Patient risk dynamic assessment system in medical care teaching

The anastomosis status evaluation system constructed by multimodal imaging technology and data processing module solves the subjectivity and inaccuracy of traditional evaluation methods, realizes the real-time generation of personalized nursing plans and teaching, and improves the quality of nursing.

CN120452801AActive Publication Date: 2025-08-08THE PEOPLES HOSPITAL SHAANXI PROV

Patent Information

Application Number
CN202510933276.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-07-08
Publication Date
2025-08-08
Estimated Expiration
2045-07-08

AI Technical Summary

Technical Problem

Traditional imaging technology is difficult to fully reflect the anastomosis status after gastric cancer. Medical staff have strong subjectivity in the assessment, lack of quantification and standardization, resulting in poor nursing plans, and traditional nursing teaching lacks interactivity and sense of reality.

Method used

Multimodal imaging technology (quantum dot fluorescence imaging, Raman spectroscopy imaging, CT tomography imaging and intraoperative high-definition white light imaging) is used for data acquisition, and a three-dimensional model is established in combination with the data processing module to build an anastomotic state evaluation model, generate a personalized intervention plan, and conduct real-time evaluation and feedback through the early warning decision module.

Benefits of technology

It improves the accuracy of anastomosis evaluation and the feasibility of nursing plans, enhances the authenticity and teaching effect of medical care, and reduces the incidence of postoperative complications.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120452801A_ABST
    Figure CN120452801A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of medical care, and provides a patient risk dynamic assessment system in medical care teaching, which comprises a data acquisition module used for acquiring multi-modal image data of a postoperative gastric organ anastomosis area of a patient; the data processing module is used for preprocessing the multi-modal image data, establishing a three-dimensional model of organs around the anastomotic stoma and analyzing edge features of the anastomotic stoma; and the risk assessment module is used for constructing an anastomotic stoma state assessment model and outputting comprehensive scores of an anastomotic stoma blood perfusion condition, a tissue metabolism abnormal probability, an anastomotic stoma leakage risk, anastomotic stoma scar hyperplasia and healing trend prediction. A quantum dot fluorescence image, a Raman spectrum image, a CT image and an intraoperative high-definition white light image are enhanced through the system and evaluated, a nursing scheme is obtained in combination with opinions of medical staff, and compared with manual evaluation of the medical staff, the accuracy is higher, the feasibility of the nursing scheme is higher, and medical nursing use is facilitated.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of medical care technology, and in particular to a patient risk dynamic assessment system in medical care teaching. Background Art

[0002] In the healthcare field, accurate patient risk assessment and personalized care plans are crucial for improving care quality and ensuring patient safety. This is especially true for anastomotic care after gastric cancer surgery, where traditional approaches present numerous challenges.

[0003] In the imaging assessment process, the previous single-modality imaging technology was unable to fully reflect the anastomotic condition, and various types of images were often interfered by noise, resulting in large errors in the extraction of anastomotic features by medical staff, which in turn affected risk assessment. In the risk assessment stage, traditional methods mostly relied on the subjective experience of medical staff, lacked quantification and standardization, and had limitations such as large subjective differences, low specificity, ignoring the relationship between indicators, and inability to conduct real-time dynamic assessment. As a result, in the formulation of nursing plans, standardized plans were mostly adopted in the past, and individual differences of patients were not fully considered, resulting in poor nursing effects and a high incidence of postoperative complications. In addition, traditional nursing teaching mainly imparts experience through nursing records, lacking realism and interactivity. It is difficult for medical students to deeply understand the logic and basis behind the nursing plan, which is not conducive to cultivating their clinical decision-making ability. Summary of the Invention

[0004] In view of the problems existing in the prior art, the purpose of the present invention is to provide a patient risk dynamic assessment system in medical care teaching to solve the problems raised by the above background technology.

[0005] To achieve the above objectives, the present invention provides a patient risk dynamic assessment system in medical care teaching, comprising: The data acquisition module is used to obtain multimodal imaging data of the patient's postoperative gastric organ anastomosis area, including quantum dot fluorescence imaging, Raman spectroscopy imaging, CT tomography imaging, and intraoperative high-definition white light imaging, as well as the patient's vital signs data; The data processing module is used to pre-process the multimodal imaging data, establish a three-dimensional model of the organs around the anastomosis, and analyze the characteristics of the anastomotic edge; The risk assessment module is used to build an anastomotic status assessment model and output a comprehensive score for anastomotic blood perfusion, the probability of abnormal tissue metabolism, the risk of anastomotic leakage, anastomotic scar hyperplasia, and healing trend prediction; The early warning decision module is used to classify risk levels based on risk assessment results, generate early warning information, and generate personalized intervention plans based on current patient vital signs data; The interactive feedback module is used to organize the acquired data, generated personalized plans, and actual treatment effects into data sets, establish a teaching case library, and create virtual operation training scenarios.

[0006] Preferably, the data acquisition module acquires multimodal image data including the following steps: S11. Establish a model for estimating the anastomotic length of patients undergoing gastrectomy surgery, inject targeted quantum dot probes before surgery, and calculate the probe dose based on the patient's weight; S12. Collect the patient's quantum dot fluorescence images, Raman spectroscopy images, CT scan images, and intraoperative high-definition white light images according to the surgical process; S13. Add a timestamp, patient ID, and acquisition device information to each frame of image and upload it to the database for storage.

[0007] Preferably, the data processing module pre-processes the collected image data including the following steps: S21. De-noising the quantum dot fluorescence image and performing baseline correction on the Raman spectrum image; S22. Perform three-dimensional reconstruction of the CT tomographic images to generate a triangular mesh model of the anastomosis, and perform contrast equalization on the high-definition white light images during the operation to enhance the details of the anastomotic suture. S23, calculating the fluorescence intensity distribution of the anastomotic site, extracting the edge features of the anastomotic degree, and analyzing the anastomotic condition; S24. Construct anastomotic feature vectors, register multimodal images, and fuse the extracted geometric features, intensity features, and texture features.

[0008] Preferably, the risk assessment module constructs an anastomotic state assessment model to output a comprehensive score, comprising the following steps: S31. Input the processed multimodal image features, perform feature extraction and transformation layer by layer on the processed multimodal image features, mine the image deep features, and provide high-dimensional, abstract image representation for subsequent risk assessment; S32. Treat the geometric, strength, and texture features of the anastomosis as graph nodes, transfer and update features based on the attention mechanism between nodes, capture the dependencies between features, strengthen the weights of effective features, and improve the accuracy of risk prediction; S33. Establish models for anastomotic blood perfusion, tissue metabolic abnormality probability, anastomotic leakage risk, anastomotic scar hyperplasia, and healing trend prediction, statistically integrate the output results of each model, and calculate the comprehensive risk value of anastomotic condition after gastric cancer surgery.

[0009] Preferably, in step S33, calculating the comprehensive risk value of the anastomotic condition after gastric cancer surgery includes the following steps: S331. Fit the probability distribution of each model output result to accurately characterize its uncertainty and fluctuation pattern, distinguish the model reliability, and optimize the comprehensive scoring fusion logic; S332. Quantify the reliability of each model's calculation results by calculating the entropy of the probability distribution output by the module, providing a basis for rationally allocating module weights when comprehensively assessing anastomotic risk; S333. By allocating weights using normalized entropy weights and combining distribution mean and variance penalties, we focus on reliable model output, suppress fluctuation interference, and accurately and comprehensively evaluate the anastomotic risk after gastric cancer surgery.

[0010] Preferably, in step S12, the order of collecting image data is to use a high-definition white light camera to collect detailed images of the patient's anastomotic suture during the operation, use a quantum dot fluorescence imager to collect anastomotic microvascular perfusion images within 24 hours after the operation, use a Raman spectrometer to collect anastomotic tissue metabolism image within 24-48 hours, and the scanning range covers the anastomosis and the surrounding 2 cm tissue. CT tomography is performed within 48-72 hours, and the layer thickness is set to 0.5 mm.

[0011] Preferably, the early warning decision module generates early warning information and personalized intervention plans including the following steps: S41. Set an early warning threshold for anastomotic-related complications. When the real-time assessment result exceeds the threshold, trigger a multi-channel early warning based on the early warning priority. S42. Conduct risk tracing analysis, tracing back the root causes of anastomotic complications after gastric cancer surgery from abnormal indicators, and assist in making precise nursing decisions. S43. Generate a risk propagation path diagram to show the causal relationship between abnormal indicators and automatically adjust the monitoring frequency according to the warning level; S44. Synchronize the warning information to the electronic medical record system, mark it as a priority, generate a personalized intervention plan, and upload it to the database for storage; S45. Analyze the generated personalized plan and use the state transition probability, immediate reward and discount factor to determine the anastomotic state value under different nursing intervention actions; S46. Adjust the intervention plan based on the feedback from medical staff and record the intervention effects for model updating.

[0012] Preferably, in step S44, the generation of the personalized plan includes controlling the sustained-release dosage of the local drug at the anastomosis site, a nutritional support formula, and an early activity plan.

[0013] The patient risk dynamic assessment system in medical care teaching provided by the present invention has the following beneficial effects: By enhancing quantum dot fluorescence images, Raman spectroscopy images, CT tomography images, and intraoperative high-definition white light images through the system and evaluating them, the nursing plan derived from the opinions of medical staff is more accurate and more feasible than manual evaluation by medical staff, which is beneficial for medical care use. In addition, this system can be used for simulation teaching, which is more realistic than traditional medical staff reading relevant nursing records, and it is easier for medical staff to understand the intention of the nursing plan and evaluate the nursing plan generated by medical staff, which is beneficial for medical teaching use. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0015] Figure 1 Schematic diagram of the system modules of the patient risk dynamic assessment system in medical care teaching provided in this application. DETAILED DESCRIPTION

[0016] The following embodiments of the present invention are described in further detail in conjunction with the accompanying drawings and examples. The following embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention.

[0017] like Figure 1 As shown, this embodiment proposes a patient risk dynamic assessment system in medical care teaching, including: The data acquisition module is used to obtain multimodal imaging data of the patient's postoperative gastric organ anastomosis area, including quantum dot fluorescence imaging, Raman spectroscopy imaging, CT tomography imaging, and intraoperative high-definition white light imaging, as well as the patient's vital signs data; The data processing module is used to pre-process the multimodal imaging data, establish a three-dimensional model of the organs around the anastomosis, and analyze the characteristics of the anastomotic edge; The risk assessment module is used to build an anastomotic status assessment model and output a comprehensive score for anastomotic blood perfusion, the probability of abnormal tissue metabolism, the risk of anastomotic leakage, anastomotic scar hyperplasia, and healing trend prediction; The early warning decision module is used to classify risk levels based on risk assessment results, generate early warning information, and generate personalized intervention plans based on current patient vital signs data; The interactive feedback module is used to organize the acquired data, generated personalized plans, and actual treatment effects into data sets, establish a teaching case library, and create virtual operation training scenarios.

[0018] The present invention systematically enhances quantum dot fluorescence images, Raman spectroscopy images, CT tomography images, and intraoperative high-definition white light images, and evaluates and obtains a nursing plan based on the opinions of medical staff. Compared with manual evaluation by medical staff, the nursing plan has higher accuracy and higher feasibility, which is beneficial for medical care use. In addition, this system can be used for simulation teaching, which is more realistic than traditional medical staff reading relevant nursing records, and makes it easier for medical staff to understand the intention of the nursing plan and evaluate the nursing plan generated by the medical staff, which is beneficial for medical teaching use.

[0019] In this embodiment, the data acquisition module acquires multimodal image data including the following steps: S11. Establish a model for estimating the anastomotic length of patients undergoing gastrectomy surgery. The formula is: Where, is the estimated anastomotic length, Clinical characteristic variables The clinical characteristic variables include but are not limited to the patient's BMI, age, maximum tumor diameter, gastric tissue resection volume, and intraoperative anastomosis coefficient. The manual anastomosis coefficient is 1, and the stapler anastomosis coefficient is 0.7. To estimate the anastomotic blood perfusion area, To estimate the mean fluorescence intensity of the anastomosis, is the tumor volume, is the gastrectomy ratio, 、 、 as well as As the model parameters, targeted quantum dot probes were injected before surgery, and the probe dose was calculated according to the patient's weight using the formula: , where For dosage, is the patient's weight; S12. Collect the patient's quantum dot fluorescence images, Raman spectroscopy images, CT scan images, and intraoperative high-definition white light images according to the surgical process; S13. Add a timestamp, patient ID, and acquisition device information to each frame of image and upload it to the database for storage.

[0020] In this embodiment, in step S12, the order of collecting image data is to use a high-definition white light camera to collect detailed images of the patient's anastomotic suture during the operation, use a quantum dot fluorescence imager to collect anastomotic microvascular perfusion images within 24 hours after the operation, and use a Raman spectrometer to collect anastomotic tissue metabolism image within 24-48 hours. The scanning range covers the anastomosis and the surrounding 2 cm tissue. CT tomography is performed within 48-72 hours, and the layer thickness is set to 0.5 mm.

[0021] Specifically, a multi-dimensional data collection system was constructed for the evaluation of anastomosis after gastric cancer surgery. By integrating the estimation model of patient characteristics, tumor load and blood supply status, the anastomosis length estimation error was controlled within a high-precision range, accurately matching the pathophysiological process of the anastomosis.

[0022] In this embodiment, the data processing module pre-processes the collected image data including the following steps: S21. De-noise the quantum dot fluorescence image and perform baseline correction on the Raman spectrum image. The formula is: , where Pixels The denoised value of and, Pixels The original value of is the similarity weight, and are the corresponding normalization factors, Pixels Belong to the neighborhood pixels, For the original signal The wavelet coefficients obtained after continuous wavelet transform are is the wavelet basis function The complex conjugate of is the scale parameter, is the translation parameter, is the integration variable, is the differential unit; S22. Perform three-dimensional reconstruction on the CT tomographic images to generate a triangular mesh model of the anastomosis. Equalize the contrast of the high-definition white light images during the operation to enhance the details of the anastomosis. The formula is: , where is the three-dimensional coordinate of the surface intersection point located on the line connecting two vertices in the CT volume data obtained by linear interpolation, and Vertex and The three-dimensional coordinates of is the threshold set during the thresholding process of CT volume data, and They are and The original grayscale value of CT at position, and The coordinates of the white light image before and after equalization are The grayscale value of the pixel, is the inverse of the cumulative distribution function, The gray value in the image block belongs to The number of pixels; S23. Calculate the fluorescence intensity distribution of the anastomotic site, extract the edge features of the anastomotic degree, and analyze the anastomotic condition. The formula is: , where and are the filtered weak edges and strong edges respectively, is the proportionality coefficient, and , and are the image grayscale mean and standard deviation, is the adjustment factor, and , The fluorescence intensity is The probability distribution when is the number of Gaussian distributions, 、 and Respectively The mixing coefficient, mean, and covariance matrix of the Gaussian distribution; S24. Construct anastomotic feature vectors, register multimodal images, and fuse the extracted geometric features, intensity features, and texture features. The formula is: , where is the final fusion decision result, is the number of types of multimodal images involved in fusion, For the The predicted probability obtained after analyzing the image features of the modalities, For the The weight coefficients of the modalities in decision fusion.

[0023] Specifically, through multimodal image preprocessing and feature fusion, the reliability and analysis depth of anastomotic assessment data have been significantly improved. By improving the signal-to-noise ratio of fluorescence images, controlling the baseline drift error of Raman spectroscopy, and performing three-dimensional reconstruction and enhanced white light imaging of CT, the surface error of the anastomotic triangular mesh model is made extremely small. The overall process achieves full-chain optimization from image denoising to feature abstraction, providing high-dimensional, low-noise quantitative features for the risk assessment model, making the subsequent comprehensive scoring more accurate.

[0024] In this embodiment, the risk assessment module constructs an anastomotic state assessment model to output a comprehensive score, including the following steps: S31. Input the processed multimodal image features, perform feature extraction and transformation layer by layer on the processed multimodal image features, mine the image depth features, and provide high-dimensional, abstract image representation for subsequent risk assessment. The formula is: , where For the Layer feature maps, is the set of input feature maps, For the Layer input feature maps, is the convolution kernel, is the bias value, is the activation function; S32. The geometric, strength, and texture features of the anastomosis are regarded as graph nodes. The features are transferred and updated based on the attention mechanism between nodes to capture the dependencies between features, strengthen the weights of effective features, and improve the accuracy of risk prediction. The formula is: , where For nodes The output features of For nodes The set of neighbor points of For nodes For Node The attention coefficient, is the weight matrix, is the parameter vector of the attention mechanism, It is a multimodal feature; S33. Establish models for anastomotic blood perfusion, tissue metabolic abnormality probability, anastomotic leakage risk, anastomotic scar hyperplasia, and healing trend prediction, statistically integrate the output results of each model, and calculate the comprehensive risk value of anastomotic condition after gastric cancer surgery.

[0025] Specifically, the quantitative model formula for anastomotic blood perfusion is: , where is the blood perfusion index, is the peak systolic blood flow velocity, is the resistance index, is the hemoglobin concentration, is the anastomotic vessel diameter, is the number of days after surgery, 、 、 、 、 For the corresponding weight coefficient, the probability model formula for tissue metabolic abnormality is: , where is the probability of metabolic abnormality, is the linear predictor, is the lactate level, is the interstitial fluid pH, is the venous oxygen saturation, is the glucose concentration, C-reactive protein, 、 、 、 、 is the corresponding regression coefficient, and the anastomotic leakage risk scoring model formula is: , where Score the risk, The number of risks, is the risk weight, As the risk variable, the anastomotic scar hyperplasia prediction model formula is: , where is the scar hyperplasia index, Transforming growth factor, is the ratio of type I to type III collagen, For wound tension, is the patient's age, A history of diabetes, 、 、 、 、 is the corresponding specific gravity coefficient.

[0026] In this embodiment, in step S33, calculating the comprehensive risk value of the anastomotic condition after gastric cancer surgery includes the following steps: S331. Fit the probability distribution of each model output result to accurately characterize its uncertainty and fluctuation pattern, distinguish the model reliability, and optimize the comprehensive scoring fusion logic. The formula is: , where For the The output of the model, is the symbol of normal distribution, and Respectively The output mean and variance of each model; S332. Quantify the reliability of each model's calculation results by calculating the entropy of the probability distribution output by the module, and provide a basis for the reasonable allocation of module weights when comprehensively evaluating anastomotic risk. The formula is: , where For the The information entropy of the model output is For the Model output results The probability density function of For the Model output results The integral variable of ; S333. By allocating weights based on normalized entropy weights and combining distribution mean and variance penalties, we can focus on reliable model output, suppress fluctuation interference, and accurately and comprehensively assess the anastomotic risk after gastric cancer surgery. The formula is: , where is the comprehensive risk score of anastomosis obtained by integrating the information of each model. is the number of models, For the The entropy of the model output distribution, and , For the The mean of the model output distribution, is the variance penalty coefficient, For the The variance of the model output distribution.

[0027] Specifically, a high-precision anastomotic risk assessment system was constructed through multi-layer feature extraction, graph attention mechanism and multi-model fusion. By mining the deep features of images layer by layer, the correlation between the extracted fluorescence intensity, texture and other indicators and anastomotic fistula was analyzed. The attention mechanism was used to improve the accuracy of capturing the dependency between geometric features, and the weight of key features such as abnormal blood perfusion was strengthened to make the Kappa value of the comprehensive risk score highly consistent with the clinical outcome. The overall process realizes intelligent evaluation from feature abstraction to uncertainty quantification, improves the prediction accuracy of complications such as anastomotic leakage and scar hyperplasia, and provides a reliable quantitative basis for personalized intervention.

[0028] In this embodiment, the early warning decision module generates early warning information and personalized intervention plans, including the following steps: S41. Set the warning threshold for anastomosis-related complications. When the real-time assessment result exceeds the threshold, trigger a multi-channel warning based on the warning priority. The formula is: , where For warning priority, is the leakage risk probability, For blood perfusion score, is the metabolic abnormality index, 、 and are the corresponding weight coefficients; S42. Conduct risk tracing analysis, tracing back the root causes of anastomotic complications after gastric cancer surgery from abnormal indicators, and assist in making precise nursing decisions. S43. Generate a risk propagation path diagram to show the causal relationship between abnormal indicators and automatically adjust the monitoring frequency according to the warning level; S44. Synchronize the warning information to the electronic medical record system, mark it as a priority, generate a personalized intervention plan, and upload it to the database for storage; S45. Analyze the generated personalized plan and use the state transition probability, immediate reward, and discount factor to determine the anastomotic state value under different nursing intervention actions. The formula is: , where is the anastomotic status value, is the state transition probability, For timely rewards, is the discount factor, To perform nursing interventions After the current state The next state to transition to; S46. Adjust the intervention plan based on the feedback from medical staff and record the intervention effects for model updating.

[0029] In this embodiment, in step S44, the generation of the personalized plan includes the control of the local drug sustained-release dosage at the anastomotic site, the nutritional support formula, and the early activity plan, and the formula is: , where and are the final dosage and basic dosage of the local sustained-release drug at the anastomosis site, is the proportion of protein in the nutritional support formula, is the albumin level test value, is the early activity intensity level of patients after gastric cancer surgery, is the boundary limit function.

[0030] Specifically, by integrating leakage risk, perfusion score and metabolic abnormalities into quantitative indicators, the warning response time is shortened compared to manual warnings. Through risk tracing and transmission path maps, medical staff can quickly locate the root causes of complications and improve the accuracy of causal relationship identification. The generated personalized intervention plan is dynamically optimized through the reinforcement learning framework, which improves the effectiveness of personalized nursing plans, reduces the incidence of serious postoperative complications, and improves the efficiency of nursing decision-making.

[0031] The above embodiments are intended to illustrate the present invention only and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the embodiments, it should be understood by those skilled in the art that various combinations, modifications, or equivalent substitutions of the technical solutions of the present invention do not depart from the spirit and scope of the technical solutions of the present invention and should be encompassed by the scope of the claims of the present invention.

Claims

1. A patient risk dynamic assessment system in medical nursing teaching, characterized by: include: The data acquisition module is used to obtain multimodal imaging data of the patient's postoperative gastric organ anastomosis area, including quantum dot fluorescence imaging, Raman spectroscopy imaging, CT tomography imaging, and intraoperative high-definition white light imaging, as well as the patient's vital signs data; The data processing module is used to pre-process the multimodal imaging data, establish a three-dimensional model of the organs around the anastomosis, and analyze the characteristics of the anastomotic edge; The risk assessment module is used to build an anastomotic status assessment model and output a comprehensive score for anastomotic blood perfusion, probability of abnormal tissue metabolism, anastomotic leakage risk, anastomotic scar hyperplasia, and healing trend prediction; The early warning decision module is used to classify risk levels based on risk assessment results, generate early warning information, and generate personalized intervention plans based on current patient vital signs data; The interactive feedback module is used to organize the acquired data, generated personalized plans, and actual treatment effects into data sets, establish a teaching case library, and create virtual operation training scenarios.

2. The patient risk dynamic assessment system in medical care teaching according to claim 1 is characterized in that: The data acquisition module acquires multimodal image data including the following steps: S11. Establish a model for estimating the anastomotic length of patients undergoing gastrectomy surgery, inject targeted quantum dot probes before surgery, and calculate the probe dose based on the patient's weight; S12. Collect the patient's quantum dot fluorescence images, Raman spectroscopy images, CT scan images, and intraoperative high-definition white light images according to the surgical process; S13. Add a timestamp, patient ID, and acquisition device information to each frame of image and upload it to the database for storage.

3. The patient risk dynamic assessment system in medical care teaching according to claim 1 is characterized in that: The data processing module pre-processes the collected image data, including the following steps: S21. De-noising the quantum dot fluorescence image and performing baseline correction on the Raman spectrum image; S22. Perform three-dimensional reconstruction of the CT tomographic images to generate a triangular mesh model of the anastomosis, and perform contrast equalization on the high-definition white light images during the operation to enhance the details of the anastomotic suture. S23, calculating the fluorescence intensity distribution of the anastomotic site, extracting the edge features of the anastomotic degree, and analyzing the anastomotic condition; S24. Construct anastomotic feature vectors, register multimodal images, and fuse the extracted geometric features, intensity features, and texture features.

4. The patient risk dynamic assessment system in medical care teaching according to claim 1 is characterized in that: The risk assessment module constructs an anastomotic state assessment model to output a comprehensive score, comprising the following steps: S31. Input the processed multimodal image features, perform feature extraction and transformation layer by layer on the processed multimodal image features, mine the image deep features, and provide high-dimensional, abstract image representation for subsequent risk assessment; S32. Treat the geometric, strength, and texture features of the anastomosis as graph nodes, transfer and update features based on the attention mechanism between nodes, capture the dependencies between features, strengthen the weights of effective features, and improve the accuracy of risk prediction; S33. Establish models for anastomotic blood perfusion, tissue metabolic abnormality probability, anastomotic leakage risk, anastomotic scar hyperplasia, and healing trend prediction, statistically integrate the output results of each model, and calculate the comprehensive risk value of anastomotic condition after gastric cancer surgery.

5. The patient risk dynamic assessment system in medical care teaching according to claim 4 is characterized in that: In step S33, calculating the comprehensive risk value of the anastomotic condition after gastric cancer surgery includes the following steps: S331. Fit the probability distribution of each model output result to accurately characterize its uncertainty and fluctuation pattern, distinguish the model reliability, and optimize the comprehensive scoring fusion logic; S332. Quantify the reliability of each model's calculation results by calculating the entropy of the probability distribution output by the module, providing a basis for rationally allocating module weights when comprehensively assessing anastomotic risk; S333. By allocating weights using normalized entropy weights and combining distribution mean and variance penalties, we focus on reliable model output, suppress fluctuation interference, and accurately and comprehensively evaluate the anastomotic risk after gastric cancer surgery.

6. The patient risk dynamic assessment system in medical care teaching according to claim 2 is characterized in that: In step S12, the order of collecting image data is to use a high-definition white light camera to collect detailed images of the patient's anastomotic suture during the operation, use a quantum dot fluorescence imager to collect anastomotic microvascular perfusion images within 24 hours after the operation, use a Raman spectrometer to collect anastomotic tissue metabolism image within 24-48 hours, and the scanning range covers the anastomosis and the surrounding 2 cm tissue. CT tomography is performed within 48-72 hours, and the layer thickness is set to 0.5 mm.

7. The patient risk dynamic assessment system in medical care teaching according to claim 1 is characterized in that: The early warning decision module generates early warning information and personalized intervention plans, including the following steps: S41. Set an early warning threshold for anastomotic-related complications. When the real-time assessment result exceeds the threshold, trigger a multi-channel early warning based on the early warning priority. S42. Conduct risk tracing analysis, tracing back the root causes of anastomotic complications after gastric cancer surgery from abnormal indicators, and assist in making precise nursing decisions. S43. Generate a risk propagation path diagram to show the causal relationship between abnormal indicators and automatically adjust the monitoring frequency according to the warning level; S44. Synchronize the warning information to the electronic medical record system, mark it as a priority, generate a personalized intervention plan, and upload it to the database for storage; S45. Analyze the generated personalized plan and use the state transition probability, immediate reward and discount factor to determine the anastomotic state value under different nursing intervention actions; S46. Adjust the intervention plan based on the feedback from medical staff and record the intervention effects for model updating.

8. The patient risk dynamic assessment system in medical care teaching according to claim 7 is characterized in that: In step S44, the generation of a personalized plan includes controlling the sustained-release dosage of local drugs at the anastomosis site, nutritional support formula, and early activity plan.

Citation Information

Patent Citations

  • Fluorescent camera based on indocyanine green, and application of fluorescent camera

    CN111887826A

  • Disease risk analysis method and system based on painless gastrointestinal endoscope and storage medium

    CN118522448A

  • System and method for evaluating potential complications of department of cardiology

    CN120199470A

  • Method and apparatus for implementing digital video modeling to generate a patient risk assessment model

    US20090005650A1

  • Systems and methods for predicting and detecting post-operative complications

    US20240006075A1

Cited By

  • Thoracic surgery postoperative complication prediction method and system

    CN120636824A

  • Vein thrombosis risk assessment method based on large language model

    CN121281734A

  • Evaluation method and system for improving postoperative sleep of elderly patient through nerve block analgesia

    CN121439268A

  • Complication risk assessment system for complex liver occupying lesion resection under body cold perfusion

    CN122455376A