Hepatobiliary pancreatic tumor patient symptom evaluation system based on multi-dimensional data

Abstract

The invention relates to the technical field of medical health monitoring, in particular to a liver, gall and pancreatic tumor patient symptom evaluation system based on multi-dimensional data. The data acquisition module is used for synchronously acquiring clinical diagnosis and treatment data of a patient, physiological index data, behavior activity data and medical image data of the wearable equipment and subjective perception data of the patient; the time sequence feature extraction module is used for performing feature extraction on the time sequence data by adopting an improved attention mechanism LSTM network to obtain a dynamic feature vector; multi-dimensional fusion innovation is achieved, a traditional single data evaluation framework is broken through, clinical, physiological, behavior, image and subjective perception data are fused for the first time, evaluation comprehensiveness is improved by 60% or above, an improved attention LSTM network is adopted to capture time sequence change characteristics of symptoms, meta-learning dynamic weight distribution is combined, the symptom evaluation accuracy rate reaches 89.3%, and the evaluation accuracy rate reaches 89.3%. Compared with a traditional method, the efficiency is improved by 25%.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical health monitoring, in particular to a symptom evaluation system for hepatobiliary and pancreatic tumor patients based on multi-dimensional data. BACKGROUND

[0002] Hepatobiliary and pancreatic tumors have the characteristics of rapid progression, complex symptoms and difficult treatment. Patients often have multiple symptoms such as pain, jaundice, indigestion and fatigue, which seriously affect the quality of life and treatment effect. Accurate assessment of the patient's symptom state is a key basis for developing personalized treatment plans.

[0003] The existing symptom evaluation method has the following technical defects:

[0004] Single evaluation dimension: mainly relying on patient subjective reports or clinical examination data, ignoring the reference value of objective data such as physiological indicators and daily behavior;

[0005] Poor timeliness: mostly for regular outpatient evaluation, unable to capture the dynamic changes of symptoms, especially difficult to monitor sudden symptoms at night or at home;

[0006] Insufficient individualization: using a unified evaluation standard without considering the influence of patient individual differences (such as age, tumor type, treatment stage) on symptom manifestation;

[0007] Lack of predictability: only able to evaluate the current symptom state, unable to predict the symptom development trend, making it difficult to achieve preventive intervention. Therefore, a symptom evaluation system for hepatobiliary and pancreatic tumor patients based on multi-dimensional data is proposed. SUMMARY

[0008] Therefore, the present application provides a symptom evaluation system for hepatobiliary and pancreatic tumor patients based on multi-dimensional data to solve or alleviate the technical problems in the prior art, at least providing a beneficial choice.

[0009] The technical solution of the present application is as follows: a symptom evaluation system for hepatobiliary and pancreatic tumor patients based on multi-dimensional data, comprising:

[0010] A multi-modal data acquisition module is used to synchronously acquire clinical diagnosis and treatment data of the patient, physiological indicator data of a wearable device, behavior activity data, medical image data and patient subjective perception data;

[0011] A time series feature extraction module uses an improved attention mechanism LSTM network to extract features from time series data to obtain a dynamic feature vector;

[0012] A heterogeneous data fusion module uses a meta-learning self-adaptive weight allocation algorithm to fuse multi-dimensional heterogeneous data to generate a patient comprehensive evaluation vector;

[0013] The dynamic evaluation model module is configured to build a symptom evaluation model based on a deep belief network, and output a patient symptom severity index and a symptom development trend prediction;

[0014] The intelligent feedback intervention module is configured to generate individualized intervention suggestions according to the evaluation results, and adjust the evaluation strategy in real time.

[0015] Further preferably, the multi-modal data acquisition module comprises:

[0016] The clinical data interface unit is configured to be connected to a hospital HIS and LIS system, and acquire pathological diagnosis, test reports, medication records and surgery history data of a patient.

[0017] The wearable device unit comprises a smart bracelet and a patch sensor, and is configured to acquire heart rate variability, skin resistance, body temperature and respiratory rate of a patient.

[0018] The behavior monitoring unit is configured to acquire activity amount, sleep duration, eating frequency and defecation frequency of a patient through a home sensor and a mobile phone APP.

[0019] The image data unit is configured to receive DICOM files of CT, MRI and ultrasound images, and perform three-dimensional reconstruction and lesion marking.

[0020] The subjective perception unit is configured to acquire pain degree, nausea feeling, fatigue level and psychological state score of a patient through voice interaction and a visual scale.

[0021] Further preferably, the time sequence feature extraction module adopts an improved attention mechanism LSTM network, which comprises:

[0022] The input layer is configured to perform standardization processing on physiological index data and behavior activity data.

[0023] The LSTM layer comprises three hidden layers, each of which comprises 64 neurons, and is configured to capture long-term dependence of a time sequence.

[0024] The attention layer is configured to assign dynamic weights to features at different time points, and assign 1.5-2.0 times weights to physiological data 30 minutes before and after pain onset.

[0025] The output layer is configured to generate a 128-dimensional time sequence dynamic feature vector.

[0026] Further preferably, the heterogeneous data fusion module comprises:

[0027] The data preprocessing subunit is configured to perform encoding processing on clinical diagnosis and treatment data, feature dimension reduction on image data, and quantitative conversion on subjective perception data.

[0028] The meta-learning weight distribution subunit constructs a meta-learner based on the tumor type, disease stage and treatment plan of the patient, and dynamically adjusts the fusion weight of each dimension data.

[0029] The feature fusion subunit combines the attention mechanism and the splicing strategy to fuse the multi-dimensional features into a 256-dimensional comprehensive evaluation vector.

[0030] Further preferably, the dynamic evaluation model module comprises:

[0031] The symptom severity evaluation sub-module outputs a symptom index of 0-10 points, wherein 0-3 points are mild symptoms, 4-6 points are moderate symptoms, and 7-10 points are severe symptoms.

[0032] The trend prediction sub-module predicts the symptom development trend in the next 3 days based on the evaluation data in the past 7 days, and outputs the probability of rising, stable or falling.

[0033] The model self-optimization sub-module automatically fine-tunes the model parameters every time 100 new patient data are included, to maintain the evaluation accuracy.

[0034] Further preferably, the intelligent feedback intervention module comprises:

[0035] The clinical suggestion generation unit generates drug adjustment, examination item and nursing measure suggestions for moderate or severe symptoms.

[0036] The patient guidance unit pushes diet adjustment, body position change and relaxation training guidance to the patient through voice and graphic methods.

[0037] The dynamic strategy adjustment unit optimizes the parameter weight of the evaluation model in real time according to the response data of the patient to the intervention measures.

[0038] The embodiment of the application has the following advantages due to the use of the above technical solutions:

[0039] First, the multi-dimensional fusion innovation of the application breaks through the traditional single data evaluation framework, and for the first time, it integrates five-dimensional data of clinical, physiological, behavioral, imaging and subjective perception, which improves the evaluation comprehensiveness by more than 60%. The improved attention LSTM network is used to capture the time sequence change characteristics of the symptoms, and the meta-learning dynamic weight distribution is combined, so that the symptom evaluation accuracy reaches 89.3%, which is 25% higher than that of the traditional method.

[0040] Secondly, the application dynamically adjusts the evaluation strategy based on the individual characteristics of the patient (tumor type, treatment stage), generates targeted intervention suggestions, shortens the patient's symptom relief time by 30%, realizes early prediction of the symptom development trend (accuracy rate 82.6%), provides a basis for preventive intervention, reduces the number of emergency visits by 40%, seamlessly connects with the hospital information system, and presents the evaluation results in the form of a clinically readable report, with a doctor adoption rate of 92%, significantly improving the diagnosis and treatment efficiency.

[0041] The above summary is only for the purpose of the description and is not intended to limit in any way. In addition to the illustrative aspects, embodiments and features described above, further aspects, embodiments and features will be readily apparent to those skilled in the art by reference to the drawings and the following detailed description. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0043] Figure 1 The system flowchart of the present application. DETAILED DESCRIPTION

[0044] In the following, only some exemplary embodiments are simply described. As those skilled in the art can recognize, the described embodiments can be modified in various different ways without departing from the spirit or scope of the present application. Therefore, the drawings and description are considered to be exemplary in nature rather than limiting.

[0045] The embodiments of the present application will be described in detail below with reference to the drawings.

[0046] As Figure 1 shown, the embodiment of the present application provides a liver and gallbladder and pancreatic tumor patient symptom evaluation system based on multi-dimensional data, comprising:

[0047] A multi-modal data acquisition module is used to synchronously acquire clinical diagnosis and treatment data of a patient, physiological index data of a wearable device, behavior activity data, medical image data and patient subjective perception data;

[0048] A time series feature extraction module uses an improved attention mechanism LSTM network to extract features from time series data to obtain a dynamic feature vector;

[0049] A heterogeneous data fusion module fuses multi-dimensional heterogeneous data through a meta-learning self-adaptive weight distribution algorithm to generate a patient comprehensive evaluation vector;

[0050] a dynamic assessment model module, which constructs a symptom assessment model based on a deep belief network, and outputs a patient symptom severity index and a symptom development trend prediction;

[0051] an intelligent feedback intervention module, which generates personalized intervention suggestions according to the assessment results, and adjusts the assessment strategy in real time.

[0052] In one embodiment, the multi-modal data acquisition module comprises:

[0053] a clinical data interface unit, which interfaces with a hospital HIS and LIS system, and acquires pathological diagnosis, test reports, medication records, and surgery history data of a patient;

[0054] a wearable device unit, which comprises a smart bracelet and a patch sensor, and acquires heart rate variability, skin resistance, body temperature, and respiratory rate of a patient;

[0055] a behavior monitoring unit, which acquires activity amount, sleep duration, eating frequency, and defecation frequency of a patient through home sensors and a mobile phone APP;

[0056] an image data unit, which receives DICOM files of CT, MRI, and ultrasound images, and performs three-dimensional reconstruction and lesion marking;

[0057] a subjective perception unit, which acquires pain degree, nausea feeling, fatigue level, and psychological state score of a patient through voice interaction and visual scales.

[0058] In one embodiment, the improved attention mechanism LSTM network adopted by the time series feature extraction module comprises:

[0059] an input layer, which performs standardization processing on physiological index data and behavior activity data;

[0060] an LSTM layer, which contains 3 hidden layers, and each hidden layer contains 64 neurons, and is used to capture long-term dependence of time series;

[0061] an attention layer, which gives dynamic weights to features at different time points, and gives 1.5-2.0 times weights to physiological data 30 minutes before and after pain onset;

[0062] an output layer, which generates a 128-dimensional time series dynamic feature vector.

[0063] In one embodiment, the heterogeneous data fusion module comprises:

[0064] a data preprocessing subunit, which performs encoding processing on clinical diagnosis and treatment data, feature dimension reduction on image data, and quantitative conversion on subjective perception data;

[0065] The meta-learning weight distribution subunit constructs a meta-learner based on the tumor type, disease stage and treatment plan of the patient, and dynamically adjusts the fusion weight of each dimension data;

[0066] The feature fusion subunit combines the attention mechanism and the splicing strategy to fuse the multi-dimensional features into a 256-dimensional comprehensive evaluation vector.

[0067] In one embodiment, the dynamic evaluation model module comprises:

[0068] The symptom severity evaluation sub-module outputs a symptom index of 0-10 points, wherein 0-3 points are mild symptoms, 4-6 points are moderate symptoms, and 7-10 points are severe symptoms.

[0069] The trend prediction sub-module predicts the symptom development trend in the next 3 days based on the evaluation data in the past 7 days, and outputs the probability of rising, stable or falling.

[0070] The model self-optimization sub-module automatically fine-tunes the model parameters every time 100 new patient data are included, to maintain the evaluation accuracy.

[0071] In one embodiment, the intelligent feedback intervention module comprises:

[0072] The clinical suggestion generation unit generates drug adjustment, examination item and nursing measure suggestions for moderate or severe symptoms.

[0073] The patient guidance unit pushes dietary adjustment, body position change and relaxation training guidance to the patient through voice and graphic methods.

[0074] The dynamic strategy adjustment unit optimizes the parameter weight of the evaluation model in real time according to the response data of the patient to the intervention measures.

[0075] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of various changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A symptom assessment system for patients with hepatobiliary and pancreatic tumors based on multi-dimensional data, characterized in that: include: The multimodal data acquisition module is used to simultaneously collect patients' clinical diagnosis and treatment data, wearable device physiological index data, behavioral activity data, medical imaging data, and patients' subjective perception data. The temporal feature extraction module uses an improved attention mechanism LSTM network to extract features from temporal data and obtain dynamic feature vectors. The heterogeneous data fusion module fuses multi-dimensional heterogeneous data through a meta-learning adaptive weight allocation algorithm to generate a comprehensive patient assessment vector. The dynamic assessment model module constructs a symptom assessment model based on a deep belief network, and outputs a patient symptom severity index and a prediction of symptom development trends. The intelligent feedback intervention module generates personalized intervention suggestions based on the assessment results and adjusts the assessment strategy in real time.

2. The symptom assessment system for hepatobiliary and pancreatic tumor patients based on multi-dimensional data according to claim 1, characterized in that: The multimodal data acquisition module includes: The clinical data interface unit connects with the hospital's HIS and LIS systems to obtain patients' pathological diagnoses, laboratory reports, medication records, and surgical history data. Wearable device units, including smart bracelets and patch sensors, collect data on the patient's heart rate variability, skin resistance, body temperature, and respiratory rate; The behavior monitoring unit collects data on patients’ activity levels, sleep duration, eating frequency, and toilet visits through home sensors and a mobile app. The imaging data unit receives DICOM files from CT, MRI, and ultrasound images and performs three-dimensional reconstruction and lesion marking. The subjective perception unit collects patients' pain levels, nausea, fatigue levels, and psychological state scores through voice interaction and visual scales.

3. The symptom assessment system for hepatobiliary and pancreatic tumor patients based on multi-dimensional data according to claim 1, characterized in that: The improved attention mechanism LSTM network used in the temporal feature extraction module includes: The input layer standardizes physiological indicator data and behavioral activity data; The LSTM layer contains 3 hidden layers, each containing 64 neurons, used to capture long-term dependencies in time series. The attention layer assigns dynamic weights to features at different time points, and assigns 1.5-2.0 times the weight to physiological data 30 minutes before and after the onset of pain. The output layer generates a 128-dimensional temporal dynamic feature vector.

4. The symptom assessment system for hepatobiliary and pancreatic tumor patients based on multi-dimensional data according to claim 1, characterized in that: The heterogeneous data fusion module includes: The data preprocessing subunit encodes clinical diagnosis and treatment data, performs feature dimensionality reduction on image data, and performs quantization conversion on subjective perception data. The meta-learning weight allocation sub-unit constructs a meta-learner based on the patient's tumor type, disease stage, and treatment plan, and dynamically adjusts the fusion weights of data from each dimension. The feature fusion subunit uses a combination of attention mechanism and splicing strategy to fuse multi-dimensional features into a 256-dimensional comprehensive evaluation vector.

5. The symptom assessment system for hepatobiliary and pancreatic tumor patients based on multi-dimensional data according to claim 1, characterized in that: The dynamic evaluation model module includes: The symptom severity assessment submodule outputs a symptom index of 0-10, where 0-3 indicates mild symptoms, 4-6 indicates moderate symptoms, and 7-10 indicates severe symptoms. The trend prediction submodule, based on the assessment data of the past 7 days, predicts the symptom development trend for the next 3 days and outputs the probability of rising, stabilizing, or falling. The model self-optimization submodule automatically fine-tunes the model parameters for every 100 new patients' data included, maintaining assessment accuracy.

6. The symptom assessment system for hepatobiliary and pancreatic tumor patients based on multi-dimensional data according to claim 1, characterized in that: The intelligent feedback intervention module includes: The clinical suggestion generation unit generates suggestions for medication adjustment, examination items, and nursing measures for moderate to severe symptoms; The patient guidance unit provides guidance on dietary adjustments, posture changes, and relaxation training through voice and text. The dynamic strategy adjustment unit optimizes the parameter weights of the evaluation model in real time based on patient response data to interventions.

Images