Intelligent tumor nursing monitoring system

Abstract

The invention provides an intelligent tumor nursing monitoring system. The intelligent tumor nursing monitoring system comprises a data acquisition module, a data processing module, a dynamic evaluation module, a multi-stage early warning module and a self-adaptive control module. The data acquisition module acquires physiological sign data, tumor local image data and body fluid biochemical data of a patient in real time through a multi-source biosensor. And the data processing module performs space-time alignment, noise suppression and cross-modal feature association on the multi-modal data to generate a comprehensive feature map. The dynamic evaluation module calculates the tumor progress rate and the complication risk coefficient based on the map, and generates a personalized risk prediction curve. And the multi-stage early warning module triggers an early warning signal according to the curve and generates a risk traceability report. And the adaptive control module generates a personalized scheme based on the early warning signal and the traceability report and controls the medical equipment to execute closed-loop feedback adjustment. The nursing quality and the treatment effect of tumor patients can be improved, and the medical risk is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of medical care, and more specifically, the present invention relates to an intelligent tumor care monitoring system. Background Art

[0002] In the field of medical care, especially in the care process of cancer patients, accurate monitoring and timely intervention are crucial for the recovery of patients. Existing tumor care monitoring methods mostly rely on single physiological parameter monitoring or regular imaging examinations. Although these methods can provide partial health information of patients to a certain extent, they have obvious limitations. For example, single physiological parameter monitoring cannot comprehensively reflect the complex pathological changes of tumors, and regular imaging examinations have problems such as long time intervals and inability to monitor in real time. In addition, traditional methods lack the ability to integrate and analyze multi-modal data, making it difficult to achieve dynamic assessment of tumor progression and personalized risk prediction. In practical applications, medical staff often need to rely on experience for care decisions, which may lead to untimely or excessive intervention, affecting the treatment effect and quality of life of patients.

[0003] In the process of implementing the embodiments of the present invention, the inventors found that there are at least the following problems or defects in the prior art: the existing monitoring system cannot synchronously collect multi-source biological data in real time, lacks the ability to deeply integrate and analyze multi-modal data, is difficult to achieve accurate dynamic assessment and personalized early warning, and cannot dynamically adjust the care plan and medical device parameters according to the real-time monitoring results, and cannot meet the complex and changeable care needs of cancer patients. Summary of the Invention

[0004] The present invention provides an intelligent tumor care monitoring system, including: A data acquisition module, configured to synchronously collect in real time the physiological sign data, local tumor imaging data, and body fluid biochemical data of a patient through a multi-source biological sensor, where the physiological sign data includes body surface temperature distribution, heart rate variability, and respiratory rhythm parameters; A data processing module, communicatively connected to the data acquisition module, configured to perform spatio-temporal alignment, noise suppression, and cross-modal feature association on the multi-modal data to generate a comprehensive feature map with a temporal dependence relationship; A dynamic assessment module, based on the comprehensive feature map, calculates the tumor progression rate and the complication risk coefficient through a dynamic deterioration probability model, and generates a personalized risk prediction curve by integrating the patient's historical treatment records; A multi-level early warning module, according to the slope change and threshold crossing event of the risk prediction curve, triggers a differential early warning signal and generates a multi-dimensional risk traceability report; The adaptive control module dynamically generates a personalized plan including drug dose adjustment, physical therapy parameter optimization, and nursing intervention timing based on the warning signal and the traceability report, and controls the medical device to perform closed-loop feedback regulation.

[0005] Further, the data acquisition module includes: A distributed physiological sensing unit, which consists of a flexible epidermal temperature sensor array, an implantable electrocardiogram monitoring chip, and a non-invasive blood component spectroscopic analyzer. The flexible epidermal temperature sensor array covers the area around the tumor in a honeycomb topology and generates a real-time temperature field distribution map; A tumor image enhancement unit, configured with a dynamic focus magnetic resonance imaging device, whose spatial resolution satisfies: where R is the spatial resolution, k is the inherent coefficient of the device, S / N is the signal-to-noise ratio, is the average temperature of the tumor area, is the tumor vascular density, m is the temperature sensitivity index, and n is the vascular density attenuation coefficient.

[0006] Further, the data processing module includes: A dynamic baseline calibration unit that performs sliding window normalization processing on physiological sign data and calculates the dynamic physiological fluctuation index: where, is the physiological fluctuation index, is the real-time physiological data, is the dynamic baseline value, is the absolute deviation weight coefficient, is the change rate sensitivity coefficient, is the non-linear adjustment factor; A cross-modal fusion unit that performs the following operations: Extracts periodic and trend components from physiological time series data and constructs a physiological feature matrix; Extracts texture features and vascular topological structures from tumor image data and constructs an image feature matrix; Performs feature interaction on the physiological feature matrix ( ) and the image feature matrix ( ) through a cross-attention weight matrix to generate a fused feature vector ( ), where represents the physiological feature matrix, represents the image feature matrix.

[0007] Further, the dynamic evaluation module includes: A deterioration probability prediction unit, based on the fused feature vector Calculate the dynamic deterioration probability of the tumor based on the temporal changes: Wherein, is the sigmoid activation function, , , are the model training parameters, represents the change rate of the feature vector, is the feature accumulation; The risk stratification unit calculates the comprehensive risk coefficient by combining the treatment response index Tr and the physiological fluctuation index Fb: Wherein, is the power-law coefficient, is the smoothing factor, is obtained by comparing the change rate of the features before and after treatment.

[0008] Furthermore, the multi-level early warning module includes: The first-level early warning trigger unit activates the primary alarm when the following composite conditions are met: Wherein, is the deterioration probability threshold, is the risk coefficient threshold, is the physiological fluctuation acceleration threshold; The second-level early warning trigger unit activates the advanced alarm when Condition1 is satisfied and there are abnormal imaging features: Wherein, is the imaging feature threshold, is the change rate threshold of the deterioration probability.

[0009] Furthermore, the adaptive control module includes: The dose optimization unit dynamically adjusts the drug dose based on the deviation between the risk coefficient and the target value : Wherein, is the adjusted dose, is the proportional adjustment coefficient, is the integral adjustment coefficient, is the preset safety probability threshold; The execution constraint unit controls the flow rate of the infusion pump to satisfy the non-linear response relationship: Wherein, is the flow velocity sensitivity coefficient, is the dose before adjustment.

[0010] Furthermore, the dynamic evaluation module further includes: A treatment feedback evaluation unit that calculates a treatment response index : where , are weight coefficients, and respectively represent the characteristic change rates before and after treatment, and are the physiological fluctuation indices before and after treatment.

[0011] Furthermore, the cross-modal fusion unit performs a bidirectional attention mechanism: where is the query matrix and key matrix of physiological features, is the query matrix and key matrix of image features, d is the feature dimension, and Concat represents the concatenation operation.

[0012] Furthermore, the dynamic baseline calibration unit includes: An abnormal data screening sub-unit that uses a mixed criterion to identify invalid data: If the real-time data exceeds the mean value of the sliding window ±3 times the standard deviation range; and at the same time exceeds the interquartile range of historical data ; then start redundant sensor cross-verification and mark the abnormal data channel.

[0013] Furthermore, the dose optimization unit includes: A taboo judgment sub-unit that constrains the dose adjustment to satisfy: and satisfies the pharmacokinetic safety condition: where is the blood drug concentration, AUC is the area under the curve, CL is the clearance rate, is the toxicity threshold, is the maximum dose increase coefficient.

[0014] The above embodiments of the present invention have at least the following beneficial effects: The intelligent tumor care monitoring system of the present invention can realize real-time synchronous acquisition and in-depth analysis of multimodal data of patients. Through cross-modal feature association and dynamic deterioration probability models, it can accurately calculate the tumor progression rate and complication risk coefficient, and generate personalized risk prediction curves. This multi-dimensional and dynamic evaluation method can provide comprehensive and real-time patient health information for medical staff, helping them more accurately grasp the changing trend of the patient's condition, so as to achieve precise and timely nursing intervention, and improve the nursing quality and treatment effect of tumor patients.

[0015] In addition, the system can also trigger differential warning signals and generate multi-dimensional risk traceability reports according to the slope change and threshold crossing events of the risk prediction curve, and then dynamically generate personalized plans including drug dose adjustment, physical therapy parameter optimization and nursing intervention timing, and control medical devices to perform closed-loop feedback regulation. This intelligent adaptive control function can effectively reduce medical risks, reduce the workload of medical staff, and at the same time improve the treatment experience and rehabilitation efficiency of patients, and has important clinical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the drawings, several embodiments of the present invention are shown by way of illustration and not limitation, wherein: Figure 1 It is a schematic structural diagram of an intelligent tumor care monitoring system provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are given only to enable those skilled in the art to better understand and then implement the present invention, and are not intended to limit the scope of the present invention in any way. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to be able to fully convey the scope of the present invention to those skilled in the art.

[0018] Those skilled in the art know that the embodiments of the present invention can be implemented as a system, device, equipment, method or computer program product. Therefore, the present invention can be specifically implemented in the following forms: completely hardware, completely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0019] It should be noted that any number of elements in the drawings is for illustration rather than limitation, and any naming is only for distinction and does not have any limiting meaning.

[0020] The following referenceFigure 1 , Figure 1 is a schematic structural diagram of an intelligent tumor care monitoring system provided by an embodiment of the present invention. As Figure 1 shown, an intelligent tumor care monitoring system 100 includes: A data acquisition module 101, configured to synchronously acquire physiological sign data, local tumor image data, and body fluid biochemical data of a patient in real time through a multi-source biosensor, where the physiological sign data includes body surface temperature distribution, heart rate variability, and respiratory rhythm parameters; A data processing module 102, communicatively connected to the data acquisition module, configured to perform spatio-temporal alignment, noise suppression, and cross-modal feature association on multi-modal data to generate a comprehensive feature map with a time series dependence relationship; A dynamic assessment module 103, based on the comprehensive feature map, calculates the tumor progression rate and complication risk coefficient through a dynamic deterioration probability model, and generates a personalized risk prediction curve by integrating the patient's historical treatment records; A multi-level warning module 104, according to the slope change and threshold crossing event of the risk prediction curve, triggers a differential warning signal and generates a multi-dimensional risk traceability report; An adaptive control module 105, based on the warning signal and traceability report, dynamically generates a personalized plan including drug dose adjustment, physical therapy parameter optimization, and nursing intervention timing, and controls medical devices to perform closed-loop feedback regulation.

[0021] It should be noted that the core of this system is to synchronously acquire multi-source biological data of patients in real time through the data acquisition module. These data include physiological sign data, local tumor image data, and body fluid biochemical data. Among them, physiological sign data such as body surface temperature distribution, heart rate variability, and respiratory rhythm parameters can reflect the overall health status of the patient; local tumor image data is obtained through high-resolution imaging technology and is used to observe the morphological and structural changes of tumors; body fluid biochemical data is obtained through non-invasive detection technology and is used to analyze the biochemical components in the patient's body fluid. The data processing module is communicatively connected to the data acquisition module and is responsible for processing the acquired multi-modal data, including spatio-temporal alignment, noise suppression, and cross-modal feature association, to generate a comprehensive feature map with a time series dependence relationship. These processing steps can ensure the accuracy and availability of the data and provide a reliable basis for subsequent dynamic assessment.

[0022] Specifically, the distributed physiological sensing unit in the data acquisition module is composed of a flexible epidermal temperature sensor array, an implantable electrocardiogram monitoring chip, and a non-invasive blood component spectral analyzer. The flexible epidermal temperature sensor array covers the peripheral area of the tumor in a honeycomb topology, generating a real-time temperature field distribution map, which can accurately monitor the temperature changes around the tumor and reflect the local blood circulation and metabolism. The tumor imaging enhancement unit is equipped with a dynamic focus magnetic resonance imaging device, and its spatial resolution is calculated by a formula, where the device intrinsic coefficient, signal-to-noise ratio, average temperature of the tumor area, tumor vascular density, temperature sensitivity index, and vascular density attenuation coefficient are the key parameters affecting the resolution. These parameters can be adjusted according to the specific tumor type and patient conditions to ensure high resolution and accuracy of the imaging. The dynamic baseline calibration unit in the data processing module performs sliding window normalization on the physiological sign data and calculates the dynamic physiological fluctuation index. This process can effectively calibrate the data, reduce noise interference, and ensure the stability and reliability of the data.

[0023] Preferably, the flexible epidermal temperature sensor array in the data acquisition module can be flexibly adjusted according to the size and location of the tumor to ensure the comprehensiveness and accuracy of the coverage. For example, for larger or irregularly shaped tumors, the number and distribution density of the sensors can be increased. In the data processing module, the cross-modal fusion unit performs feature interaction on the physiological feature matrix and the imaging feature matrix through a cross-attention weight matrix to generate a fused feature vector. This process can be further improved in terms of the accuracy and efficiency of feature extraction by introducing more feature extraction algorithms, such as convolutional neural networks (CNNs) in deep learning. In addition, the deterioration probability prediction unit in the dynamic evaluation module can combine more clinical data, such as the patient's age, medical history, etc., to further optimize the model parameters and improve the accuracy of the prediction.

[0024] In some embodiments, the data acquisition module includes: A distributed physiological sensing unit, which is composed of a flexible epidermal temperature sensor array, an implantable electrocardiogram monitoring chip, and a non-invasive blood component spectral analyzer. The flexible epidermal temperature sensor array covers the peripheral area of the tumor in a honeycomb topology and generates a real-time temperature field distribution map; A tumor imaging enhancement unit, which is equipped with a dynamic focus magnetic resonance imaging device, and its spatial resolution satisfies: where R is the spatial resolution, k is the device intrinsic coefficient, S / N is the signal-to-noise ratio, is the average temperature of the tumor area, is the tumor vascular density, m is the temperature sensitivity index, and n is the vascular density attenuation coefficient.

[0025] It should be noted that the data acquisition module in this system includes a distributed physiological sensing unit and a tumor imaging enhancement unit. The distributed physiological sensing unit consists of a flexible epidermal temperature sensor array, an implantable electrocardiogram monitoring chip, and a non-invasive blood component spectroscopic analyzer, which is used to collect the physiological sign data of patients in real time. Among them, the flexible epidermal temperature sensor array covers the area around the tumor in a honeycomb topology structure, and can generate a temperature field distribution map in real time, reflecting the temperature changes around the tumor. The tumor imaging enhancement unit is equipped with a dynamic focusing magnetic resonance imaging device, which is used to obtain high-resolution local tumor image data. Its spatial resolution is calculated by a specific formula, which comprehensively considers parameters such as the inherent coefficient of the device, signal-to-noise ratio, average temperature of the tumor area, tumor vascular density, temperature sensitivity index, and vascular density attenuation coefficient, to ensure the high resolution and accuracy of imaging.

[0026] Specifically, the flexible epidermal temperature sensor array adopts a honeycomb topology structure, which is similar to the hexagonal arrangement of a honeycomb, and can ensure the uniform distribution of the sensor array, so as to accurately monitor the temperature changes around the tumor and reflect the local blood circulation and metabolism. The implantable electrocardiogram monitoring chip is used to monitor the heart rate variability of patients in real time and provide key information on the heart health status. The non-invasive blood component spectroscopic analyzer detects the biochemical components in the patient's body fluid through spectroscopic analysis technology, without invasive operation, reducing the patient's pain and infection risk. In the tumor imaging enhancement unit, the spatial resolution of the dynamic focusing magnetic resonance imaging device is calculated by a specific formula, where the inherent coefficient k of the device is the performance parameter of the device itself, and the signal-to-noise ratio S / N reflects the clarity of the imaging signal, and the average temperature of the tumor area and the tumor vascular density are the pathological characteristic parameters of the tumor, and the temperature sensitivity index m and the vascular density attenuation coefficient n are the weight parameters used to adjust the resolution calculation. These parameters can be adjusted according to the specific tumor type and patient conditions to ensure the high resolution and accuracy of imaging.

[0027] Preferably, the flexible epidermal temperature sensor array can be flexibly adjusted according to the size and location of the tumor to ensure the comprehensiveness and accuracy of the coverage range. For example, for larger or irregularly shaped tumors, the number and distribution density of sensors can be increased to improve the accuracy of temperature monitoring. In the tumor imaging enhancement unit, the spatial resolution of the dynamic focusing magnetic resonance imaging device can be optimized by adjusting the inherent coefficient k of the device and the signal-to-noise ratio S / N. For example, by improving the signal processing ability of the imaging device, the signal-to-noise ratio can be increased, thereby improving the spatial resolution. In addition, the temperature sensitivity index m and the vascular density attenuation coefficient n can be adjusted according to the pathological characteristics of the tumor to better adapt to different types of tumors. For example, for tumors with a higher vascular density, the vascular density attenuation coefficient n can be appropriately increased to more accurately reflect the imaging characteristics of the tumor.

[0028] In some embodiments, the data processing module includes: A dynamic baseline calibration unit that performs sliding window normalization on physiological sign data and calculates a dynamic physiological fluctuation index: Where is the physiological fluctuation index, is the real-time physiological data, is the dynamic baseline value, is the absolute deviation weight coefficient, is the rate of change sensitivity coefficient, is the non-linear adjustment factor; A cross-modal fusion unit that performs the following operations: Extracts periodic and trend components from the physiological time series data and constructs a physiological feature matrix; Extracts texture features and vascular topological structures from the tumor image data and constructs an image feature matrix; Performs feature interaction on the physiological feature matrix ( ) and the image feature matrix ( ) through a cross-attention weight matrix to generate a fused feature vector ( ), where represents the physiological feature matrix, represents the image feature matrix.

[0029] It should be noted that the data processing module in this system includes a dynamic baseline calibration unit and a cross-modal fusion unit. The dynamic baseline calibration unit performs sliding window normalization on physiological sign data and calculates a dynamic physiological fluctuation index. This process can effectively calibrate the data, reduce noise interference, and ensure the stability and reliability of the data. The cross-modal fusion unit performs feature interaction on the physiological feature matrix and the image feature matrix through a cross-attention weight matrix to generate a fused feature vector. This process can deeply fuse data of different modalities and provide more comprehensive feature information for subsequent dynamic evaluation.

[0030] Specifically, when the dynamic baseline calibration unit performs sliding window normalization on physiological sign data, it calculates the dynamic physiological fluctuation index. This index is calculated through a specific formula, in which the difference between real-time physiological data and the dynamic baseline value is weighted, combined with the absolute deviation weight coefficient, the change rate sensitivity coefficient, and the non-linear adjustment factor, and can reflect the dynamic change trend of physiological data. The cross-modal fusion unit extracts periodic and trend components from physiological time-series data to construct a physiological feature matrix; it extracts texture features and vascular topological structures from tumor image data to construct an image feature matrix. Through the cross-attention weight matrix, the physiological feature matrix and the image feature matrix perform feature interaction to generate a fused feature vector. Among them, the physiological feature matrix and the image feature matrix respectively contain key features extracted from physiological data and image data. These features interact through the cross-attention mechanism and can better capture the correlation between different modal data.

[0031] Preferably, the size of the sliding window in the dynamic baseline calibration unit can be adjusted according to the sampling frequency and fluctuation characteristics of physiological data. For example, for physiological data with high-frequency sampling, a smaller sliding window can be set to capture short-term changes in the data more quickly; while for data with low-frequency sampling, a larger sliding window can be set to smooth the long-term trend of the data. In addition, the weights in the cross-attention weight matrix can be optimized through machine learning algorithms to better adapt to the data characteristics of different patients. For example, the attention mechanism network in deep learning can be used to dynamically adjust the weights, thereby improving the effect of feature interaction. During the cross-modal fusion process, other types of feature extraction methods can also be introduced, such as feature extraction based on wavelet transform, to further enrich the content of the feature matrix and improve the quality of the fused feature vector.

[0032] In some embodiments, the dynamic evaluation module includes: A deterioration probability prediction unit, based on the temporal variation of the fused feature vector to calculate the dynamic tumor deterioration probability: where, is the sigmoid activation function, , , are model training parameters, represents the change rate of the feature vector, is the feature accumulation; A risk stratification unit, combining the treatment response index T_r and the physiological fluctuation index F_b, to calculate the comprehensive risk coefficient: where, is the power-law coefficient, is the smoothing factor, obtained by comparing the change rates of features before and after treatment.

[0033] It should be noted that the dynamic assessment module is one of the core parts of this system. Its main function is to calculate the dynamic deterioration probability of tumors based on the temporal changes of the fused feature vectors, and combine the treatment response index and the physiological fluctuation index to calculate the comprehensive risk coefficient. The deterioration probability prediction unit calculates the dynamic deterioration probability of tumors through a specific model, and this probability reflects the possibility of tumor deterioration within a certain period of time. The risk stratification unit further refines the risk assessment according to the treatment response index and the physiological fluctuation index, generates the comprehensive risk coefficient, and provides a basis for the generation of personalized risk prediction curves. These assessment results can help medical staff better understand the changes in the patient's condition, so as to formulate more accurate treatment and nursing plans.

[0034] Specifically, when calculating the dynamic deterioration probability of tumors, the deterioration probability prediction unit will consider the change rate and cumulative amount of the fused feature vectors. These parameters are weighted and calculated through a specific model, and finally a probability value between 0 and 1 is obtained through an activation function, indicating the possibility of tumor deterioration. The risk stratification unit combines the treatment response index and the physiological fluctuation index to calculate the comprehensive risk coefficient through a power-law function. The treatment response index reflects the impact of treatment measures on the patient's condition, while the physiological fluctuation index represents the stability of the patient's physiological state. The specific settings of these parameters can be adjusted according to the individual differences of patients and clinical experience. For example, for certain specific types of tumors, it may be necessary to adjust the weights of the model training parameters to more accurately reflect their deterioration trends.

[0035] Preferably, the model training parameters in the deterioration probability prediction unit can be optimized through machine learning algorithms to improve the accuracy of prediction. For example, historical case data can be used to train the model, and the parameters can be adjusted through methods such as cross-validation to ensure the applicability of the model in different patient groups. In the risk stratification unit, the power-law coefficient and the smoothing factor can be set personalized according to the type of tumor and the physiological characteristics of the patient. For example, for tumors with a higher degree of malignancy, the value of the power-law coefficient can be appropriately increased to more sensitively reflect the risk changes. In addition, other risk assessment indicators, such as the patient's age, family medical history, etc., can be introduced to further enrich the calculation dimension of the comprehensive risk coefficient and make it more comprehensively reflect the patient's disease risk.

[0036] In some embodiments, the multi-level warning module includes: A first-level warning trigger unit that activates a primary alarm when the following composite conditions are met: wherein, is the deterioration probability threshold, is the risk coefficient threshold, is the physiological fluctuation acceleration threshold; The secondary warning trigger unit activates the high - level alarm when both Condition1 is satisfied and there are abnormal imaging features: Among them, is the imaging feature threshold, is the deterioration probability change rate threshold.

[0037] It should be noted that the multi - level warning module is a key part of this system for triggering warning signals according to the risk prediction curve generated by the dynamic assessment module. It includes a primary warning trigger unit and a secondary warning trigger unit, which are used to activate the primary alarm and the high - level alarm under different risk conditions respectively. The primary warning trigger unit activates the primary alarm when the specific deterioration probability threshold, risk coefficient threshold, and physiological fluctuation acceleration threshold are met, to remind medical staff to pay attention to the possible risks of patients. The secondary warning trigger unit, on the basis of meeting the primary warning conditions, further checks whether the imaging features are abnormal. If there are abnormalities, it activates the high - level alarm to indicate a more urgent risk situation. This hierarchical warning mechanism can provide warnings of different levels according to the severity of the risk, helping medical staff to take corresponding intervention measures in a timely manner.

[0038] Specifically, the deterioration probability threshold, risk coefficient threshold, and physiological fluctuation acceleration threshold in the primary warning trigger unit are parameters set according to clinical experience and historical data, and are used to judge the risk level of patients. For example, the deterioration probability threshold can be set to 0.7, indicating that when the probability of tumor deterioration reaches 70%, the primary alarm is triggered; the risk coefficient threshold can be set to 1.5, indicating that when the comprehensive risk coefficient exceeds 1.5, the alarm is triggered; the physiological fluctuation acceleration threshold can be set to 0.3, indicating that when the acceleration of physiological fluctuation exceeds 0.3, the alarm is triggered. The imaging feature threshold and the deterioration probability change rate threshold in the secondary warning trigger unit are also set according to clinical data, and are used to further confirm the severity of the risk. The imaging feature threshold can be set as the abnormal range of a specific texture or vascular structure parameter, while the deterioration probability change rate threshold indicates the rapid increase in the deterioration probability. The specific settings of these parameters need to be adjusted according to different tumor types and patient conditions to ensure the accuracy and timeliness of the warning.

[0039] Preferably, the threshold value in the primary warning trigger unit can be dynamically adjusted according to the individual differences of patients. For example, for elderly patients or patients with other underlying diseases, the threshold value can be appropriately reduced to trigger the alarm earlier and intervene in advance. In the secondary warning trigger unit, more imaging features can be introduced for comprehensive judgment, such as the volume change rate and edge sharpness of the tumor, to improve the accuracy of the advanced alarm. In addition, real-time clinical data, such as the changes in the patient's symptoms and laboratory test results, can be combined to further optimize the warning conditions. For example, when the patient's symptoms suddenly worsen, even if the imaging features have not reached the threshold value, the advanced alarm can be triggered to remind the medical staff to conduct examinations and treatments in a timely manner.

[0040] In some embodiments, the adaptive control module includes: A dose optimization unit that dynamically adjusts the drug dose based on the deviation between the risk coefficient and the target value : where is the adjusted dose, is the proportional adjustment coefficient, is the integral adjustment coefficient, is the preset safety probability threshold; An execution constraint unit that controls the flow rate of the infusion pump to satisfy the non-linear response relationship: where is the flow rate sensitivity coefficient, is the dose before adjustment.

[0041] It should be noted that the adaptive control module is a key part of this system for dynamically generating a personalized care plan and controlling the medical device to perform closed-loop feedback regulation according to the warning signals and risk traceability reports generated by the multi-level warning module. This module includes a dose optimization unit and an execution constraint unit. The dose optimization unit dynamically adjusts the drug dose based on the deviation between the risk coefficient and the target value, while the execution constraint unit controls the flow rate of the infusion pump to ensure the safety and effectiveness of drug treatment. This adaptive control mechanism can automatically adjust the care plan and medical device parameters according to the patient's real-time health status and risk assessment results, achieving precise personalized care.

[0042] Specifically, the dose optimization unit dynamically adjusts the drug dose according to the deviation between the risk coefficient and the target value. Among them, the risk coefficient is a comprehensive risk coefficient calculated by the dynamic evaluation module, reflecting the current health risk level of the patient; the target value is an expected risk coefficient preset according to the treatment goal of the patient. The dose adjustment formula includes a proportional adjustment coefficient and an integral adjustment coefficient, which are used to control the amplitude and speed of dose adjustment. The proportional adjustment coefficient determines the direct response intensity of dose adjustment, while the integral adjustment coefficient is used to eliminate the cumulative effect of deviation and ensure the stability of dose adjustment. The execution constraint unit controls the infusion pump flow rate according to the adjusted dose, and the flow rate sensitivity coefficient is used to adjust the response speed of the flow rate to dose changes, ensuring that the drug can be infused to the patient according to the predetermined dose and speed.

[0043] Preferably, the proportional adjustment coefficient and the integral adjustment coefficient in the dose optimization unit can be adjusted according to the individual differences of the patient and the drug characteristics. For example, for some patients with high drug sensitivity, the proportional adjustment coefficient can be appropriately reduced to avoid adverse reactions caused by too rapid dose adjustment. In the execution constraint unit, the flow rate sensitivity coefficient can be optimized according to the pharmacokinetic characteristics of the drug. For example, for drugs with a short half-life, the flow rate sensitivity coefficient can be appropriately increased to ensure that the drug can quickly reach the effective concentration. In addition, upper and lower limits for drug dose adjustment can be introduced to ensure the safety of dose adjustment. For example, according to the safe dose range of the drug and the physiological state of the patient, a maximum dose increase coefficient is set to avoid drug overdose. At the same time, real-time physiological monitoring data, such as blood drug concentration, can be combined to further optimize the dose adjustment strategy and ensure the safety and effectiveness of drug treatment.

[0044] In some embodiments, the dynamic evaluation module further includes: A treatment feedback evaluation unit that calculates a treatment response index : Wherein, 、 Are weight coefficients, And Respectively represent the characteristic change rates before and after treatment, And Are the physiological fluctuation indexes before and after treatment.

[0045] It should be noted that the treatment feedback evaluation unit in the dynamic evaluation module is a key part for calculating the treatment response index. The treatment response index is a quantitative indicator for evaluating the treatment effect by comparing the change rates of features before and after treatment and the change of the physiological fluctuation index. This indicator can help medical staff understand the patient's response to treatment measures, so as to dynamically adjust the subsequent treatment plan. In this way, the system can better adapt to the individual differences of patients and improve the accuracy and effectiveness of treatment.

[0046] Specifically, the calculation of the treatment response index involves the setting of multiple parameters. Among them, the weight coefficient is used to balance the influence of different feature change rates and physiological fluctuation indexes on the treatment effect. The change rate of features before and after treatment is calculated by comparing the physiological or imaging features before and after treatment, which reflects the specific impact of treatment on the patient's health status. The physiological fluctuation index reflects the stability of the patient's physiological state, and its change rate can be used to evaluate the regulatory effect of treatment on the patient's physiological state. The specific settings of these parameters can be adjusted according to different treatment goals and patient conditions. For example, for certain specific tumor treatments, more attention may be paid to the changes in imaging features, so the weight coefficient of the change rate of imaging features can be appropriately increased; for some treatments with a greater impact on the physiological state, the weight of the physiological fluctuation index can be increased.

[0047] Preferably, the calculation of the treatment response index can be further refined. For example, the weight coefficient can be dynamically adjusted according to the treatment stage. In the initial stage of treatment, more attention may be paid to the changes in the physiological fluctuation index, so its weight can be appropriately increased; in the later stage of treatment, the changes in imaging features may be more representative, so the weight of the change rate of imaging features can be increased. In addition, more feature change rates can be introduced, such as the change rate of body fluid biochemical indicators, to more comprehensively evaluate the treatment effect. For example, for some patients undergoing immunotherapy, the change rate of immune indicators can be included in the calculation to more accurately reflect the regulatory effect of treatment on the patient's immune system. At the same time, in order to improve the calculation accuracy, machine learning algorithms can be used to train and optimize the calculation model of the treatment response index, so that it can better adapt to the differences of different patients and treatment plans.

[0048] In some embodiments, the cross-modal fusion unit performs a bidirectional attention mechanism: Among them, is the query matrix and key matrix of physiological features, is the query matrix and key matrix of imaging features, d is the feature dimension, and Concat represents the concatenation operation.

[0049] It should be noted that the cross-modal fusion unit performs a bidirectional attention mechanism to further improve the effect of multi-modal data fusion. During the data processing, physiological features and imaging features are respectively extracted from different data sources, each of which contains unique information. The bidirectional attention mechanism can more effectively explore the correlation between the two features by allowing the two features to pay attention to each other, so as to generate a more representative fused feature vector. This mechanism helps to improve the accuracy of subsequent evaluation and early warning, because it can better integrate data from different modalities and provide a more comprehensive basis for the system's decision-making.

[0050] Specifically, the bidirectional attention mechanism involves a query matrix and a key matrix for physiological features, as well as a query matrix and a key matrix for imaging features. These matrices are obtained by transforming the original feature data, and the transformation process usually involves linear transformation or other mathematical operations. The dimension of the query matrix and the key matrix is an important parameter of the feature dimension, which determines the expression ability of the feature vector in space. During the calculation, the attention weights can be obtained through the dot product operation between the query matrix and the key matrix, and these weights reflect the degree of mutual correlation between different features. Subsequently, the weighted feature vectors are combined through a concatenation operation to form the final fused feature vector. This process can not only retain the information of the original features, but also enhance the interaction between the features, thus improving the fusion effect.

[0051] Preferably, the feature dimension in the bidirectional attention mechanism can be adjusted according to the requirements of actual applications. For example, when processing high-resolution image data, the feature dimension can be appropriately increased to better capture the detailed information in the image. At the same time, in order to improve the calculation efficiency, an approximate method can be used to calculate the attention weights, such as using a fast matrix multiplication algorithm. In addition, a multi-head attention mechanism can be introduced to decompose the feature vector into multiple subspaces, calculate the attention weights independently in each subspace, and then merge the results. This method can further improve the effect of feature fusion, because it allows the model to learn different feature interaction patterns in different subspaces. For example, when processing complex tumor images and physiological signals, the multi-head attention mechanism can better capture the complex relationship between the morphological features of the tumor and the physiological state, thus providing more accurate information for subsequent evaluation and early warning.

[0052] In some embodiments, the dynamic baseline calibration unit includes: An abnormal data screening sub-unit that identifies invalid data using a mixed criterion: If the real-time data exceeds the moving window mean ±3 times the standard deviation range; and at the same time exceeds the interquartile range of historical data ; Then start the redundant sensor cross - verification and mark the abnormal data channel.

[0053] It should be noted that the abnormal data screening sub - unit in the dynamic baseline calibration unit is a key part for identifying and processing invalid data. During the data acquisition process, due to sensor failures, signal interference, or other reasons, invalid or abnormal data may be generated. The abnormal data screening sub - unit identifies these invalid data through a combined criterion and starts the redundant sensor cross - verification to ensure the accuracy and reliability of the data. This process is crucial for improving the overall performance and stability of the system because it can effectively avoid misjudgments and wrong decisions caused by abnormal data.

[0054] Specifically, the combined criterion adopted by the abnormal data screening sub - unit includes two main conditions. First, if the real - time data exceeds the range of the moving window mean ± 3 times the standard deviation, then this data point may be abnormal. The moving window mean and standard deviation are statistical quantities calculated based on recent data and are used to reflect the normal fluctuation range of the data. Second, if the real - time data simultaneously exceeds the inter - quartile range [Q1 - 1.5IQR, Q3 + 1.5IQR] of the historical data, then this data point is further confirmed as abnormal data. The inter - quartile range (IQR) is a statistical quantity describing the degree of data dispersion, and Q1 and Q3 are the first quartile and the third quartile respectively. The combination of these two conditions can effectively identify those data points that significantly deviate from the normal range. Once abnormal data is identified, the system will start the redundant sensor cross - verification, confirm the validity of this data point by comparing the data of other sensors, and mark the abnormal data channel for subsequent processing.

[0055] Preferably, the size of the moving window in the abnormal data screening sub - unit can be adjusted according to the sampling frequency and fluctuation characteristics of the data. For example, for physiological data with high - frequency sampling, a smaller moving window can be set to detect abnormal data more quickly; while for data with low - frequency sampling, a larger moving window can be set to smooth the long - term trend of the data. In addition, the threshold of the inter - quartile range can also be optimized according to the specific application scenario. For example, in some applications with high requirements for data accuracy, the inter - quartile range can be appropriately narrowed to improve the sensitivity of abnormal data detection. At the same time, to further improve the efficiency of abnormal data processing, machine learning algorithms can be introduced to automatically identify and process abnormal data. For example, by training an anomaly detection model, this model can learn the pattern of normal data from historical data and thus more accurately identify abnormal data points.

[0056] In some embodiments, the dose optimization unit includes: A taboo judgment sub - unit that restricts the dose adjustment to satisfy: and satisfy the pharmacokinetic safety conditions: Wherein, is the blood drug concentration, AUC is the area under the curve, CL is the clearance rate, is the toxicity threshold, is the maximum dose increment coefficient.

[0057] It should be noted that the taboo judgment subunit in the dose optimization unit is a key part for restricting the adjustment of drug doses. During the process of drug dose adjustment, it is necessary to ensure that the adjusted dose can effectively respond to the risk changes of patients and will not exceed the safety range, so as to avoid drug overdose or underdose. The taboo judgment subunit ensures the safety and effectiveness of drug dose adjustment by setting the maximum dose increment coefficient and pharmacokinetic safety conditions. This process is crucial for ensuring the safety of patients' medication, because it can effectively avoid drug adverse reactions or poor treatment effects caused by improper dose adjustment.

[0058] Specifically, the maximum dose increment coefficient in the taboo judgment subunit is used to limit the maximum amplitude of single-dose adjustment. For example, if the current drug dose is , and the maximum dose increment coefficient is , then the adjusted dose shall not exceed . The pharmacokinetic safety conditions ensure that the adjusted dose conforms to the metabolism and excretion laws of the drug. For example, the blood drug concentration shall not exceed the toxicity threshold , and at the same time, the area under the curve (AUC) and clearance rate (CL) of the drug must also be within the safe range. The specific settings of these parameters need to be adjusted according to the characteristics of the drug, the physiological state of the patient, and the treatment goal. For example, for some patients with high drug sensitivity, the maximum dose increment coefficient can be appropriately reduced to ensure the safety of medication.

[0059] Preferably, the maximum dose increment coefficient in the taboo judgment subunit can be dynamically adjusted according to the therapeutic window of the drug and the individual differences of the patient. For example, for elderly patients or patients with other underlying diseases, the maximum dose increment coefficient can be appropriately reduced to reduce the risk of drug adverse reactions. In the pharmacokinetic safety conditions, the blood drug concentration data monitored in real time can be introduced, and the dose can be dynamically adjusted through a closed-loop feedback mechanism. For example, if the blood drug concentration monitored in real time is close to the toxicity threshold, the system can automatically reduce the dose adjustment amplitude to ensure the safety of medication. In addition, the clinical symptoms and laboratory test results of the patient can be combined to further optimize the dose adjustment strategy. For example, if the patient has impaired liver and kidney function, the clearance rate parameter can be appropriately adjusted to adapt to the patient's metabolic ability.

[0060] The above-mentioned various embodiments of the present invention have the following beneficial effects: The intelligent tumor care monitoring system of the present invention can comprehensively monitor the multimodal data of patients by collecting the physiological sign data, local tumor image data and body fluid biochemical data of patients in real time through multi-source biosensors, providing a rich data basis for subsequent accurate evaluation. The system performs spatio-temporal alignment, noise suppression and cross-modal feature correlation on the collected multimodal data to generate a comprehensive feature map with time-series dependence relationships, which can effectively improve the accuracy and usability of the data, providing a reliable basis for calculating the tumor progression rate and the complication risk coefficient. Based on the comprehensive feature map, the system calculates the tumor progression rate and the complication risk coefficient through a dynamic deterioration probability model, and generates a personalized risk prediction curve by integrating the patient's historical treatment records, which can realize the accurate dynamic evaluation of the patient's condition and provide scientific guidance for the formulation of personalized care plans. According to the slope change and threshold crossing events of the risk prediction curve, the system triggers differential warning signals and generates a multi-dimensional risk traceability report, which can timely detect potential risks, provide warning information for medical staff, and facilitate the adoption of timely intervention measures. Based on the warning signals and the traceability report, the system dynamically generates a personalized plan including drug dose adjustment, physical therapy parameter optimization and nursing intervention timing, and controls medical devices to perform closed-loop feedback regulation, which can realize accurate personalized nursing intervention and improve the treatment effect and the patient's recovery speed.

[0061] In terms of data acquisition, the system adopts distributed physiological sensing units and tumor imaging enhancement units, which can achieve comprehensive coverage and high-resolution imaging of the area around the tumor, providing high-quality data support for subsequent feature extraction and analysis. The dynamic baseline calibration unit and cross-modal fusion unit in the data processing module can effectively calibrate physiological sign data and achieve deep fusion of multi-modal data, further improving the accuracy and reliability of data processing. The deterioration probability prediction unit and risk stratification unit in the dynamic assessment module can accurately calculate the dynamic deterioration probability of the tumor and the comprehensive risk coefficient based on the temporal changes of the fused feature vectors and the treatment response index, providing more accurate assessment results for risk warning and the generation of personalized treatment plans. The primary warning trigger unit and secondary warning trigger unit of the multi-level warning module can trigger primary alarms and high-level alarms according to different risk thresholds and feature anomaly situations, providing hierarchical warning information for medical staff and facilitating their corresponding intervention measures according to the risk level. The dose optimization unit and execution constraint unit of the adaptive control module can dynamically adjust the drug dose according to the deviation between the risk coefficient and the target value and control the flow rate of the infusion pump to ensure the safety and effectiveness of drug treatment. In addition, the two-way attention mechanism of the treatment feedback assessment unit and the cross-modal fusion unit in the dynamic assessment module can further improve the evaluation accuracy of the treatment effect and the effect of multi-modal data fusion. The abnormal data screening sub-unit and the taboo judgment sub-unit can effectively identify and process abnormal data, ensuring the accuracy and safety of the data and avoiding misjudgment and incorrect intervention caused by data anomalies.

[0062] Furthermore, the storage medium of the embodiment of the present application stores program instructions capable of implementing all the above methods. Among them, the program instructions can be stored in the above storage medium in the form of a software product, including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods described in various embodiments of the present application. The aforementioned storage medium includes: various media that can store program codes such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs, or terminal devices such as computers, servers, mobile phones, and tablets.

[0063] The above description is only some preferred embodiments of the present invention and an explanation of the technical principles applied. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present invention is not limited to the technical solutions formed by the specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the technical solutions formed by mutually replacing the above features with the technical features (but not limited to) having similar functions disclosed in the embodiments of the present invention.

Claims

1. An intelligent tumor nursing monitoring system, characterized in that: include: A data acquisition module is used to synchronously acquire the patient's physiological sign data, tumor local imaging data and body fluid biochemical data in real time through multi-source biosensors. The physiological sign data includes body surface temperature distribution, heart rate variability and respiratory rhythm parameters; A data processing module, which is in communication with the data acquisition module and is used to perform spatiotemporal alignment, noise suppression, and cross-modal feature association on the multimodal data to generate a comprehensive feature map with temporal dependency; A dynamic assessment module, based on the comprehensive feature map, calculates the tumor progression rate and complication risk coefficient through a dynamic deterioration probability model, and integrates the patient's historical treatment records to generate a personalized risk prediction curve; A multi-level early warning module triggers differentiated early warning signals and generates a multi-dimensional risk tracing report based on the slope change of the risk prediction curve and the threshold crossing event; The adaptive control module dynamically generates a personalized plan including drug dosage adjustment, physical therapy parameter optimization and nursing intervention timing based on the early warning signal and traceability report, and controls the medical equipment to perform closed-loop feedback regulation.

2. The system according to claim 1, characterized in that The data acquisition module comprises: A distributed physiological sensing unit, which is composed of a flexible epidermal temperature sensor array, an implantable ECG monitoring chip, and a non-invasive blood component spectrometer. The flexible epidermal temperature sensor array covers the tumor surrounding area with a honeycomb topology structure and generates a temperature field distribution map in real time. The tumor image enhancement unit is equipped with a dynamic focusing magnetic resonance imaging device, and its spatial resolution meets the following requirements: Where R is the spatial resolution, k is the device intrinsic coefficient, S / N is the signal-to-noise ratio, is the average temperature of the tumor area, is the tumor vascular density, m is the temperature sensitivity index, and n is the vascular density attenuation coefficient.

3. The system according to claim 2, characterized in that The data processing module comprises: The dynamic baseline calibration unit performs sliding window normalization processing on the physiological sign data and calculates the dynamic physiological fluctuation index: in, is the physiological fluctuation index, For real-time physiological data, is the dynamic baseline value, is the absolute deviation weight coefficient, is the rate of change sensitivity coefficient, is the nonlinear adjustment factor; The cross-modal fusion unit performs the following operations: Extract periodic and trend components from physiological time series data and construct physiological feature matrix; Extract texture features and vascular topology from tumor image data and construct an image feature matrix; The physiological feature matrix ( ) and the image feature matrix ( ) to interact with features and generate fused feature vectors ( ),in represents the physiological feature matrix, Represents the image feature matrix.

4. The system according to claim 3, characterized in that The dynamic evaluation module includes: Deterioration probability prediction unit, based on fusion feature vector The temporal changes of , and the probability of dynamic tumor deterioration are calculated: in, is the sigmoid activation function, , , are the model training parameters, represents the rate of change of the eigenvector, is the characteristic accumulation; The risk stratification unit combines the treatment response index T_r and the physiological fluctuation index F_b to calculate the comprehensive risk coefficient: in, is the power law coefficient, is the smoothing factor, Obtained by comparing the rate of change of characteristics before and after treatment.

5. The system according to claim 4, characterized in that The multi-level early warning module includes: The primary warning trigger unit activates the primary alarm when the following composite conditions are met: in, is the deterioration probability threshold, is the risk factor threshold, is the physiological fluctuation acceleration threshold; The secondary warning trigger unit activates the advanced alarm when Condition 1 is met and there is an abnormal image feature: in, is the image feature threshold, is the deterioration probability change rate threshold.

6. The system according to claim 5, characterized in that The adaptive control module comprises: Dose optimization unit, based on risk factor With target value Deviations in the dosage of drugs can be adjusted dynamically: in, After adjusting the dose, is the proportional adjustment coefficient, is the integral adjustment coefficient, is a preset safety probability threshold; Execute constraint unit to control the flow rate of infusion pump Satisfies the nonlinear response relationship: in, is the velocity sensitivity coefficient, For pre-dose adjustment.

7. The system according to claim 6, characterized in that The dynamic evaluation module also includes: Treatment feedback evaluation unit, calculates treatment response index : in, , is the weight coefficient, and Represent the characteristic change rate before and after treatment, and It is the physiological fluctuation index before and after treatment.

8. The system according to claim 7, characterized in that The cross-modal fusion unit performs a bidirectional attention mechanism: in, are the query matrix and key matrix of physiological features, are the query matrix and key matrix of image features, d is the feature dimension, and Concat represents the concatenation operation.

9. The system according to claim 8, characterized in that The dynamic baseline calibration unit comprises: Abnormal data screening sub-unit, using mixed criteria to identify invalid data: If real-time data Exceeding the sliding window mean ±3 times the standard deviation scope; And at the same time it exceeds the interquartile range of historical data ; Redundant sensor cross-validation is then initiated and abnormal data channels are marked.

10. The system according to claim 9, characterized in that The dosage optimization unit comprises: Contraindication judgment subunit, constraining dose adjustment to meet: And meet the pharmacokinetic safety conditions: in, is the blood drug concentration, AUC is the area under the curve, CL is the clearance, is the toxicity threshold, is the maximum dose increase factor.

Images