weak dimension; a risk early warning unit is used to mark a patient as high risk if the rehabilitation evaluation grade is lower than a preset grade or a dimension characteristic index exceeds a preset clinical safety range, and trigger a priority intervention mechanism. 4.The AI-based tumor radiotherapy rehabilitation nursing intervention system of claim 3, wherein, The specific identification process of the rehabilitation key weak dimension comprises the following steps: the target data characteristics in the standardized rehabilitation data are classified according to clinical dimensions, and are divided into a physiological index dimension, a symptom control dimension, a psychological state dimension and a life ability dimension; Calculate the influence weight of each target data feature on the comprehensive rehabilitation score in each dimension, and correct the influence weight of each target data feature according to the clinical priority system; Arrange the corrected influence weights in descending order, select target data features based on the influence weight queue, and take the clinical dimension of the target data feature that exceeds the preset clinical reference range as the rehabilitation key weak dimension; Retrieve the rehabilitation data of patients with similar radiotherapy schemes from the preset rehabilitation case database, compare the difference amplitude of the weak dimension of the patient and the similar patients, and if the difference amplitude is greater than the preset difference amplitude threshold, the rehabilitation key weak dimension is confirmed as the final rehabilitation key weak dimension. 5.The tumor radiotherapy rehabilitation nursing intervention system based on artificial intelligence according to claim 4, wherein, The correction of the influence weight of each target data feature also includes: Quantify the marginal contribution of each target feature in the standardized rehabilitation data to the comprehensive rehabilitation score in the rehabilitation state evaluation model, and determine the influence amplitude of the feature value change of the target feature on the score result; Summarize and calculate the mean of multiple evaluation samples of the same patient as the original influence weight of the target feature, and mark the influence direction of the target feature; Obtain the target organ sensitivity coefficient of the radiotherapy site, the correlation degree of the underlying disease, and the complication risk level according to the clinical priority system, and obtain the priority coefficient based on the target organ sensitivity coefficient of the radiotherapy site, the correlation degree of the underlying disease, and the complication risk level; Adjust the original influence weight based on the priority coefficient and the original influence weight of the target feature, and combine the direction calibration coefficient corresponding to the influence direction of the target feature to obtain the corrected influence weight. 6.The tumor radiotherapy rehabilitation nursing intervention system based on artificial intelligence according to claim 5, wherein, Triggering the priority intervention mechanism also includes: risk grading for patients marked as high risk, and integrating the standardized rehabilitation data and rehabilitation key weak dimension data corresponding to the patients according to the risk level, and marking the personal file processing state priority of the patient. 7.The tumor radiotherapy rehabilitation nursing intervention system based on artificial intelligence according to claim 6, wherein, The nursing intervention strategy generation module includes: A target function construction unit for constructing a multi-objective optimization target function based on the rehabilitation key weak dimension and the risk level, and combining the preset clinical priority system to determine the constraint conditions of the multi-objective optimization target function; An optimization strategy generation unit for retrieving effective intervention schemes of similar patients, inputting each group of schemes into the multi-objective optimization target function, calculating the scheme fitness, and verifying whether the scheme meets the constraint conditions of the target function; Iterative optimization based on the multi-objective optimization target function, clinical feasibility verification, selection of the scheme with the highest fitness based on the verification result, generation of the optimal nursing intervention strategy, synchronization to the personal file of the patient, and marking of the execution state of the nursing intervention strategy. 8.The tumor radiotherapy rehabilitation nursing intervention system based on artificial intelligence according to claim 7, wherein, The intervention effect feedback module also includes acquiring core feature indicator data of patients marked as high risk at a preset collection time interval, comparing the core feature indicator data with the preset improvement range, and if the core feature indicator collected for a continuous preset number of times is not lower than the preset improvement range, the risk level of the patient is increased and the priority intervention mechanism is retriggered until the core feature indicator returns to the preset improvement range. 9.The tumor radiotherapy rehabilitation nursing intervention system based on artificial intelligence according to claim 8, wherein, Further comprising a dynamic dose-effect adaptive adjustment module for calculating the cumulative biological equivalent dose of radiotherapy in real time during the execution of the nursing intervention strategy and dynamically calibrating the intervention intensity threshold, performing the following steps: From the core feature index data collected in real time by the intervention effect feedback module, the radiotherapy fraction dose record D, the fraction irradiation time interval ΔT, the target volume change rate RV, the oxygen enhancement ratio OER, and the cell survival curve αβ value are extracted; wherein the fraction dose record D is obtained through the dosimetry report output by the radiotherapy equipment; the fraction irradiation time interval ΔT is obtained through the schedule record of the radiotherapy information management system; the target volume change rate RV is calculated by comparing the target contouring volume at the current time and the initial time; the oxygen enhancement ratio OER is measured by the blood oxygen level dependent signal of real-time functional magnetic resonance imaging; the cell survival curve αβ value is matched and obtained from the radiobiology database according to the tumor pathology type; Based on the extracted radiotherapy physical parameters and biological response parameters, the cumulative biological equivalent dose BED at the current time is calculated, and the formula is as follows: ； Wherein: BED represents the cumulative biological equivalent dose up to the current time, the higher the value, the higher the probability of tumor control and the greater the risk of normal tissue complications; D represents the physical absorbed dose of single radiotherapy, which is obtained by analyzing the dose output log of the radiotherapy equipment; αβ represents the linear-quadratic model parameter of the target tissue, which is matched from the pre-set radiobiology parameter library according to the tumor pathology type; k represents the oxygen enhancement ratio correction coefficient; OER represents the real-time measured tumor oxygen enhancement ratio, which is calculated by the blood oxygen level dependent signal of functional magnetic resonance imaging; RV represents the relative change rate of the target volume compared with the initial volume, which is calculated by three-dimensional registration and segmentation of the continuously collected CT images through medical image analysis software; Based on the calculated BED value, it is compared with the pre-set normal tissue tolerance window [Tlow, Tup] in real time and a decision is made: When BED When BED When Tlow≤BED≤Tup, the existing intervention intensity remains unchanged, and the current BED value is recorded as the reference value for the next fraction dose calibration; The calibrated intervention intensity threshold is written into the dynamic intervention tag field of the patient's personal file in real time, and is pushed to the intervention effect feedback module as the weight allocation basis for the next collection cycle core feature index. 10.The tumor radiotherapy rehabilitation nursing intervention system based on artificial intelligence according to claim 9, wherein, Further comprising a personalized dose-symptom correlation module for establishing a quantitative correlation model between radiotherapy dose distribution and symptom manifestation and realizing symptom warning, performing the following steps: The historical radiotherapy dose distribution data of the patient is extracted based on the user profile acquisition module, including the target dose homogeneity index HI, the irradiated volume of the critical organ Vx, and the maximum dose Dmax. Meanwhile, the corresponding symptom manifestation records are obtained from the intervention effect feedback module, including the radiation dermatitis grading, the mucositis score, and the fatigue degree assessment. The dose homogeneity index HI is obtained by calculating the uniformity of the target dose distribution. The irradiated volume of the critical organ Vx represents the volume percentage receiving a dose of x Gy or more, which is obtained by dose-volume histogram analysis. The maximum dose Dmax is obtained by analyzing the peak dose in the dose distribution cloud map. The symptom manifestation records are extracted from the electronic medical record system structure, and are quantitatively graded according to the CTCAE standard. Based on the extracted dose parameters and symptom data, a dose-symptom association model is constructed, with the formula as follows: ； Wherein: S is the comprehensive symptom prediction score, the higher the value, the greater the risk of symptoms; , , is a weight coefficient, determined based on multiple regression analysis, satisfying ; HI is the target area dose uniformity index; is the percentage of the volume of the lung irradiated above 20Gy; α is the organ-specific sensitivity coefficient; Dmax is the maximum dose of the organ at risk; Dref is the reference dose, taking the tolerance dose limit of the organ; EUD is the equivalent uniform dose of the organ at risk; based on the calculated comprehensive symptom prediction score S, with a preset symptom risk threshold comparison is made: When Low risk, maintain current care intervention strategy; When at risk, increasing targeted symptom monitoring and preventive interventions in the care intervention strategy; When high risk, a symptom warning is triggered and an intensive symptom management protocol is initiated, including adjustment of radiotherapy plan parameters and optimization of supportive drug therapy; The symptom risk assessment results are updated in real time to the symptom warning field of the patient's personal profile, and are linked with the nursing intervention strategy generation module to dynamically adjust the intervention focus and monitoring frequency, forming a dose-symptom closed-loop management.

Images