Multidisciplinary collaborative system and method for differentiated nutritional management of cancer patients
By integrating a multidisciplinary collaborative, differentiated nutrition management system for cancer patients, the problem of malnutrition risk management during radiotherapy has been solved. This system enables proactive prediction of radiation-induced mucosal damage and timely implementation of nutritional interventions, thereby improving the safety of radiotherapy and the effectiveness of the management system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANGHAI TONGREN HOSPITAL
- Filing Date
- 2026-04-16
- Publication Date
- 2026-07-10
AI Technical Summary
The existing clinical management system cannot effectively predict radiation-induced mucosal damage and malnutrition caused by dysphagia during radiotherapy in elderly patients with malignant tumors. It suffers from problems such as outdated screening tools, fragmented information in multidisciplinary collaboration, rigid gaps in monitoring frequency, and lack of hardware constraints for intervention implementation, resulting in distorted risk management and safety hazards.
A differentiated nutrition management system for cancer patients, integrating multidisciplinary collaboration, is adopted, including a risk prediction module, a collaborative audit module, an adaptive frequency diagnosis module, and a mandatory closed-loop logic lock module. The system calculates the predicted slope of nutritional decline through a prediction model, quantifies the entropy of the collaborative process, dynamically adjusts the monitoring frequency, and forcibly suspends radiotherapy access in high-risk states, forming a rigid closed loop.
It enables proactive prediction of radiotherapy risks, precise correction of risk severity, dynamic adjustment of monitoring frequency, and timely implementation of nutritional interventions, thereby improving radiotherapy safety and the effectiveness of the management system.
Smart Images
Figure SMS_21 
Figure SMS_52 
Figure QLYQS_1
Abstract
Description
Technical Field
[0001] This invention belongs to the field of smart healthcare and clinical risk management. Specifically, it relates to a differentiated nutrition management system and method for cancer patients that integrates multidisciplinary collaboration. Background Technology
[0002] During radiotherapy cycles in elderly patients with malignant tumors (such as head and neck cancer, esophageal cancer, etc.), radiation-induced mucosal damage and severe malnutrition due to dysphagia are the core factors forcing radiotherapy interruption. Current clinical management often incorporates Failure Mode and Effects Analysis (HFMEA) models for risk control, but significant technical deficiencies exist in its practical implementation with information technology.
[0003] Screening tools lack foresight: static screening scales cannot detect the cumulative damage to the gastrointestinal mucosa caused by high-risk physical doses in the radiotherapy planning system (TPS), resulting in a serious lag in early warning.
[0004] The multidisciplinary collaboration process is "black boxed": information from doctors, nurses, patients, and their families is highly fragmented, and traditional systems cannot quantify the information loss rate during the interaction process, resulting in distorted risk detection.
[0005] There is a rigid gap in the monitoring sampling frequency: due to the limitations of manual scheduling, the detection of biochemical indicators is often set at a fixed baseline frequency (such as once a week), which cannot cover the "outbreak period" of radiotherapy side effects after a specific dose threshold;
[0006] The lack of hardware-level constraints in intervention execution: Traditional systems have become soft prompts that "only alarm but do not block", lacking a rigid closed-loop mechanism that strongly binds the "high-risk non-intervention state" with the "physical radiotherapy execution" at the underlying level.
[0007] Therefore, a rigid technological closed loop is needed, encompassing risk prediction, dynamic monitoring, and mandatory enforcement, to maximize the safety of radiotherapy. Summary of the Invention
[0008] To address the above deficiencies, this invention provides a differentiated nutrition management system for cancer patients that integrates multidisciplinary collaboration, comprising:
[0009] Risk simulation module: used to obtain patients' radiotherapy plan dose data and historical biochemical indicators, build a simulation model to calculate the slope of nutritional decline prediction, and adjust the risk severity factor based on the slope;
[0010] Collaborative Audit Module: Used to collect interactive log data of multi-terminal clinical collaboration, generate process entropy by quantifying the disorder of system collaboration, and correct the risk detection factor based on the process entropy;
[0011] Adaptive frequency diagnosis module: used to dynamically calculate the target sampling frequency of nutritional indicators based on the predicted slope and process entropy, and output a critical control point out of control command when it is determined that the current actual sampling interval exceeds the frequency requirement;
[0012] Forced closed-loop logic lock module: When the out-of-control command is received, it changes the lock bit of the system state machine, suspends the corresponding radiotherapy execution permission, and unlocks the permission after listening to the preset nutritional intervention data packet.
[0013] Furthermore, the pre-simulation model in the risk pre-simulation module is a long short-term memory network based on an attention mechanism, and its state update follows the following relationship:
[0014] in, Let the vector representing the patient's current biochemical indicators serve as the system state variables of the model. The cumulative radiation dose distribution matrix characterizes the nonlinear perturbation of biological tissues by the radiotherapy physical dose. The weight parameters are trained based on historical queues. This is the nonlinear mapping function of the pre-model, which characterizes the model's internal mechanism of using long short-term memory cell state gating units and attention weight allocation to adjust the input variables. With disturbance variables This involves a nonlinear computational process for feature extraction and state transition.
[0015] Furthermore, the collaborative audit module defines key medical behaviors (such as education delivery and vital sign recording) in the clinical nutrition management pathway as discrete events. The probability that it will be completed within the time window is The system calculates process entropy using the following formula. :
[0016]
[0017] Furthermore, the adaptive frequency diagnostic module dynamically calculates the target sampling frequency. The model is:
[0018]
[0019] in, As the reference sampling frequency, The system's preset sensitivity gain coefficient, To simulate the model, the downward prediction slope of biochemical indicators is extracted based on the output curve. This represents the instantaneous increment of process entropy.
[0020] Furthermore, the adaptive frequency diagnostic module also includes a sliding window verification unit, used to calculate the actual sampling interval between the current time and the last valid detection time point. When the following inequality is satisfied (i.e., the actual sampling interval) Greater than the target sampling frequency When the countdown begins, the system triggers and outputs a loss-of-control command for the critical control point, forcibly assigning the highest penalty weight to the detectivity value in the risk model.
[0021]
[0022] Furthermore, the forced closed-loop logic lock module includes a three-state finite state machine logic lock controller. When a loss of control command is received, the logic lock is forcibly closed and a radiotherapy suspension command is sent to the underlying radiotherapy execution device to cut off the physical execution permission of radiotherapy.
[0023] This invention also discloses a closed-loop control method for radiotherapy nutrition applied to the above-mentioned system, comprising the following steps:
[0024] S1. By integrating physical radiation dose and biochemical characteristics through a pre-model, a forward-looking nutritional decline prediction slope is output.
[0025] S2. Audit the interaction logs of multidisciplinary information systems to quantify the entropy of collaborative processes;
[0026] S3. Utilize the prediction slope and process entropy to dynamically increase the target sampling frequency and implement sliding window verification to intercept monitoring vacuum periods;
[0027] S4. In a state where the assessment is high-risk and intervention measures are lacking, the physical radiotherapy privileges are forcibly suspended through a logic lock until the feedback data packet of the intervention closed loop is obtained.
[0028] Compared with the prior art, the present invention has the following advantages:
[0029] 1. By introducing the underlying physical parameters of radiotherapy (such as spatial radiation dose) as a perturbation into the pre-model, the predicted slope of nutritional decline is output. This mechanism significantly advances the risk detection window, realizing a leap from passive response to active pre-modeling, and effectively corrects the severity of the risk.
[0030] 2. By discretizing the probability calculation of multi-terminal interaction behavior logs, the originally subjective and unpredictable risk of communication failure is transformed into an objective "process entropy". This provides an extremely accurate mathematical correction basis for the "detectability" factor, which is easily distorted in risk management.
[0031] 3. It overcomes the missed detection defects of the fixed detection frequency in the traditional approach. By coupling the prediction slope and process entropy to dynamically calculate the target sampling frequency, it saves medical resources during the system's stable period and exponentially increases the detection frequency during periods of sudden parameter changes, accurately intercepting any potential risks that deviate from the monitoring time window;
[0032] 4. By deploying finite state machine logic locks in the system, high-risk states are strongly bound to radiotherapy execution permissions, thereby achieving the mandatory constraint that "radiotherapy cannot be unlocked without intervention feedback data packets," and increasing the nutritional intervention execution rate to 100%. Detailed Implementation
[0033] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.
[0034] Example 1
[0035] This embodiment provides a differentiated nutrition management system for cancer patients that integrates multidisciplinary collaboration, including the following four modules:
[0036] Module 1: Risk Pre-simulation Module, designed to overcome the limitations of static screening, uses digital twins to build a forward-looking assessment system, including the following units:
[0037] Multidimensional data acquisition unit: Through the underlying interface, it automatically extracts the cumulative radiation dose distribution matrix (such as maximum radiation dose and high-risk volume percentage) of patients from the radiotherapy planning system (TPS), and extracts the radiation hotspots of the target area and organs at risk; it also simultaneously connects to the laboratory information system (LIS) to dynamically acquire biochemical flux data such as prealbumin and transferrin.
[0038] An integrated long short-term memory network shadow patient model (pre-model) based on an attention mechanism uses the physical dose of radiotherapy as an external perturbation and biochemical indicators as system state variables. The operating formula is as follows:
[0039]
[0040] in, Let the vector representing the patient's current biochemical indicators serve as the system state variables of the model. The cumulative radiation dose distribution matrix characterizes the nonlinear perturbation of biological tissues by the radiotherapy physical dose. The weight parameters are trained based on historical queues. This is the nonlinear mapping function of the pre-model, which characterizes the model's internal mechanism of using long short-term memory cell state gating units and attention weight allocation to adjust the input variables. With disturbance variables This involves a nonlinear computational process for feature extraction and state transition.
[0041] It should be noted that the pre-model uses the extracted cumulative radiation dose distribution matrix as a high-dimensional spatial feature input, i.e., it defines... It is a feature matrix that includes the spatial distribution of irradiated hotspots, used to quantify the instantaneous perturbation of biochemical state by physical irradiation dose.
[0042] The pre-simulation model calculates and outputs the nutrient loss prediction curve for the next 72 hours, and extracts the downward prediction slope of key biochemical indicators. The system's built-in HFMEA management logic will As a corrective input, the risk severity score (S) is dynamically increased; simultaneously, the adaptive frequency diagnostic module calls this in real time. The value, combined with the instantaneous increment of process entropy output by the collaborative audit module, is used as the target sampling frequency for calculation. The core variables are: the HFMEA model is responsible for the macro-level qualitative assessment and early warning of risk levels, while the adaptive frequency diagnosis module is responsible for the micro-level quantitative assessment and linkage of execution frequency (HFMEA model (Medical Failure Mode and Effects Analysis) is the existing basic framework for hospital risk management).
[0043] It should be noted that the dynamic adjustment of this severity score (S) provides a basic risk benchmark for the process entropy calculation in the collaborative audit module, and is synchronously read by the adaptive frequency diagnostic module as the trigger threshold for runaway (minimum effective monitoring frequency). The weighting factor is used to achieve full-link coupling of risk simulation results to the execution end.
[0044] Module Two: Collaborative Audit Module. This module uses information theory algorithms to quantify the coherence of the doctor-patient collaboration chain and includes the following units:
[0045] Interaction Log Auditing Unit: The system captures the digital footprint timestamps of the interaction logs of the four parties involved (medical staff, nurses, caregivers, and families) across multiple platforms in real time (such as the delay in family members uploading dietary logs, the duration of reading feedback on doctors' educational tasks, etc.), and uses this data to compare and calculate the aforementioned interaction nodes. The probability of completing within the specified time window ;
[0046] Process entropy calculation unit: Defines key medical behaviors (such as education and vital sign recording) in the clinical nutrition management pathway as interaction nodes. The probability of successfully completing a handshake within the system's specified effective time window is: The system calculates the process entropy of the current collaborative flow in real time. :
[0047]
[0048] When information gaps occur in clinical nutrition management pathways, the probability... Attenuation leads to process entropy in the system. Rapid surge indicates that the system has fallen into a state of high disorder (disorder is divided into high, medium and low). The system directly uses the instantaneous increment brought by the high entropy value as the correction factor of the detectivity (D) in the HFMEA model.
[0049] Module 3: Adaptive Frequency Diagnostic Module, used to eliminate fatal missed detections caused by fixed monitoring cycles and determine the status of critical control points (CCPs), including the following units:
[0050] Dynamic sampling frequency factor modeling unit: used for real-time monitoring of the absolute value of the downslope of biochemical prediction and the instantaneous increment of process entropy. Dynamically calculate the minimum effective monitoring frequency required to maintain system safety at the current moment. To ensure the minimum required effective monitoring frequency is maintained when indicators decline rapidly or coordination becomes chaotic. It exhibits an exponential leap;
[0051]
[0052] in, As the reference sampling frequency, The system's preset sensitivity gain coefficient, To simulate the model, the downward prediction slope of biochemical indicators is extracted based on the output curve. This represents the instantaneous increment of process entropy.
[0053] The monitoring failure determination logic unit is used to calculate the actual sampling interval between the current time and the last valid detection time. The system continuously performs the following inequality checks:
[0054]
[0055] If this inequality holds, that is, the actual sampling interval Greater than the target sampling frequency - i.e., the minimum effective monitoring frequency When the reciprocal of (i.e.) (That is, the maximum tolerance period allowed by the algorithm), the system immediately triggers a "detection frequency insufficient" failure signal;
[0056] Module 4: Forced Closed-Loop Logic Lock Module, used to change the lock bit of the system state machine upon receiving a runaway command, suspend the corresponding radiotherapy execution permission, and unlock the execution permission after listening for a preset nutritional intervention data packet. It includes the following units:
[0057] State machine logic lock controller: includes unlocked state (Green), warning state (Amber), and locked state (Red). When the third step determines that the CCP is seriously out of control and there is no matching intervention order, the state machine logic lock controller switches to the locked state.
[0058] Locking trigger unit: After entering the locked state, the state machine logic lock controller forcibly writes a lock bit signal to the system and sends a radiotherapy suspension command to the underlying radiotherapy execution equipment, completely cutting off the physical execution authority of radiotherapy for the day and popping up a window requesting medical intervention;
[0059] Automatic Unlocking Unit: This unit incorporates a cross-system intervention feature recognition engine. The system automatically activates a monitoring mode for the coupled database layer of the RIS (Radiotherapy Information System) and HIS (Hospital Information System). The engine captures the medical order change flow of the HIS system in real time through preset database shadow triggers and matches the business feature fields with the built-in "Safety Intervention Dictionary Table" using feature codes. Simultaneously, the engine executes cross-system correlation verification logic, checking the consistency between the patient ID in the intervention data packet and the logic lock ID, as well as the validity of the intervention timestamp, to determine if the unlocking conditions are met. Only after successful verification does the state machine logic lock controller automatically transition back to the unlocked state and issue an unlock command.
[0060] Example 2
[0061] This embodiment is a method based on the system of Embodiment 1, including the following steps:
[0062] S1. By integrating physical radiation dose and biochemical characteristics through a pre-model, a forward-looking nutritional decline prediction slope is output.
[0063] S2. Audit the interaction logs of multidisciplinary information systems to quantify the entropy of collaborative processes;
[0064] S3. Utilize the prediction slope and process entropy to dynamically increase the target sampling frequency and implement sliding window verification to intercept monitoring vacuum periods;
[0065] S4. In a state where the assessment is high-risk and intervention measures are lacking, the physical radiotherapy privileges are forcibly suspended through a logic lock until the feedback data packet of the intervention closed loop is obtained.
[0066] Example 3
[0067] This embodiment is based on the system in Embodiment 1 above, and uses the method in Embodiment 2 to perform nutritional management on patient Li (72 years old, esophageal squamous cell carcinoma):
[0068] P1 (Pre-training Trigger): During the 18th radiotherapy session, the system extracted TPS data and found a sharp increase in the target area radiation dose. The pre-training model (Attention-LSTM) ingested the cumulative radiation dose distribution matrix. Subsequently, it was calculated that the albumin level would rapidly drop below 130 mg / L within the next 72 hours, and the downward prediction slope of the biochemical indicators at this time was extracted. The value is -0.45, which the system judges as extremely high risk. The system proactively adjusts the risk severity score (S) from the default of 5 to 9.
[0069] P2 (Collaborative Collapse Detection): The system's backend probe detected that the family had not uploaded their dietary information to the client for 50 consecutive hours, and the nurse had not recorded vital signs. The system calculated the interaction node. completion probability A sudden drop leads to an increase in process entropy. The disordered state of the system is captured by quantization;
[0070] P3 (Monitoring Network Collapse and Loss of Control Determination): The adaptive frequency diagnosis module substitutes the predicted slope extracted from P1 above with... =-0.45 and Calculate the minimum sampling frequency required at this moment (i.e., the target sampling frequency). ≈1.35 times / day; sliding window verification revealed that the patient's last blood draw was 4 days ago (i.e. (day), then determine Upon successful establishment, the system immediately sounds an alarm and adjusts the correction factor in the HFMEA model. Pushed to 10 points, the CCP declares itself seriously out of control;
[0071] P4 (Hardcore Interception and Closed-Loop Unlocking): The following morning, when the radiotherapy technician was preparing to perform radiotherapy in the RIS system, the physical equipment was suspended due to the forced closure of the logic lock. The system initiated monitoring of the RIS and HIS coupled database layer through the intervention feature recognition engine in the automatic unlocking unit. This engine detected an ONS prescription record generated by the HIS system in real time, parsed its feature code to match a valid intervention item in the dictionary table, and confirmed through primary key verification that the prescription belonged to patient Li and was generated due to the closed logic lock. After the verification passed, the state machine automatically switched back to the unlocked state and released the execution lock in the RIS system, allowing the radiotherapy to be performed smoothly.
[0072] It should be noted that the structure described in this invention can be implemented in many different forms and is not limited to the embodiments described. Any equivalent transformations made by those skilled in the art based on the content of this specification, or direct or indirect applications in other related technical fields, such as the loading and unloading of other items, are included within the protection scope of this invention.
Claims
1. A differentiated nutrition management system for cancer patients integrating multidisciplinary collaboration, characterized in that: include: Risk simulation module: used to obtain patients' radiotherapy plan dose data and historical biochemical indicators, build a simulation model to calculate the slope of nutritional decline prediction, and adjust the risk severity factor based on the slope; Collaborative Audit Module: Used to collect interactive log data of multi-terminal clinical collaboration, generate process entropy by quantifying the disorder of system collaboration, and correct the risk detection factor based on the process entropy; Adaptive frequency diagnosis module: used to dynamically calculate the target sampling frequency of nutritional indicators based on the predicted slope and process entropy, and output a critical control point out of control command when it is determined that the current actual sampling interval exceeds the frequency requirement; Forced closed-loop logic lock module: When the out-of-control command is received, it changes the lock bit of the system state machine, suspends the corresponding radiotherapy execution permission, and unlocks the permission after listening to the preset nutritional intervention data packet.
2. The differentiated nutrition management system for cancer patients integrating multidisciplinary collaboration as described in claim 1, characterized in that: The pre-simulation model in the risk pre-simulation module is a long short-term memory network based on an attention mechanism, and its state update follows the following relationship: in, Let the vector representing the patient's current biochemical indicators serve as the system state variables of the model. The cumulative radiation dose distribution matrix characterizes the nonlinear perturbation of biological tissues by the radiotherapy physical dose. The weight parameters are trained based on historical queues. This is the nonlinear mapping function of the pre-model.
3. The differentiated nutrition management system for cancer patients integrating multidisciplinary collaboration as described in claim 2, characterized in that: The collaborative auditing module defines key medical behaviors in the clinical nutrition management pathway as discrete events. The probability that it will be completed within the time window is The system calculates process entropy using the following formula. :
4. The differentiated nutrition management system for cancer patients integrating multidisciplinary collaboration as described in claim 3, characterized in that: The adaptive frequency diagnostic module dynamically calculates the target sampling frequency. The model is: in, As the reference sampling frequency, The system's preset sensitivity gain coefficient, To simulate the model, the downward prediction slope of biochemical indicators is extracted based on the output curve. This represents the instantaneous increment of process entropy.
5. The differentiated nutrition management system for cancer patients integrating multidisciplinary collaboration as described in claim 4, characterized in that: The adaptive frequency diagnostic module also includes a sliding window verification unit for calculating the actual sampling interval between the current time and the last valid detection time. When the following inequality is satisfied, the system triggers and outputs a loss-of-control command for the critical control point, forcibly assigning the highest penalty weight to the detectivity value in the risk model:
6. The differentiated nutrition management system for cancer patients integrating multidisciplinary collaboration as described in claim 5, characterized in that: The forced closed-loop logic lock module includes a three-state finite state machine logic lock controller. When a loss of control instruction is received, the logic lock is forcibly closed and a device suspension instruction is sent to cut off the physical execution permission for radiotherapy.
7. A closed-loop control method for radiotherapy nutrition applied to the system described in any one of claims 1-6, characterized in that, Includes the following steps: S1. By integrating physical radiation dose and biochemical characteristics through a pre-model, a forward-looking nutritional decline prediction slope is output. S2. Audit the interaction logs of multidisciplinary information systems to quantify the entropy of collaborative processes; S3. Utilize the prediction slope and process entropy to dynamically increase the target sampling frequency and implement sliding window verification to intercept monitoring vacuum periods; S4. In a state where the assessment is high-risk and intervention measures are lacking, the physical radiotherapy privileges are forcibly suspended through a logic lock until the feedback data packet of the intervention closed loop is obtained.