[go: up one dir, main page]

CN118658633A - Lymphedema limb volume measurement data management system and method - Google Patents

Lymphedema limb volume measurement data management system and method Download PDF

Info

Publication number
CN118658633A
CN118658633A CN202411135190.8A CN202411135190A CN118658633A CN 118658633 A CN118658633 A CN 118658633A CN 202411135190 A CN202411135190 A CN 202411135190A CN 118658633 A CN118658633 A CN 118658633A
Authority
CN
China
Prior art keywords
limb volume
semantic
limb
lymphedema
temporal
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202411135190.8A
Other languages
Chinese (zh)
Other versions
CN118658633B (en
Inventor
冯越
马欣蕾
徐东升
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jilin University
Original Assignee
Jilin University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Jilin University filed Critical Jilin University
Priority to CN202411135190.8A priority Critical patent/CN118658633B/en
Publication of CN118658633A publication Critical patent/CN118658633A/en
Application granted granted Critical
Publication of CN118658633B publication Critical patent/CN118658633B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/103—Measuring devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
    • A61B5/107—Measuring physical dimensions, e.g. size of the entire body or parts thereof
    • A61B5/1073—Measuring volume, e.g. of limbs
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/74—Details of notification to user or communication with user or patient ; user input means
    • A61B5/746—Alarms related to a physiological condition, e.g. details of setting alarm thresholds or avoiding false alarms
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/211—Selection of the most significant subset of features
    • G06F18/2113—Selection of the most significant subset of features by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2123/00—Data types
    • G06F2123/02—Data types in the time domain, e.g. time-series data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • Public Health (AREA)
  • Medical Informatics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Pathology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • Veterinary Medicine (AREA)
  • Databases & Information Systems (AREA)
  • Epidemiology (AREA)
  • Primary Health Care (AREA)
  • Physiology (AREA)
  • Dentistry (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Medical Treatment And Welfare Office Work (AREA)

Abstract

本申请提供了一种淋巴水肿肢体体积测量数据管理系统及方法，涉及智能数据分析领域，其采用基于AI的数据分析和处理技术来分别对被监测淋巴水肿肢体患者的淋巴水肿肢体体积数据和治疗数据进行局部时序特征提取、语义推理和语义融合，以此根据所述肢体体积数据和所述治疗数据之间的语义交互融合特征来自动地判断患者的淋巴水肿肢体的体积是否存在异常，并产生相应的预警。这样，能够实现自动化的肢体体积异常检测和预警功能，减少因医生主观经验导致的误判，提升了淋巴水肿肢体体积评估的客观程度和准确程度。

The present application provides a lymphedema limb volume measurement data management system and method, which relates to the field of intelligent data analysis. It uses AI-based data analysis and processing technology to extract local temporal features, semantic reasoning and semantic fusion of the lymphedema limb volume data and treatment data of the monitored lymphedema limb patients, so as to automatically judge whether the volume of the patient's lymphedema limb is abnormal according to the semantic interaction fusion features between the limb volume data and the treatment data, and generate corresponding warnings. In this way, the automated limb volume abnormality detection and warning functions can be realized, the misjudgment caused by the doctor's subjective experience can be reduced, and the objectivity and accuracy of the lymphedema limb volume assessment can be improved.

Description

Lymphedema limb volume measurement data management system and method
Technical Field
The application relates to the field of intelligent data analysis, and more particularly, to a lymphedema limb volume measurement data management system and method.
Background
Lymphedema is a chronic and difficult to cure disease with a lengthy and complex treatment cycle. During this process, periodic assessment of limb volume is critical to guiding treatment.
The existing evaluation technology, such as a water replacement method, has the disadvantages of resource waste, complicated operation, insufficient precision and potential infection risk; lymphoscintigraphy may cause allergic reactions due to the use of contrast agents and is costly. In addition, magnetic Resonance Imaging (MRI) and Computed Tomography (CT), while providing detailed images, are not only expensive, but are also unsuitable for patients carrying cardiac pacemakers, cochlear implants or other metal implants. Moreover, the current evaluation of the treatment effect of lymphedema often depends on subjective judgment of doctors, and the dependence can be deviated due to personal experience difference, mood fluctuation or fatigue state of the doctors, so that accurate evaluation of the illness state and treatment decision are affected. Furthermore, due to the lack of quantitative assessment tools, doctors may have difficulty capturing subtle changes in the condition, leading to erroneous judgment of the treatment outcome.
Thus, there is a need for an optimized lymphedema limb volume measurement data management scheme.
Disclosure of Invention
The application provides a lymphedema limb volume measurement data management system and method aiming at the defects in the prior art.
According to one aspect of the present application there is provided a lymphedema limb volume measurement data management system comprising:
The system comprises a patient lymphedema limb volume data acquisition module, a monitoring module and a monitoring module, wherein the patient lymphedema limb volume data acquisition module is used for acquiring a time sequence of lymphedema limb volume data of a monitored lymphedema limb patient;
the system comprises a patient lymphedema limb treatment data acquisition module, a monitoring module and a monitoring module, wherein the patient lymphedema limb treatment data acquisition module is used for acquiring treatment data of the monitored lymphedema limb patient, and the treatment data comprise physical treatment data and drug treatment data;
the limb volume time sequence feature extraction module is used for arranging the time sequence of the lymphedema limb volume data into a lymphedema limb volume time sequence input vector and then inputting the lymphedema limb volume time sequence pattern feature extractor to obtain a sequence of limb volume local time sequence pattern feature vectors;
the limb volume time sequence feature semantic reasoning module is used for enabling the sequence of the limb volume local time sequence mode feature vectors to pass through a limb volume time sequence node semantic propagation reasoner based on a node energy attenuation mechanism so as to obtain limb volume time sequence semantic propagation reasoning feature vectors;
The treatment data multi-mode fusion module is used for carrying out semantic coding on the treatment data of the monitored lymphedema limb patient to obtain a physical treatment data semantic coding feature vector and a drug treatment data semantic coding feature vector, and then fusing the physical treatment data semantic coding feature vector and the drug treatment data semantic coding feature vector to obtain a treatment multi-mode semantic fusion feature vector;
The treatment multi-mode semantic feature fusion module is used for enabling the treatment multi-mode semantic fusion feature vector and the limb volume time sequence semantic propagation inference feature vector to be used as treatment multi-mode semantic-limb volume time sequence inference semantic dynamic selection interaction fusion features through the dynamic feature selection interaction fusion module based on a bidirectional attention network;
the abnormal detection result generation module is used for dynamically selecting interactive fusion characteristics based on the therapeutic multi-mode semantic-limb volume time sequence reasoning semantic to obtain an abnormal detection result, and generating a corresponding early warning prompt signal to medical staff based on the abnormal detection result;
Wherein, the limbs volume time sequence characteristic draws the module, includes: a lymphedema limb volume data structuring unit, configured to arrange a time sequence of the lymphedema limb volume data into the lymphedema limb volume time sequence input vector according to a time dimension; and the lymphedema limb volume local time sequence characteristic generating unit is used for inputting the lymphedema limb volume time sequence input vector into a lymphedema limb volume time sequence mode characteristic extractor based on a 1D convolution neural network model so as to obtain a sequence of the limb volume local time sequence mode characteristic vector.
In the above lymphedema limb volume measurement data management system, the limb volume time sequence feature semantic reasoning module includes: a limb volume local time sequence pattern feature energy statistical pattern value calculation unit, configured to calculate, based on a maximum value, an average value, and a variance of each limb volume local time sequence pattern feature vector in the sequence of limb volume local time sequence pattern feature vectors, an energy statistical pattern value of each limb volume local time sequence pattern feature vector to obtain a sequence of limb volume local time sequence pattern energy statistical pattern values, where an energy statistical pattern value corresponding to a current limb volume local time sequence pattern feature vector in the sequence of limb volume local time sequence pattern energy statistical pattern values is used as a current node energy statistical pattern value, and other energy statistical pattern values are used as historical node energy statistical pattern values to obtain a sequence of current limb volume local time sequence pattern energy statistical pattern values and historical limb volume local time sequence pattern energy statistical pattern values; the limb volume local time sequence mode node propagation space span value calculation unit is used for counting node propagation space span values between each other limb volume local time sequence mode characteristic vector and the current limb volume local time sequence mode characteristic vector in the sequence of the limb volume local time sequence mode characteristic vectors so as to obtain a sequence of limb volume local time sequence mode propagation space span values; a limb volume local time sequence mode energy transmission attenuation coefficient value calculation unit, configured to determine energy transmission attenuation coefficient values of other limb volume local time sequence mode feature vectors in the sequence of limb volume local time sequence mode feature vectors based on the sequence of limb volume local time sequence mode transmission spatial span values and the sequence of historical limb volume local time sequence mode energy statistics paradigm values to obtain a sequence of limb volume local time sequence mode energy transmission attenuation coefficient values, where the limb volume local time sequence mode energy transmission attenuation coefficient values and the historical limb volume local time sequence mode energy statistics paradigm values form an inverse relationship; the limb volume local time sequence pattern feature weighting unit is used for calculating the weighted sum among the sequences of all other limb volume local time sequence pattern feature vectors in the sequence of the limb volume local time sequence pattern feature vectors by taking the sequence of the limb volume local time sequence pattern energy transmission attenuation coefficient values as a weight sequence so as to obtain a historical limb volume local time sequence pattern energy attenuation time sequence aggregation feature vector; and the limb volume local time sequence mode propagation inference unit is used for fusing the current limb volume local time sequence mode energy statistical norm value to calculate the weighted sum of the historical limb volume local time sequence mode energy attenuation time sequence aggregation feature vector and the current limb volume local time sequence mode feature vector so as to obtain the limb volume time sequence semantic propagation inference feature vector.
In the above lymphedema limb volume measurement data management system, the limb volume local time sequence pattern feature energy statistical paradigm value calculating unit is configured to: extracting the maximum value of the feature vector of the local time sequence mode of the limb volume to obtain the feature maximum value of the local time sequence mode of the limb volume; respectively calculating the average value and variance of the feature vectors of the local time sequence modes of the limb volumes to obtain the feature average value of the local time sequence modes of the limb volumes and the feature variance of the local time sequence modes of the limb volumes; multiplying a value obtained by adding the characteristic variance of the local time sequence mode of the limb volume and a preset super parameter by four to obtain a first energy statistical norm value of the characteristic of the local time sequence mode of the limb volume; calculating the square of the difference between the characteristic maximum value of the local time sequence mode of the limb volume and the characteristic average value of the local time sequence mode of the limb volume to obtain a characteristic difference value of the local time sequence mode of the limb volume; the feature variance of the local time sequence pattern of the limb volume is multiplied by two to obtain a feature variance of the local time sequence pattern of the modulated limb volume, and the value obtained by multiplying the preset super parameter by two and the feature difference value of the local time sequence pattern of the limb volume are added to obtain a second energy statistical norm value of the feature of the local time sequence pattern of the limb volume; dividing the first energy statistical paradigm value of the limb volume local time sequence pattern feature by the second energy statistical paradigm value of the limb volume local time sequence pattern feature to obtain the limb volume local time sequence pattern energy statistical paradigm value.
In the above-described lymphedema limb volume measurement data management system, the limb volume local time series pattern energy propagation attenuation coefficient value calculation unit is configured to: multiplying each limb volume local time sequence mode propagation space span value in the sequence of limb volume local time sequence mode propagation space span values with a first weight super parameter to obtain a sequence of first weight propagation space span values; taking each limb volume local time sequence mode propagation space span value in the sequence of limb volume local time sequence mode propagation space span values as an index of a natural constant to calculate an index function value based on e according to positions so as to obtain a sequence of limb volume local time sequence mode propagation type support space span values; multiplying each limb volume local time sequence mode propagation class support space span value in the sequence of limb volume local time sequence mode propagation class support space span values with a second weight super parameter to obtain a sequence of second weighted propagation space span values; calculating a per-position sum of the sequence of first weighted propagation space span values and the sequence of second weighted propagation space span values to obtain a sequence of weighted sum propagation space span values; dividing the historical limb volume local time sequence pattern energy statistical normal form value corresponding to each group in the sequence of the historical limb volume local time sequence pattern energy statistical normal form value and the sequence of the weighted sum propagation space span value by the weighted sum propagation space span value to obtain a sequence of the limb volume local time sequence pattern energy propagation attenuation coefficient values.
In the lymphedema limb volume measurement data management system, the therapeutic multi-modality-limb volume semantic feature fusion module comprises: the limb volume time sequence semantic enhancement unit is used for carrying out feature semantic interaction enhancement on the limb volume time sequence semantic propagation inference feature vector based on the treatment multi-mode semantic fusion feature vector so as to obtain an enhanced limb volume time sequence semantic propagation inference feature vector containing treatment multi-mode semantic-limb volume time sequence inference semantic interaction information; the treatment multi-mode semantic enhancement unit is used for carrying out characteristic semantic interaction enhancement on the treatment multi-mode semantic fusion feature vector based on the limb volume time sequence semantic propagation inference feature vector so as to obtain an enhanced treatment multi-mode semantic fusion feature vector containing limb volume time sequence inference semantic-treatment multi-mode semantic interaction information; the intensive therapy multi-mode semantic-limb volume interaction fusion unit is used for carrying out cascade processing on the intensive therapy multi-mode semantic propagation inference feature vector and the intensive therapy multi-mode semantic fusion feature vector to obtain an intensive therapy multi-mode semantic-limb volume interaction fusion feature vector; the interaction information fusion response gating value calculation unit is used for inputting the intensive therapy multi-mode semantic-limb volume interaction fusion feature vector into a gating response function to obtain an intensive therapy multi-mode semantic-limb volume interaction information fusion response gating value; the treatment multi-mode semantic gating adjustment unit is used for calculating the position-based product between the treatment multi-mode semantic fusion feature vector and the reinforced treatment multi-mode semantic-limb volume interaction information fusion response gating value to obtain a treatment multi-mode semantic gating adjustment vector; the limb volume time sequence semantic gating and adjusting unit is used for calculating the position-based product between the limb volume time sequence semantic propagation inference feature vector and a subtracted multi-mode semantic-limb volume interaction information fusion response gating value of the intensive treatment so as to obtain a limb volume time sequence semantic gating and adjusting vector; the semantic dynamic selection interaction fusion unit is used for carrying out position-based addition on the therapeutic multi-mode semantic gating adjustment vector and the limb volume time sequence semantic gating adjustment vector to obtain the therapeutic multi-mode semantic-limb volume time sequence reasoning semantic dynamic selection interaction fusion feature vector.
In the lymphedema limb volume measurement data management system, the limb volume time sequence semantic enhancement unit is used for: multiplying the first weight matrix with the therapeutic multi-modal semantic fusion feature vector to obtain a therapeutic multi-modal semantic first weight vector; multiplying the limb volume time sequence semantic propagation inference feature vector with a second weight matrix and a third weight matrix respectively to obtain a limb volume time sequence semantic propagation inference second weight vector and a limb volume time sequence semantic propagation inference third weight vector; after calculating the product between the transpose vector of the therapeutic multi-modal semantic first weight vector and the limb volume time sequence semantic propagation inference second weight vector to obtain a therapeutic multi-modal-limb volume correlation value, dividing the therapeutic multi-modal-limb volume correlation value by the square root of the length of the limb volume time sequence semantic propagation inference feature vector to obtain a therapeutic multi-modal-limb volume weight value; inputting the therapeutic multi-modality-limb volume weight value into a softmax function to obtain a therapeutic multi-modality-limb volume activation weight value; and taking the therapeutic multi-mode-limb volume activation weight value as a weight value, and weighting the transposed vector of the third weight vector of the limb volume time sequence semantic reasoning so as to obtain the reinforced limb volume time sequence semantic reasoning feature vector containing therapeutic multi-mode semantic-limb volume time sequence reasoning semantic interaction information.
In the lymphedema limb volume measurement data management system, the interactive information fusion response gating value calculation unit is used for: after calculating the matrix product of the intensive therapy multi-mode semantic-limb volume interaction fusion feature vector and the parameter matrix, adding the obtained feature vector and the bias vector according to positions to obtain the intensive therapy multi-mode semantic-limb volume interaction fusion bias feature vector; inputting the intensive therapy multimode semantic-limb volume interaction fusion bias feature vector into a sigmoid activation function to obtain a intensive therapy multimode semantic-limb volume interaction information fusion response gating value.
In the lymphedema limb volume measurement data management system, the abnormality detection result generation module is configured to: the therapeutic multi-mode semantic-limb volume time sequence reasoning semantic dynamic selection interaction fusion feature vector is passed through a lymphedema limb volume anomaly identifier based on a classifier to obtain an anomaly detection result, wherein the anomaly detection result is used for indicating whether the volume of a lymphedema limb of a monitored lymphedema limb patient is abnormal or not; and responding to the abnormal detection result to detect that the volume of the lymphedema limb of the monitored lymphedema limb patient is abnormal, and generating an early warning prompt signal to medical staff.
According to another aspect of the present application, there is provided a lymphedema limb volume measurement data management method comprising:
obtaining a time series of lymphedema limb volume data of a monitored lymphedema limb patient;
acquiring treatment data of the monitored lymphedema limb patient, wherein the treatment data comprises physical treatment data and drug treatment data;
The time sequence of the lymphedema limb volume data is arranged into a lymphedema limb volume time sequence input vector, and then the lymphedema limb volume time sequence pattern feature extractor is input to obtain a limb volume local time sequence pattern feature vector sequence;
passing the sequence of the feature vectors of the local time sequence mode of the limb volume through a semantic propagation reasoner of the limb volume time sequence nodes based on a node energy attenuation mechanism to obtain a semantic propagation reasoning feature vector of the limb volume time sequence;
After carrying out semantic coding on the treatment data of the monitored lymphedema limb patient to obtain a physical treatment data semantic coding feature vector and a drug treatment data semantic coding feature vector, fusing the physical treatment data semantic coding feature vector and the drug treatment data semantic coding feature vector to obtain a treatment multi-mode semantic fusion feature vector;
The therapeutic multi-mode semantic fusion feature vector and the limb volume time sequence semantic propagation inference feature vector are processed through a dynamic feature selection interaction fusion module based on a bidirectional attention network to obtain a therapeutic multi-mode semantic-limb volume time sequence inference semantic dynamic selection interaction fusion feature vector serving as a therapeutic multi-mode semantic-limb volume time sequence inference semantic dynamic selection interaction fusion feature;
dynamically selecting interactive fusion characteristics based on the therapeutic multi-mode semantics-limb volume time sequence reasoning semantics to obtain an abnormal detection result, and generating corresponding early warning prompt signals to medical staff based on the abnormal detection result;
The method for obtaining the lymphedema limb volume time sequence comprises the steps of arranging the time sequence of the lymphedema limb volume data into a lymphedema limb volume time sequence input vector, and then inputting the lymphedema limb volume time sequence mode feature extractor to obtain a limb volume local time sequence mode feature vector sequence, and comprises the following steps:
arranging the time sequence of the lymphedema limb volume data into a time sequence input vector of the lymphedema limb volume according to a time dimension;
inputting the lymphedema limb volume time sequence input vector into a lymphedema limb volume time sequence pattern feature extractor based on a 1D convolution neural network model to obtain a sequence of the limb volume local time sequence pattern feature vector.
Compared with the prior art, the lymphedema limb volume measurement data management system and method provided by the application adopt the AI-based data analysis and processing technology to respectively conduct local time sequence feature extraction, semantic reasoning and semantic fusion on the lymphedema limb volume data and the treatment data of the monitored lymphedema limb patient, so as to automatically judge whether the volume of the lymphedema limb of the patient is abnormal or not according to the semantic interaction fusion features between the limb volume data and the treatment data, and generate corresponding early warning. Therefore, the automatic limb volume abnormality detection and early warning functions can be realized, misjudgment caused by subjective experience of doctors is reduced, and objective degree and accuracy degree of lymphedema limb volume assessment are improved.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, and it is obvious that the drawings in the following description are some embodiments of the present application, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art. In the drawings:
fig. 1 is a system block diagram of a lymphedema limb volume measurement data management system in accordance with an embodiment of the present application.
Fig. 2 is a schematic diagram of the architecture of a lymphedema limb volume measurement data management system according to an embodiment of the present application.
Fig. 3 is a flow chart of a method of lymphedema limb volume measurement data management in accordance with an embodiment of the present application.
Detailed Description
Hereinafter, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be apparent that the described embodiments are only some embodiments of the present application and not all embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described herein.
Lymphedema is a chronic disease that is difficult to cure and its treatment requires long-term and complex assessment. The existing evaluation methods such as a water replacement method have the disadvantages of resource waste, complex operation, low precision and infection risk; lymphoscintigraphy is costly and may cause allergies due to the use of contrast agents. MRI and CT, while imaging clearly, are expensive and unsuitable for patients with metallic implants. Moreover, treatment effect evaluation often depends on subjective judgment of doctors, which may generate deviation due to personal differences of doctors, and lack of quantification tools may cause misjudgment on disease changes.
Therefore, in view of the above problems, the technical idea of the present application is to automatically determine whether there is an abnormality in the volume of the lymphedema limb of the monitored lymphedema limb patient according to the semantic interactive fusion characteristics between the limb volume data and the treatment data by acquiring the time series of the lymphedema limb volume data of the monitored lymphedema limb patient and the treatment data (physical treatment and medication treatment) of the monitored lymphedema limb patient and performing local time sequence feature extraction, semantic reasoning and semantic fusion on the lymphedema limb volume data and the treatment data respectively by adopting the AI-based data analysis and processing technology, and generate corresponding early warning. Therefore, the automatic limb volume abnormality detection and early warning function can be realized, misjudgment caused by subjective experience of doctors is reduced, time sequence information and semantic association in volume data and treatment data are effectively captured, medical staff is helped to comprehensively understand the condition of patients, and more intelligent lymphedema limb volume measurement data management is realized.
Fig. 1 is a system block diagram of a lymphedema limb volume measurement data management system in accordance with an embodiment of the present application. Fig. 2 is a schematic diagram of the architecture of a lymphedema limb volume measurement data management system according to an embodiment of the present application. As shown in fig. 1 and 2, in the lymphedema limb volume measurement data management system 100, it includes: a patient lymphedema limb volume data acquisition module 110 for acquiring a time series of lymphedema limb volume data of a monitored lymphedema limb patient; a patient lymphedema limb therapy data acquisition module 120 for acquiring therapy data of the monitored lymphedema limb patient, the therapy data including physical therapy data and medication therapy data; a limb volume time sequence feature extraction module 130, configured to arrange the time sequence of the lymphedema limb volume data into a lymphedema limb volume time sequence input vector, and then input the lymphedema limb volume time sequence pattern feature extractor to obtain a sequence of limb volume local time sequence pattern feature vectors; the limb volume time sequence feature semantic reasoning module 140 is used for passing the sequence of the limb volume local time sequence pattern feature vectors through a limb volume time sequence node semantic propagation reasoner based on a node energy attenuation mechanism to obtain limb volume time sequence semantic propagation reasoning feature vectors; the therapeutic data multi-mode fusion module 150 is configured to perform semantic coding on therapeutic data of the monitored lymphedema limb patient to obtain a physical therapeutic data semantic coding feature vector and a drug therapeutic data semantic coding feature vector, and then fuse the physical therapeutic data semantic coding feature vector and the drug therapeutic data semantic coding feature vector to obtain a therapeutic multi-mode semantic fusion feature vector; a therapeutic multi-mode-limb volume semantic feature fusion module 160, configured to use the therapeutic multi-mode semantic fusion feature vector and the limb volume time-sequence semantic propagation inference feature vector as therapeutic multi-mode semantic-limb volume time-sequence inference semantic dynamic selection interaction fusion features by using a dynamic feature selection interaction fusion module based on a bidirectional attention network to obtain a therapeutic multi-mode semantic-limb volume time-sequence inference semantic dynamic selection interaction fusion feature vector; the abnormality detection result generation module 170 is configured to dynamically select the interaction fusion feature based on the treatment multi-mode semantic-limb volume time sequence inference semantic, obtain an abnormality detection result, and generate a corresponding early warning prompt signal to a medical staff based on the abnormality detection result.
In an embodiment of the present application, the patient lymphedema limb volume data acquisition module 110 and the patient lymphedema limb treatment data acquisition module 120 are respectively configured to acquire a time sequence of lymphedema limb volume data of a monitored lymphedema limb patient and acquire treatment data of the monitored lymphedema limb patient, where the treatment data includes physical treatment data and drug treatment data. It will be appreciated that the time series of lymphedema limb volume data can reflect changes in limb volume over time, and that by analysing trends in limb volume, potential abnormal conditions of limb volume can be identified. While patient physical therapy data and medication data may help to understand the specifics of physical therapy and medication that the patient receives, and thus analyze the effects of these treatments on changes in limb volume. By combining and analyzing the lymphedema limb volume data and the treatment data, a more perfect limb volume abnormality identification model can be constructed, so that the accuracy of limb volume abnormality detection is improved. In particular, in one embodiment of the present application, the acquisition of lymphedema limb volume data and treatment data may be performed by clinical treatment recordings of the patient.
In the embodiment of the present application, the limb volume time sequence feature extraction module 130 is configured to arrange the time sequence of the lymphedema limb volume data into a lymphedema limb volume time sequence input vector, and then input the lymphedema limb volume time sequence pattern feature extractor to obtain a sequence of limb volume local time sequence pattern feature vectors. Specifically, in an embodiment of the present application, the limb volume time sequence feature extraction module includes: a lymphedema limb volume data structuring unit, configured to arrange a time sequence of the lymphedema limb volume data into the lymphedema limb volume time sequence input vector according to a time dimension; and the lymphedema limb volume local time sequence characteristic generating unit is used for inputting the lymphedema limb volume time sequence input vector into a lymphedema limb volume time sequence mode characteristic extractor based on a 1D convolution neural network model so as to obtain a sequence of the limb volume local time sequence mode characteristic vector. It should be understood that the lymphedema limb volume data has time sequence characteristic information, so in order to better understand the time sequence variation fluctuation trend of the limb volume, in the technical scheme of the application, the time sequence of the lymphedema limb volume data is arranged into a lymphedema limb volume time sequence input vector according to the time dimension. Then, in order to capture and mine out the local time sequence characteristic information in the lymphedema limb volume time sequence input vector, in the technical scheme of the application, the lymphedema limb volume time sequence input vector is input into a lymphedema limb volume time sequence mode characteristic extractor based on a 1D convolution neural network model so as to obtain a sequence of limb volume local time sequence mode characteristic vectors.
In the embodiment of the present application, the semantic reasoning module 140 for the limb volume time sequence features is configured to pass the sequence of the feature vectors of the local time sequence pattern of the limb volume through a semantic propagation reasoner for the limb volume time sequence nodes based on a node energy attenuation mechanism to obtain the semantic propagation reasoning feature vectors for the limb volume time sequence. Accordingly, considering that each limb volume local time sequence pattern feature vector in the sequence of limb volume local time sequence pattern feature vectors expresses time sequence change information of limb volume in a local time period, a limb volume time sequence semantic propagation inference feature vector is obtained by taking the sequence of the limb volume local time sequence pattern feature vectors as a reference to better identify limb volume change feature points which are more critical and remarkable in the whole time sequence, so that the whole lymphedema limb volume feature is accurately inferred. It is worth mentioning that the limb volume time sequence node semantic propagation reasoner based on the node energy attenuation mechanism utilizes the information attenuation process of time change in the time sequence data to help the model to better understand the dynamic change and development trend in the time sequence data. That is, by dynamically evaluating the energy level of the feature vector of each local time sequence mode of the limb volume, the topological structure among the nodes is captured, and the energy propagation attenuation and the propagation of time sequence space features are realized, so that the sensitivity of the model to the time sequence dynamic change of the node features can be improved, the trend and the mode of the change of the limb volume features along with the time are effectively captured, and the evolution process of the limb volume time sequence data is better deduced.
Specifically, in the embodiment of the present application, the semantic reasoning module 140 for the time sequence characteristics of the limb volume includes: a limb volume local time sequence pattern feature energy statistical pattern value calculation unit, configured to calculate, based on a maximum value, an average value, and a variance of each limb volume local time sequence pattern feature vector in the sequence of limb volume local time sequence pattern feature vectors, an energy statistical pattern value of each limb volume local time sequence pattern feature vector to obtain a sequence of limb volume local time sequence pattern energy statistical pattern values, where an energy statistical pattern value corresponding to a current limb volume local time sequence pattern feature vector in the sequence of limb volume local time sequence pattern energy statistical pattern values is used as a current node energy statistical pattern value, and other energy statistical pattern values are used as historical node energy statistical pattern values to obtain a sequence of current limb volume local time sequence pattern energy statistical pattern values and historical limb volume local time sequence pattern energy statistical pattern values; the limb volume local time sequence mode node propagation space span value calculation unit is used for counting node propagation space span values between each other limb volume local time sequence mode characteristic vector and the current limb volume local time sequence mode characteristic vector in the sequence of the limb volume local time sequence mode characteristic vectors so as to obtain a sequence of limb volume local time sequence mode propagation space span values; a limb volume local time sequence mode energy transmission attenuation coefficient value calculation unit, configured to determine energy transmission attenuation coefficient values of other limb volume local time sequence mode feature vectors in the sequence of limb volume local time sequence mode feature vectors based on the sequence of limb volume local time sequence mode transmission spatial span values and the sequence of historical limb volume local time sequence mode energy statistics paradigm values to obtain a sequence of limb volume local time sequence mode energy transmission attenuation coefficient values, where the limb volume local time sequence mode energy transmission attenuation coefficient values and the historical limb volume local time sequence mode energy statistics paradigm values form an inverse relationship; the limb volume local time sequence pattern feature weighting unit is used for calculating the weighted sum among the sequences of all other limb volume local time sequence pattern feature vectors in the sequence of the limb volume local time sequence pattern feature vectors by taking the sequence of the limb volume local time sequence pattern energy transmission attenuation coefficient values as a weight sequence so as to obtain a historical limb volume local time sequence pattern energy attenuation time sequence aggregation feature vector; and the limb volume local time sequence mode propagation inference unit is used for fusing the current limb volume local time sequence mode energy statistical norm value to calculate the weighted sum of the historical limb volume local time sequence mode energy attenuation time sequence aggregation feature vector and the current limb volume local time sequence mode feature vector so as to obtain the limb volume time sequence semantic propagation inference feature vector.
More specifically, in an embodiment of the present application, the limb volume local time series pattern feature energy statistical paradigm value calculating unit is configured to: extracting the maximum value of the feature vector of the local time sequence mode of the limb volume to obtain the feature maximum value of the local time sequence mode of the limb volume; respectively calculating the average value and variance of the feature vectors of the local time sequence modes of the limb volumes to obtain the feature average value of the local time sequence modes of the limb volumes and the feature variance of the local time sequence modes of the limb volumes; multiplying a value obtained by adding the characteristic variance of the local time sequence mode of the limb volume and a preset super parameter by four to obtain a first energy statistical norm value of the characteristic of the local time sequence mode of the limb volume; calculating the square of the difference between the characteristic maximum value of the local time sequence mode of the limb volume and the characteristic average value of the local time sequence mode of the limb volume to obtain a characteristic difference value of the local time sequence mode of the limb volume; the feature variance of the local time sequence pattern of the limb volume is multiplied by two to obtain a feature variance of the local time sequence pattern of the modulated limb volume, and the value obtained by multiplying the preset super parameter by two and the feature difference value of the local time sequence pattern of the limb volume are added to obtain a second energy statistical norm value of the feature of the local time sequence pattern of the limb volume; dividing the first energy statistical paradigm value of the limb volume local time sequence pattern feature by the second energy statistical paradigm value of the limb volume local time sequence pattern feature to obtain the limb volume local time sequence pattern energy statistical paradigm value.
More specifically, in an embodiment of the present application, the limb volume local time series pattern energy propagation attenuation coefficient value calculating unit is configured to: multiplying each limb volume local time sequence mode propagation space span value in the sequence of limb volume local time sequence mode propagation space span values with a first weight super parameter to obtain a sequence of first weight propagation space span values; taking each limb volume local time sequence mode propagation space span value in the sequence of limb volume local time sequence mode propagation space span values as an index of a natural constant to calculate an index function value based on e according to positions so as to obtain a sequence of limb volume local time sequence mode propagation type support space span values; multiplying each limb volume local time sequence mode propagation class support space span value in the sequence of limb volume local time sequence mode propagation class support space span values with a second weight super parameter to obtain a sequence of second weighted propagation space span values; calculating a per-position sum of the sequence of first weighted propagation space span values and the sequence of second weighted propagation space span values to obtain a sequence of weighted sum propagation space span values; dividing the historical limb volume local time sequence pattern energy statistical normal form value corresponding to each group in the sequence of the historical limb volume local time sequence pattern energy statistical normal form value and the sequence of the weighted sum propagation space span value by the weighted sum propagation space span value to obtain a sequence of the limb volume local time sequence pattern energy propagation attenuation coefficient values.
In an embodiment of the present application, specifically, the limb volume timing feature semantic reasoning module 140 is configured to: processing the sequence of the feature vectors of the local time sequence mode of the limb volume through a semantic propagation reasoner of the time sequence nodes of the limb volume based on a node energy attenuation mechanism according to the following propagation reasoning formula to obtain the semantic propagation reasoning feature vectors of the time sequence of the limb volume; wherein, the propagation reasoning formula is:
；
；
；
；
；
Wherein, For the sequence of limb volume local time sequence pattern feature vectors,AndRespectively, the first in the sequence of the feature vectors of the local time sequence mode of the limb volumeAnd (b)Each of the limb volume local time sequence pattern feature vectors,Representing the number of feature vectors in the sequence of feature vectors of the limb volume local time sequence pattern,AndThe first in the sequence of the feature vectors of the local time sequence mode of the limb volume respectivelyA local time sequence pattern characteristic average value of the individual limb volumes and a local time sequence pattern characteristic variance of the limb volumes,Is the firstThe maximum value in the individual limb volume local time series pattern feature vectors,Is the preset super-parameter of the device,Is the firstThe energy statistical paradigm value of the individual limb volume local time series pattern feature vector,Is the firstThe first of the feature vectors of the time sequence pattern of the body volume partThe characteristic value of the individual position is used,Is the number of eigenvalues in the eigenvector of each limb volume local time sequence mode,AndRespectively the firstAnd (b)The energy statistical paradigm value of the individual limb volume local time series pattern feature vector,Representation ofAnd (3) withThe number of feature vectors that are spaced apart,、、AndIs the weight of the parameter to be exceeded,Is the limb volume time sequence semantic propagation inference feature vector.
In the embodiment of the present application, the therapeutic data multi-mode fusion module 150 is configured to perform semantic encoding on therapeutic data of the monitored lymphedema limb patient to obtain a physical therapeutic data semantic encoding feature vector and a pharmaceutical therapeutic data semantic encoding feature vector, and then fuse the physical therapeutic data semantic encoding feature vector and the pharmaceutical therapeutic data semantic encoding feature vector to obtain a therapeutic multi-mode semantic fusion feature vector. It should be appreciated that the treatment data of the monitored lymphedema limb patient includes semantic associations between respective categories of information, and therefore, the treatment data of the monitored lymphedema limb patient is semantically encoded to obtain a physical treatment data semantically encoded feature vector and a pharmaceutical treatment data semantically encoded feature vector. Then, in order to more comprehensively and comprehensively display the semantic information of the treatment data, the semantic coding feature vector of the physical treatment data and the semantic coding feature vector of the drug treatment data are further fused to obtain a therapeutic multi-mode semantic fusion feature vector.
In the embodiment of the present application, the therapeutic multi-mode-limb volume semantic feature fusion module 160 is configured to use the therapeutic multi-mode semantic fusion feature vector and the limb volume time-sequence semantic propagation inference feature vector as therapeutic multi-mode semantic-limb volume time-sequence inference semantic dynamic selection interaction fusion features by using the dynamic feature selection interaction fusion module based on the bidirectional attention network to obtain the therapeutic multi-mode semantic-limb volume time-sequence inference semantic dynamic selection interaction fusion feature vector. Correspondingly, in order to realize the feature interaction and fusion of the two, the semantic relevance between the treatment data and the limb volume time sequence data is fully considered, so that the complex relationship between the two is more comprehensively captured, and the therapeutic multi-mode semantic fusion feature vector and the limb volume time sequence semantic propagation inference feature vector are subjected to a dynamic feature selection interaction fusion module based on a bidirectional attention network to obtain a therapeutic multi-mode semantic-limb volume time sequence inference semantic dynamic selection interaction fusion feature vector.
Specifically, the dynamic feature selection interaction fusion module based on the bidirectional attention network dynamically selects and focuses on the most relevant features by respectively calculating interaction semantic feature information between the therapeutic multi-mode semantic fusion feature vector and the limb volume time sequence semantic propagation reasoning feature vector. And then merging the information interacted by the two, performing gating selection to obtain a gating value, and finally flexibly adjusting the importance and the weight of the two features based on the gating value to better capture the semantic association between the two data, thereby realizing more accurate and effective feature fusion and finally obtaining the comprehensive therapeutic multi-mode semantic-limb volume time sequence reasoning semantic dynamic selection interaction fusion feature vector.
Specifically, in an embodiment of the present application, the therapeutic multi-modality-limb volume semantic feature fusion module includes: the limb volume time sequence semantic enhancement unit is used for carrying out feature semantic interaction enhancement on the limb volume time sequence semantic propagation inference feature vector based on the treatment multi-mode semantic fusion feature vector so as to obtain an enhanced limb volume time sequence semantic propagation inference feature vector containing treatment multi-mode semantic-limb volume time sequence inference semantic interaction information; the treatment multi-mode semantic enhancement unit is used for carrying out characteristic semantic interaction enhancement on the treatment multi-mode semantic fusion feature vector based on the limb volume time sequence semantic propagation inference feature vector so as to obtain an enhanced treatment multi-mode semantic fusion feature vector containing limb volume time sequence inference semantic-treatment multi-mode semantic interaction information; the intensive therapy multi-mode semantic-limb volume interaction fusion unit is used for carrying out cascade processing on the intensive therapy multi-mode semantic propagation inference feature vector and the intensive therapy multi-mode semantic fusion feature vector to obtain an intensive therapy multi-mode semantic-limb volume interaction fusion feature vector; the interaction information fusion response gating value calculation unit is used for inputting the intensive therapy multi-mode semantic-limb volume interaction fusion feature vector into a gating response function to obtain an intensive therapy multi-mode semantic-limb volume interaction information fusion response gating value; the treatment multi-mode semantic gating adjustment unit is used for calculating the position-based product between the treatment multi-mode semantic fusion feature vector and the reinforced treatment multi-mode semantic-limb volume interaction information fusion response gating value to obtain a treatment multi-mode semantic gating adjustment vector; the limb volume time sequence semantic gating and adjusting unit is used for calculating the position-based product between the limb volume time sequence semantic propagation inference feature vector and a subtracted multi-mode semantic-limb volume interaction information fusion response gating value of the intensive treatment so as to obtain a limb volume time sequence semantic gating and adjusting vector; the semantic dynamic selection interaction fusion unit is used for carrying out position-based addition on the therapeutic multi-mode semantic gating adjustment vector and the limb volume time sequence semantic gating adjustment vector to obtain the therapeutic multi-mode semantic-limb volume time sequence reasoning semantic dynamic selection interaction fusion feature vector.
More specifically, in an embodiment of the present application, the limb volume timing semantic enhancement unit is configured to: multiplying the first weight matrix with the therapeutic multi-modal semantic fusion feature vector to obtain a therapeutic multi-modal semantic first weight vector; multiplying the limb volume time sequence semantic propagation inference feature vector with a second weight matrix and a third weight matrix respectively to obtain a limb volume time sequence semantic propagation inference second weight vector and a limb volume time sequence semantic propagation inference third weight vector; after calculating the product between the transpose vector of the therapeutic multi-modal semantic first weight vector and the limb volume time sequence semantic propagation inference second weight vector to obtain a therapeutic multi-modal-limb volume correlation value, dividing the therapeutic multi-modal-limb volume correlation value by the square root of the length of the limb volume time sequence semantic propagation inference feature vector to obtain a therapeutic multi-modal-limb volume weight value; inputting the therapeutic multi-modality-limb volume weight value into a softmax function to obtain a therapeutic multi-modality-limb volume activation weight value; and taking the therapeutic multi-mode-limb volume activation weight value as a weight value, and weighting the transposed vector of the third weight vector of the limb volume time sequence semantic reasoning so as to obtain the reinforced limb volume time sequence semantic reasoning feature vector containing therapeutic multi-mode semantic-limb volume time sequence reasoning semantic interaction information.
More specifically, in the embodiment of the present application, the interactive information fusion response gating value calculation unit is configured to: after calculating the matrix product of the intensive therapy multi-mode semantic-limb volume interaction fusion feature vector and the parameter matrix, adding the obtained feature vector and the bias vector according to positions to obtain the intensive therapy multi-mode semantic-limb volume interaction fusion bias feature vector; inputting the intensive therapy multimode semantic-limb volume interaction fusion bias feature vector into a sigmoid activation function to obtain a intensive therapy multimode semantic-limb volume interaction information fusion response gating value.
In an embodiment of the present application, specifically, the therapeutic multi-modality-limb volume semantic feature fusion module 160 is configured to: processing the therapeutic multi-mode semantic fusion feature vector and the limb volume time sequence semantic propagation inference feature vector through a dynamic feature selection interaction fusion module based on a bidirectional attention network by using the following interaction fusion formula to obtain the therapeutic multi-mode semantic-limb volume time sequence inference semantic dynamic selection interaction fusion feature vector; the interactive fusion formula is as follows:
；
；
；
；
Wherein, AndThe therapeutic multi-modal semantic fusion feature vector and the limb volume time sequence semantic propagation inference feature vector are respectively,、AndRespectively representing a first weight matrix, a second weight matrix and a third weight matrix of the therapeutic multi-mode semantic fusion feature vector,The transpose of the vector is represented,、AndA first weight matrix, a second weight matrix and a third weight matrix of the limb volume time sequence semantic propagation reasoning feature vector are respectively adopted,Is the length of the therapeutic multi-modal semantic fusion feature vector and the limb volume time sequence semantic propagation inference feature vector,Is thatThe function of the function is that,Is the reinforced limb volume time sequence semantic propagation inference feature vector containing the therapeutic multi-mode semantic-limb volume time sequence inference semantic interaction information,Is the intensive therapy multi-mode semantic fusion feature vector containing limb volume time sequence reasoning semantic-therapy multi-mode semantic interaction information,Is a cascade of processes which are carried out,Is a matrix of parameters that are selected from the group consisting of,Is the offset vector of the reference signal,Is a sigmoid function of the number of bits,Is an information fusion gating value between the reinforced limb volume time sequence semantic propagation reasoning feature vector and the reinforced treatment multi-mode semantic fusion feature vector,The therapeutic multi-mode semantic-limb volume time sequence reasoning semantic dynamic selection interactive fusion feature vector is provided.
In the embodiment of the present application, the abnormality detection result generating module 170 is configured to dynamically select the interaction fusion feature based on the treatment multi-mode semantic-limb volume time sequence inference semantic, obtain an abnormality detection result, and generate a corresponding early warning prompt signal to a medical staff based on the abnormality detection result. Specifically, in the embodiment of the present application, the abnormality detection result generating module 170 is configured to: the therapeutic multi-mode semantic-limb volume time sequence reasoning semantic dynamic selection interaction fusion feature vector is passed through a lymphedema limb volume anomaly identifier based on a classifier to obtain an anomaly detection result, wherein the anomaly detection result is used for indicating whether the volume of a lymphedema limb of a monitored lymphedema limb patient is abnormal or not; and responding to the abnormal detection result to detect that the volume of the lymphedema limb of the monitored lymphedema limb patient is abnormal, and generating an early warning prompt signal to medical staff. The therapeutic multi-mode semantic-limb volume time sequence reasoning semantic dynamic selection interactive fusion feature is obtained by carrying out bidirectional interactive fusion on the therapeutic multi-mode semantic fusion feature vector and the limb volume time sequence semantic propagation reasoning feature vector, so as to carry out classification treatment, automatically judge whether the volume of the lymphedema limb of the monitored lymphedema limb patient is abnormal, and generate corresponding early warning. Therefore, the automatic limb volume abnormality detection and early warning functions can be realized, misjudgment caused by subjective experience of doctors is reduced, time sequence information and semantic association in volume data and treatment data are effectively captured, and medical staff is helped to comprehensively understand the condition of patients.
Particularly, considering that the therapeutic multi-mode semantic fusion feature vector and the limb volume time sequence semantic propagation inference feature vector respectively represent the multi-mode semantic coding feature of the therapeutic data of the monitored lymphedema limb patient and the propagation semantic feature of the one-dimensional time sequence local association of the lymphedema limb volume data based on time sequence node energy attenuation, when the therapeutic multi-mode semantic fusion feature vector and the limb volume time sequence semantic propagation inference feature vector are subjected to dynamic feature selection interactive fusion based on a bidirectional attention network, dynamic selection interactive fusion imbalance caused by imbalance of corresponding proportion under fine granularity distribution of different feature expression dimensional sequences is required to be compensated, so that the expression effect of the therapeutic multi-mode semantic-limb volume time sequence inference semantic dynamic selection interactive fusion feature vector is improved, and the accuracy of classification results is improved.
Therefore, preferably, when the therapeutic multi-mode semantic-limb volume time sequence reasoning semantic dynamic selection interaction fusion feature vector passes through a lymphedema limb volume anomaly identifier based on a classifier to obtain an anomaly detection result, the therapeutic multi-mode semantic-limb volume time sequence reasoning semantic dynamic selection interaction fusion feature vector is corrected, and the method comprises the following steps of: calculating a first norm and a second norm of a mean value vector of the therapeutic multi-mode semantic fusion feature vector and the limb volume time sequence semantic propagation reasoning feature vector; calculating the weighted sum of the square root of the two norms and the one norms, and carrying out point multiplication on the weighted sum and the point-added sum vector of the therapeutic multi-mode semantic fusion feature vector and the limb volume time sequence semantic propagation inference feature vector to obtain a first therapeutic multi-mode semantic-limb volume time sequence inference semantic dynamic selection interaction fusion syndrome vector; performing point multiplication product vector of the therapeutic multi-mode semantic fusion feature vector and the limb volume time sequence semantic propagation inference feature vector and square root of length of the therapeutic multi-mode semantic-limb volume time sequence inference semantic dynamic selection interaction fusion feature vector to obtain a second therapeutic multi-mode semantic-limb volume time sequence inference semantic dynamic selection interaction fusion syndrome vector; calculating a weighted sum of the first therapeutic multi-modal semantic-limb volume time sequence reasoning semantic dynamic selection interaction fusion syndrome vector and the second therapeutic multi-modal semantic-limb volume time sequence reasoning semantic dynamic selection interaction fusion syndrome vector to obtain a therapeutic multi-modal semantic-limb volume time sequence reasoning semantic dynamic selection interaction fusion correction vector; and carrying out point multiplication on the therapeutic multi-mode semantic-limb volume time sequence reasoning semantic dynamic selection interaction fusion correction vector and the therapeutic multi-mode semantic-limb volume time sequence reasoning semantic dynamic selection interaction fusion feature vector to obtain a corrected therapeutic multi-mode semantic-limb volume time sequence reasoning semantic dynamic selection interaction fusion feature vector, and enabling the corrected therapeutic multi-mode semantic-limb volume time sequence reasoning semantic dynamic selection interaction fusion feature vector to pass through a lymphedema limb volume anomaly identifier based on a classifier so as to obtain an anomaly detection result.
Wherein, the therapeutic multi-modal semantic-limb volume time sequence reasoning semantic dynamic selection interaction fusion correction vector is expressed as:
；
Wherein, AndThe therapeutic multi-modal semantic fusion feature vector and the limb volume time sequence semantic propagation inference feature vector are respectively,Is the mean value vector of the therapeutic multi-mode semantic fusion feature vector and the limb volume time sequence semantic propagation reasoning feature vector,Is a norm of the feature vector that is,Is the inverse of the square root of the second norm of the feature vector,Is added according to the position point,Is multiplied by the position point,Is the length of the interactive fusion feature vector of the therapeutic multi-mode semantic-limb volume time sequence reasoning semantic dynamic selection, andAndIs a weighted sum weight as a super-parameter,The therapeutic multi-mode semantic-limb volume time sequence reasoning semantic dynamic selection interactive fusion correction vector is provided.
The method comprises the steps of modeling feature level key correspondence between feature vectors of a therapeutic multi-mode semantic fusion feature vector and a limb volume time sequence semantic propagation inference feature vector based on constraint representation of mean vector norms among the therapeutic multi-mode semantic fusion feature vector and the limb volume time sequence semantic propagation inference feature vector on structural foreground and background distinction of a superfluid of the therapeutic multi-mode semantic fusion feature vector and the limb volume time sequence semantic propagation inference feature vector, and adjusting global association relations among corresponding features, so that positive fine-grained correspondence suggestion is carried out through unbalanced proportional control between corresponding feature values of the therapeutic multi-mode semantic fusion feature vector and the limb volume time sequence semantic propagation inference feature vector, and fusion domain offset imbalance between the therapeutic multi-mode semantic fusion feature vector and the limb volume time sequence semantic propagation inference feature vector is avoided through a focus-of-attention mode.
In this way, the therapeutic multi-mode semantic-limb volume time sequence reasoning semantic dynamic selection interaction fusion correction vector is used for correcting the therapeutic multi-mode semantic-limb volume time sequence reasoning semantic dynamic selection interaction fusion feature vector, so that the expression effect of the therapeutic multi-mode semantic-limb volume time sequence reasoning semantic dynamic selection interaction fusion feature vector can be improved, and the accuracy of the abnormal detection result obtained by the classifier-based lymphedema limb volume abnormal identifier is improved.
In summary, the lymphedema limb volume measurement data management system 100 according to the embodiment of the present application is illustrated, which adopts the AI-based data analysis and processing technology to perform local time sequence feature extraction, semantic reasoning and semantic fusion on the lymphedema limb volume data and the treatment data of the monitored lymphedema limb patient, so as to automatically determine whether the volume of the lymphedema limb of the patient is abnormal according to the semantic interaction fusion features between the limb volume data and the treatment data, and generate corresponding early warning. Therefore, the automatic limb volume abnormality detection and early warning functions can be realized, misjudgment caused by subjective experience of doctors is reduced, and objective degree and accuracy degree of lymphedema limb volume assessment are improved.
As described above, the lymphedema limb volume measurement data management system 100 according to the embodiment of the present application may be implemented in various wireless terminals, such as a server or the like having a lymphedema limb volume measurement data management algorithm. In one possible implementation, the lymphedema limb volume measurement data management system 100 according to embodiments of the present application may be integrated into the wireless terminal as a software module and/or hardware module. For example, the lymphedema limb volume measurement data management system 100 can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the lymphedema limb volume measurement data management system 100 can also be one of the many hardware modules of the wireless terminal.
Alternatively, in another example, the lymphedema limb volume measurement data management system 100 and the wireless terminal may also be separate devices, and the lymphedema limb volume measurement data management system 100 may be connected to the wireless terminal via a wired and/or wireless network and communicate interactive information in accordance with a agreed data format.
Fig. 3 is a flow chart of a method of lymphedema limb volume measurement data management in accordance with an embodiment of the present application. As shown in fig. 3, in the lymphedema limb volume measurement data management method, it includes: s110, acquiring a time sequence of lymphedema limb volume data of a monitored lymphedema limb patient; s120, acquiring treatment data of the monitored lymphedema limb patient, wherein the treatment data comprise physical treatment data and drug treatment data; s130, arranging the time sequence of the lymphedema limb volume data into a lymphedema limb volume time sequence input vector, and then inputting the lymphedema limb volume time sequence pattern feature extractor to obtain a sequence of limb volume local time sequence pattern feature vectors; s140, passing the sequence of the feature vectors of the local time sequence mode of the limb volume through a semantic propagation reasoner of the limb volume time sequence nodes based on a node energy attenuation mechanism to obtain semantic propagation reasoning feature vectors of the limb volume time sequence; s150, carrying out semantic coding on the treatment data of the monitored lymphedema limb patient to obtain a physical treatment data semantic coding feature vector and a drug treatment data semantic coding feature vector, and then fusing the physical treatment data semantic coding feature vector and the drug treatment data semantic coding feature vector to obtain a treatment multi-mode semantic fusion feature vector; s160, the therapeutic multi-mode semantic fusion feature vector and the limb volume time sequence semantic propagation inference feature vector are processed through a dynamic feature selection interaction fusion module based on a bidirectional attention network to obtain a therapeutic multi-mode semantic-limb volume time sequence inference semantic dynamic selection interaction fusion feature vector serving as a therapeutic multi-mode semantic-limb volume time sequence inference semantic dynamic selection interaction fusion feature; s170, dynamically selecting interactive fusion characteristics based on the therapeutic multi-mode semantic-limb volume time sequence reasoning semantic, obtaining an abnormal detection result, and generating corresponding early warning prompt signals to medical staff based on the abnormal detection result.
Here, it will be understood by those skilled in the art that the specific operations of the respective steps in the above-described lymphedema limb volume measurement data management method have been described in detail in the above description of the lymphedema limb volume measurement data management system with reference to fig. 1 to 2, and thus, repetitive descriptions thereof will be omitted.
Implementations of the present disclosure have been described above, the foregoing description is exemplary rather than exhaustive. And is not limited to the implementations disclosed, many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the implementations described. The terminology used herein was chosen in order to best explain the principles of each implementation, the practical application, or the improvement of technology in the marketplace, or to enable others of ordinary skill in the art to understand each of the implementations disclosed herein.

Claims (9)

1.一种淋巴水肿肢体体积测量数据管理系统，其特征在于，包括：1. A lymphedema limb volume measurement data management system, characterized by comprising: 患者淋巴水肿肢体体积数据获取模块，用于获取被监测淋巴水肿肢体患者的淋巴水肿肢体体积数据的时间序列；A patient lymphedema limb volume data acquisition module, used to acquire a time series of lymphedema limb volume data of a monitored lymphedema limb patient; 患者淋巴水肿肢体治疗数据获取模块，用于获取所述被监测淋巴水肿肢体患者的治疗数据，所述治疗数据包括物理治疗数据和药物治疗数据；A patient lymphedema limb treatment data acquisition module, used to acquire treatment data of the monitored lymphedema limb patient, the treatment data including physical therapy data and drug therapy data; 肢体体积时序特征提取模块，用于将所述淋巴水肿肢体体积数据的时间序列排列为淋巴水肿肢体体积时序输入向量后输入淋巴水肿肢体体积时序模式特征提取器以得到肢体体积局部时序模式特征向量的序列；A limb volume time series feature extraction module, used for arranging the time series of the lymphedema limb volume data into a lymphedema limb volume time series input vector and then inputting the vector into a lymphedema limb volume time series pattern feature extractor to obtain a sequence of limb volume local time series pattern feature vectors; 肢体体积时序特征语义推理模块，用于将所述肢体体积局部时序模式特征向量的序列通过基于节点能量衰减机制的肢体体积时序节点语义传播推理器以得到肢体体积时序语义传播推理特征向量；A limb volume temporal feature semantic reasoning module, used for passing the sequence of the limb volume local temporal pattern feature vectors through a limb volume temporal node semantic propagation reasoner based on a node energy decay mechanism to obtain a limb volume temporal semantic propagation reasoning feature vector; 治疗数据多模态融合模块，用于对所述被监测淋巴水肿肢体患者的治疗数据进行语义编码以得到物理治疗数据语义编码特征向量和药物治疗数据语义编码特征向量后，融合所述物理治疗数据语义编码特征向量和所述药物治疗数据语义编码特征向量以得到治疗多模态语义融合特征向量；a treatment data multimodal fusion module, configured to perform semantic encoding on the treatment data of the monitored lymphedema limb patient to obtain a physical therapy data semantic encoding feature vector and a drug therapy data semantic encoding feature vector, and then fuse the physical therapy data semantic encoding feature vector and the drug therapy data semantic encoding feature vector to obtain a treatment multimodal semantic fusion feature vector; 治疗多模态-肢体体积语义特征融合模块，用于将所述治疗多模态语义融合特征向量和所述肢体体积时序语义传播推理特征向量通过基于双向注意力网络的动态特征选择交互融合模块以得到治疗多模态语义-肢体体积时序推理语义动态选择交互融合特征向量作为治疗多模态语义-肢体体积时序推理语义动态选择交互融合特征；A treatment multimodal-limb volume semantic feature fusion module, used for fusing the treatment multimodal semantic fusion feature vector and the limb volume temporal semantic propagation reasoning feature vector through a dynamic feature selection interactive fusion module based on a bidirectional attention network to obtain a treatment multimodal semantic-limb volume temporal reasoning semantic dynamic selection interactive fusion feature vector as a treatment multimodal semantic-limb volume temporal reasoning semantic dynamic selection interactive fusion feature; 异常检测结果生成模块，用于基于所述治疗多模态语义-肢体体积时序推理语义动态选择交互融合特征，得到异常检测结果，并基于所述异常检测结果，产生相应的预警提示信号给医护人员；An abnormal detection result generation module is used to dynamically select interactive fusion features based on the treatment multimodal semantics-limb volume temporal reasoning semantics to obtain abnormal detection results, and based on the abnormal detection results, generate corresponding early warning prompt signals to medical staff; 其中，所述肢体体积时序特征提取模块，包括：Wherein, the limb volume temporal feature extraction module includes: 淋巴水肿肢体体积数据结构化单元，用于将所述淋巴水肿肢体体积数据的时间序列按照时间维度排列为所述淋巴水肿肢体体积时序输入向量；A lymphedema limb volume data structuring unit, used for arranging the time series of the lymphedema limb volume data into the lymphedema limb volume time series input vector according to the time dimension; 淋巴水肿肢体体积局部时序特征生成单元，用于将所述淋巴水肿肢体体积时序输入向量输入基于1D卷积神经网络模型的淋巴水肿肢体体积时序模式特征提取器以得到所述肢体体积局部时序模式特征向量的序列。The lymphedema limb volume local temporal feature generating unit is used to input the lymphedema limb volume temporal input vector into the lymphedema limb volume temporal pattern feature extractor based on the 1D convolutional neural network model to obtain a sequence of the limb volume local temporal pattern feature vectors. 2.根据权利要求1所述的淋巴水肿肢体体积测量数据管理系统，其特征在于，所述肢体体积时序特征语义推理模块，包括：2. The lymphedema limb volume measurement data management system according to claim 1, characterized in that the limb volume temporal feature semantic reasoning module comprises: 肢体体积局部时序模式特征能量统计范式值计算单元，用于基于所述肢体体积局部时序模式特征向量的序列中的各个肢体体积局部时序模式特征向量的最大值、平均值和方差，来计算所述各个肢体体积局部时序模式特征向量的能量统计范式值以得到肢体体积局部时序模式能量统计范式值的序列，其中，将所述肢体体积局部时序模式能量统计范式值的序列中当前的肢体体积局部时序模式特征向量对应的能量统计范式值作为当前节点能量统计范式值且将其他能量统计范式值作为历史节点能量统计范式值以得到当前肢体体积局部时序模式能量统计范式值和历史肢体体积局部时序模式能量统计范式值的序列；A limb volume local temporal pattern feature energy statistical paradigm value calculation unit, used for calculating the energy statistical paradigm value of each limb volume local temporal pattern feature vector in the sequence of the limb volume local temporal pattern feature vectors based on the maximum value, average value and variance of each limb volume local temporal pattern feature vector in the sequence of the limb volume local temporal pattern feature vectors to obtain a sequence of limb volume local temporal pattern energy statistical paradigm values, wherein the energy statistical paradigm value corresponding to the current limb volume local temporal pattern feature vector in the sequence of limb volume local temporal pattern energy statistical paradigm values is used as the current node energy statistical paradigm value and other energy statistical paradigm values are used as historical node energy statistical paradigm values to obtain a sequence of current limb volume local temporal pattern energy statistical paradigm values and historical limb volume local temporal pattern energy statistical paradigm values; 肢体体积局部时序模式节点传播空间跨度值计算单元，用于统计所述肢体体积局部时序模式特征向量的序列中的各个其他肢体体积局部时序模式特征向量与当前肢体体积局部时序模式特征向量之间的节点传播空间跨度值以得到肢体体积局部时序模式传播空间跨度值的序列；A limb volume local time series pattern node propagation space span value calculation unit, used for counting the node propagation space span values between each other limb volume local time series pattern feature vector in the sequence of the limb volume local time series pattern feature vector and the current limb volume local time series pattern feature vector to obtain a sequence of limb volume local time series pattern propagation space span values; 肢体体积局部时序模式能量传播衰减系数值计算单元，用于基于所述肢体体积局部时序模式传播空间跨度值的序列和所述历史肢体体积局部时序模式能量统计范式值的序列，确定所述肢体体积局部时序模式特征向量的序列中的其他各个肢体体积局部时序模式特征向量的能量传播衰减系数值以得到肢体体积局部时序模式能量传播衰减系数值的序列，其中，所述肢体体积局部时序模式能量传播衰减系数值与所述历史肢体体积局部时序模式能量统计范式值成反相关关系；A limb volume local time series pattern energy propagation attenuation coefficient value calculation unit is used to determine the energy propagation attenuation coefficient values of other limb volume local time series pattern feature vectors in the sequence of limb volume local time series pattern feature vectors based on the sequence of limb volume local time series pattern propagation space span values and the sequence of historical limb volume local time series pattern energy statistical paradigm values to obtain a sequence of limb volume local time series pattern energy propagation attenuation coefficient values, wherein the limb volume local time series pattern energy propagation attenuation coefficient values are in an inverse correlation with the historical limb volume local time series pattern energy statistical paradigm values; 肢体体积局部时序模式特征加权单元，用于以所述肢体体积局部时序模式能量传播衰减系数值的序列作为权重序列，计算所述肢体体积局部时序模式特征向量的序列中的其他所有肢体体积局部时序模式特征向量的序列之间的加权和以得到历史肢体体积局部时序模式能量衰减时序聚合特征向量；A limb volume local time series pattern feature weighting unit, used to calculate the weighted sum of all other limb volume local time series pattern feature vector sequences in the sequence of the limb volume local time series pattern feature vector, using the sequence of the limb volume local time series pattern energy propagation attenuation coefficient values as a weight sequence, to obtain a historical limb volume local time series pattern energy attenuation time series aggregation feature vector; 肢体体积局部时序模式传播推理单元，用于融合所述当前肢体体积局部时序模式能量统计范式值来计算所述历史肢体体积局部时序模式能量衰减时序聚合特征向量和所述当前肢体体积局部时序模式特征向量的加权和以得到所述肢体体积时序语义传播推理特征向量。The limb volume local temporal pattern propagation reasoning unit is used to fuse the current limb volume local temporal pattern energy statistical paradigm value to calculate the weighted sum of the historical limb volume local temporal pattern energy attenuation temporal aggregation feature vector and the current limb volume local temporal pattern feature vector to obtain the limb volume temporal semantic propagation reasoning feature vector. 3.根据权利要求2所述的淋巴水肿肢体体积测量数据管理系统，其特征在于，所述肢体体积局部时序模式特征能量统计范式值计算单元，用于：3. The lymphedema limb volume measurement data management system according to claim 2, characterized in that the limb volume local time series pattern feature energy statistical paradigm value calculation unit is used to: 提取所述肢体体积局部时序模式特征向量的最大值以得到肢体体积局部时序模式特征最大值；Extracting the maximum value of the limb volume local time series pattern feature vector to obtain the limb volume local time series pattern feature maximum value; 分别计算所述肢体体积局部时序模式特征向量的平均值和方差以得到肢体体积局部时序模式特征平均值和肢体体积局部时序模式特征方差；Respectively calculating the average value and variance of the limb volume local time series pattern feature vector to obtain the limb volume local time series pattern feature average value and the limb volume local time series pattern feature variance; 将所述肢体体积局部时序模式特征方差和预设超参数进行相加后得到的数值乘以四以得到肢体体积局部时序模式特征第一能量统计范式值；The value obtained by adding the variance of the local temporal pattern feature of the limb volume and the preset hyperparameter is multiplied by four to obtain the first energy statistical paradigm value of the local temporal pattern feature of the limb volume; 计算所述肢体体积局部时序模式特征最大值与所述肢体体积局部时序模式特征平均值的差值的平方以得到肢体体积局部时序模式特征差异值；Calculating the square of the difference between the maximum value of the local temporal pattern feature of the limb volume and the average value of the local temporal pattern feature of the limb volume to obtain a difference value of the local temporal pattern feature of the limb volume; 将所述肢体体积局部时序模式特征方差乘以二得到的调制肢体体积局部时序模式特征方差与所述预设超参数乘以二得到的数值和所述肢体体积局部时序模式特征差异值进行相加以得到肢体体积局部时序模式特征第二能量统计范式值；The second energy statistical paradigm value of the limb volume local temporal pattern feature is obtained by adding the modulated limb volume local temporal pattern feature variance obtained by multiplying the limb volume local temporal pattern feature variance by two, the value obtained by multiplying the preset hyperparameter by two, and the limb volume local temporal pattern feature difference value; 将所述肢体体积局部时序模式特征第一能量统计范式值除以所述肢体体积局部时序模式特征第二能量统计范式值以得到所述肢体体积局部时序模式能量统计范式值。The first energy statistical paradigm value of the local temporal pattern characteristic of the limb volume is divided by the second energy statistical paradigm value of the local temporal pattern characteristic of the limb volume to obtain the energy statistical paradigm value of the local temporal pattern of the limb volume. 4.根据权利要求3所述的淋巴水肿肢体体积测量数据管理系统，其特征在于，所述肢体体积局部时序模式能量传播衰减系数值计算单元，用于：4. The lymphedema limb volume measurement data management system according to claim 3, characterized in that the limb volume local time series pattern energy propagation attenuation coefficient value calculation unit is used to: 将所述肢体体积局部时序模式传播空间跨度值的序列中各个肢体体积局部时序模式传播空间跨度值与第一权重超参数相乘以得到第一加权传播空间跨度值的序列；Multiplying each limb volume local temporal pattern propagation spatial span value in the sequence of limb volume local temporal pattern propagation spatial span values by a first weight hyperparameter to obtain a sequence of first weighted propagation spatial span values; 以所述肢体体积局部时序模式传播空间跨度值的序列中各个肢体体积局部时序模式传播空间跨度值作为自然常数的指数以计算按位置的以e为底的指数函数值以得到肢体体积局部时序模式传播类支持空间跨度值的序列；Using each limb volume local time series pattern propagation space span value in the sequence of limb volume local time series pattern propagation space span values as an exponent of a natural constant to calculate an exponential function value with base e according to position to obtain a sequence of limb volume local time series pattern propagation class support space span values; 将所述肢体体积局部时序模式传播类支持空间跨度值的序列中的每个肢体体积局部时序模式传播类支持空间跨度值与第二权重超参数相乘以得到第二加权传播空间跨度值的序列；Multiplying each limb volume local temporal pattern propagation class support space span value in the sequence of limb volume local temporal pattern propagation class support space span values by a second weight hyperparameter to obtain a sequence of second weighted propagation space span values; 计算所述第一加权传播空间跨度值的序列和所述第二加权传播空间跨度值的序列的按位置加和以得到加权总和传播空间跨度值的序列；Calculating a position-wise sum of the sequence of the first weighted propagation spatial span values and the sequence of the second weighted propagation spatial span values to obtain a sequence of weighted sum propagation spatial span values; 将所述历史肢体体积局部时序模式能量统计范式值的序列与所述加权总和传播空间跨度值的序列中每组对应的历史肢体体积局部时序模式能量统计范式值与加权总和传播空间跨度值进行相除以得到所述肢体体积局部时序模式能量传播衰减系数值的序列。Divide each group of corresponding historical limb volume local time series pattern energy statistical paradigm values and weighted sum propagation space span values in the sequence of the historical limb volume local time series pattern energy statistical paradigm values and the sequence of the weighted sum propagation space span values to obtain the sequence of the limb volume local time series pattern energy propagation attenuation coefficient values. 5.根据权利要求4所述的淋巴水肿肢体体积测量数据管理系统，其特征在于，所述治疗多模态-肢体体积语义特征融合模块，包括：5. The lymphedema limb volume measurement data management system according to claim 4, characterized in that the treatment multimodal-limb volume semantic feature fusion module comprises: 肢体体积时序语义强化单元，用于基于所述治疗多模态语义融合特征向量，对所述肢体体积时序语义传播推理特征向量进行特征语义交互强化以得到包含治疗多模态语义-肢体体积时序推理语义交互信息的强化肢体体积时序语义传播推理特征向量；A limb volume temporal semantics reinforcement unit, configured to perform feature semantics interaction reinforcement on the limb volume temporal semantics propagation reasoning feature vector based on the treatment multimodal semantics fusion feature vector to obtain a reinforced limb volume temporal semantics propagation reasoning feature vector containing treatment multimodal semantics-limb volume temporal reasoning semantics interaction information; 治疗多模态语义强化单元，用于基于所述肢体体积时序语义传播推理特征向量，对所述治疗多模态语义融合特征向量进行特征语义交互强化以得到包含肢体体积时序推理语义-治疗多模态语义交互信息的强化治疗多模态语义融合特征向量；A treatment multimodal semantic enhancement unit, configured to perform feature semantic interaction enhancement on the treatment multimodal semantic fusion feature vector based on the limb volume temporal semantic propagation reasoning feature vector to obtain an enhanced treatment multimodal semantic fusion feature vector containing limb volume temporal reasoning semantics-treatment multimodal semantic interaction information; 强化治疗多模态语义-肢体体积交互融合单元，用于将所述强化肢体体积时序语义传播推理特征向量和所述强化治疗多模态语义融合特征向量进行级联处理以得到强化治疗多模态语义-肢体体积交互融合特征向量；An intensive therapy multimodal semantic-limb volume interactive fusion unit, used for cascading the intensive limb volume temporal semantic propagation reasoning feature vector and the intensive therapy multimodal semantic fusion feature vector to obtain an intensive therapy multimodal semantic-limb volume interactive fusion feature vector; 交互信息融合响应门控值计算单元，用于将所述强化治疗多模态语义-肢体体积交互融合特征向量输入门控响应函数以得到强化治疗多模态语义-肢体体积交互信息融合响应门控值；An interactive information fusion response gating value calculation unit, used for inputting the intensive treatment multimodal semantics-limb volume interactive fusion feature vector into a gating response function to obtain an intensive treatment multimodal semantics-limb volume interactive information fusion response gating value; 治疗多模态语义门控调整单元，用于计算所述治疗多模态语义融合特征向量与所述强化治疗多模态语义-肢体体积交互信息融合响应门控值之间的按位置乘积以得到治疗多模态语义门控调整向量；a treatment multimodal semantic gating adjustment unit, used for calculating the positional product between the treatment multimodal semantic fusion feature vector and the enhanced treatment multimodal semantic-limb volume interaction information fusion response gating value to obtain a treatment multimodal semantic gating adjustment vector; 肢体体积时序语义门控调整单元，用于计算所述肢体体积时序语义传播推理特征向量与一减去所述强化治疗多模态语义-肢体体积交互信息融合响应门控值之间的按位置乘积以得到肢体体积时序语义门控调整向量；A limb volume temporal semantic gating adjustment unit, used for calculating the positional product between the limb volume temporal semantic propagation reasoning feature vector and one minus the intensive treatment multimodal semantic-limb volume interaction information fusion response gating value to obtain a limb volume temporal semantic gating adjustment vector; 语义动态选择交互融合单元，用于将所述治疗多模态语义门控调整向量与所述肢体体积时序语义门控调整向量进行按位置相加以得到所述治疗多模态语义-肢体体积时序推理语义动态选择交互融合特征向量。The semantic dynamic selection interactive fusion unit is used to add the treatment multimodal semantic gating adjustment vector and the limb volume temporal semantic gating adjustment vector by position to obtain the treatment multimodal semantic-limb volume temporal reasoning semantic dynamic selection interactive fusion feature vector. 6.根据权利要求5所述的淋巴水肿肢体体积测量数据管理系统，其特征在于，所述肢体体积时序语义强化单元，用于：6. The lymphedema limb volume measurement data management system according to claim 5, characterized in that the limb volume temporal semantic enhancement unit is used to: 将第一权重矩阵与所述治疗多模态语义融合特征向量进行相乘以得到治疗多模态语义第一权重向量；Multiplying the first weight matrix by the treatment multimodal semantic fusion feature vector to obtain a treatment multimodal semantic first weight vector; 将所述肢体体积时序语义传播推理特征向量分别与第二权重矩阵和第三权重矩阵进行相乘以得到肢体体积时序语义传播推理第二权重向量和肢体体积时序语义传播推理第三权重向量；Multiplying the limb volume temporal semantic propagation reasoning feature vector with the second weight matrix and the third weight matrix respectively to obtain a limb volume temporal semantic propagation reasoning second weight vector and a limb volume temporal semantic propagation reasoning third weight vector; 计算所述治疗多模态语义第一权重向量的转置向量与所述肢体体积时序语义传播推理第二权重向量之间的乘积以得到治疗多模态-肢体体积关联值后，将所述治疗多模态-肢体体积关联值除以所述肢体体积时序语义传播推理特征向量的长度的平方根以得到治疗多模态-肢体体积权重值；After calculating the product between the transposed vector of the treatment multimodal semantic first weight vector and the limb volume temporal semantic propagation reasoning second weight vector to obtain the treatment multimodal-limb volume association value, the treatment multimodal-limb volume association value is divided by the square root of the length of the limb volume temporal semantic propagation reasoning feature vector to obtain the treatment multimodal-limb volume weight value; 将所述治疗多模态-肢体体积权重值输入softmax函数以得到治疗多模态-肢体体积激活权重值；Inputting the treatment multimodality-limb volume weight value into a softmax function to obtain a treatment multimodality-limb volume activation weight value; 将所述治疗多模态-肢体体积激活权重值作为权重值，对所述肢体体积时序语义传播推理第三权重向量的转置向量进行加权以得到所述包含治疗多模态语义-肢体体积时序推理语义交互信息的强化肢体体积时序语义传播推理特征向量。The treatment multimodal-limb volume activation weight value is used as the weight value, and the transposed vector of the third weight vector of the limb volume temporal semantic propagation reasoning is weighted to obtain the enhanced limb volume temporal semantic propagation reasoning feature vector containing the treatment multimodal semantics-limb volume temporal reasoning semantic interaction information. 7.根据权利要求6所述的淋巴水肿肢体体积测量数据管理系统，其特征在于，所述交互信息融合响应门控值计算单元，用于：7. The lymphedema limb volume measurement data management system according to claim 6, characterized in that the interactive information fusion response gate value calculation unit is used to: 计算所述强化治疗多模态语义-肢体体积交互融合特征向量与参数矩阵的矩阵乘积后，再将得到的特征向量与偏置向量进行按位置相加以得到强化治疗多模态语义-肢体体积交互融合偏置特征向量；After calculating the matrix product of the intensive therapy multimodal semantic-limb volume interactive fusion feature vector and the parameter matrix, the obtained feature vector and the bias vector are added according to the position to obtain the intensive therapy multimodal semantic-limb volume interactive fusion bias feature vector; 将所述强化治疗多模态语义-肢体体积交互融合偏置特征向量输入sigmoid激活函数以得到所述强化治疗多模态语义-肢体体积交互信息融合响应门控值。The intensive therapy multimodal semantic-limb volume interaction fusion bias feature vector is input into a sigmoid activation function to obtain the intensive therapy multimodal semantic-limb volume interaction information fusion response gating value. 8.根据权利要求7所述的淋巴水肿肢体体积测量数据管理系统，其特征在于，所述异常检测结果生成模块，用于：8. The lymphedema limb volume measurement data management system according to claim 7, characterized in that the abnormal detection result generation module is used to: 将所述治疗多模态语义-肢体体积时序推理语义动态选择交互融合特征向量通过基于分类器的淋巴水肿肢体体积异常识别器以得到所述异常检测结果，所述异常检测结果用于表示被监测淋巴水肿肢体患者的淋巴水肿肢体的体积是否存在异常；The treatment multimodal semantics-limb volume temporal reasoning semantics dynamic selection interactive fusion feature vector is passed through a classifier-based lymphedema limb volume abnormality identifier to obtain the abnormality detection result, and the abnormality detection result is used to indicate whether the volume of the lymphedema limb of the monitored lymphedema limb patient is abnormal; 响应于所述异常检测结果为被监测淋巴水肿肢体患者的淋巴水肿肢体的体积存在异常，产生预警提示信号给医护人员。In response to the abnormal detection result that the volume of the lymphedema limb of the monitored lymphedema limb patient is abnormal, an early warning signal is generated to the medical staff. 9.一种淋巴水肿肢体体积测量数据管理方法，其特征在于，包括：9. A method for managing lymphedema limb volume measurement data, comprising: 获取被监测淋巴水肿肢体患者的淋巴水肿肢体体积数据的时间序列；Obtaining a time series of lymphedema limb volume data of a patient with a monitored lymphedema limb; 获取所述被监测淋巴水肿肢体患者的治疗数据，所述治疗数据包括物理治疗数据和药物治疗数据；Acquiring treatment data of the monitored lymphedema limb patient, wherein the treatment data includes physical therapy data and drug therapy data; 将所述淋巴水肿肢体体积数据的时间序列排列为淋巴水肿肢体体积时序输入向量后输入淋巴水肿肢体体积时序模式特征提取器以得到肢体体积局部时序模式特征向量的序列；Arranging the time series of the lymphedema limb volume data into a lymphedema limb volume time series input vector and then inputting the vector into a lymphedema limb volume time series pattern feature extractor to obtain a sequence of limb volume local time series pattern feature vectors; 将所述肢体体积局部时序模式特征向量的序列通过基于节点能量衰减机制的肢体体积时序节点语义传播推理器以得到肢体体积时序语义传播推理特征向量；The sequence of the limb volume local temporal pattern feature vectors is passed through a limb volume temporal node semantic propagation reasoner based on a node energy decay mechanism to obtain a limb volume temporal semantic propagation reasoning feature vector; 对所述被监测淋巴水肿肢体患者的治疗数据进行语义编码以得到物理治疗数据语义编码特征向量和药物治疗数据语义编码特征向量后，融合所述物理治疗数据语义编码特征向量和所述药物治疗数据语义编码特征向量以得到治疗多模态语义融合特征向量；After semantically encoding the treatment data of the monitored lymphedema limb patient to obtain a physical therapy data semantic encoding feature vector and a drug therapy data semantic encoding feature vector, the physical therapy data semantic encoding feature vector and the drug therapy data semantic encoding feature vector are fused to obtain a treatment multimodal semantic fusion feature vector; 将所述治疗多模态语义融合特征向量和所述肢体体积时序语义传播推理特征向量通过基于双向注意力网络的动态特征选择交互融合模块以得到治疗多模态语义-肢体体积时序推理语义动态选择交互融合特征向量作为治疗多模态语义-肢体体积时序推理语义动态选择交互融合特征；The treatment multimodal semantic fusion feature vector and the limb volume temporal semantic propagation reasoning feature vector are passed through a dynamic feature selection interactive fusion module based on a bidirectional attention network to obtain a treatment multimodal semantic-limb volume temporal reasoning semantic dynamic selection interactive fusion feature vector as a treatment multimodal semantic-limb volume temporal reasoning semantic dynamic selection interactive fusion feature; 基于所述治疗多模态语义-肢体体积时序推理语义动态选择交互融合特征，得到异常检测结果，并基于所述异常检测结果，产生相应的预警提示信号给医护人员；Based on the treatment multimodal semantics-limb volume temporal reasoning semantics dynamic selection interactive fusion features, an abnormal detection result is obtained, and based on the abnormal detection result, a corresponding early warning prompt signal is generated to the medical staff; 其中，将所述淋巴水肿肢体体积数据的时间序列排列为淋巴水肿肢体体积时序输入向量后输入淋巴水肿肢体体积时序模式特征提取器以得到肢体体积局部时序模式特征向量的序列，包括：The method of arranging the time series of the lymphedema limb volume data into a lymphedema limb volume time series input vector and then inputting the vector into a lymphedema limb volume time series pattern feature extractor to obtain a sequence of limb volume local time series pattern feature vectors includes: 将所述淋巴水肿肢体体积数据的时间序列按照时间维度排列为所述淋巴水肿肢体体积时序输入向量；Arranging the time series of the lymphedema limb volume data according to the time dimension into the lymphedema limb volume time series input vector; 将所述淋巴水肿肢体体积时序输入向量输入基于1D卷积神经网络模型的淋巴水肿肢体体积时序模式特征提取器以得到所述肢体体积局部时序模式特征向量的序列。The lymphedema limb volume temporal input vector is input into a lymphedema limb volume temporal pattern feature extractor based on a 1D convolutional neural network model to obtain a sequence of the limb volume local temporal pattern feature vectors.
CN202411135190.8A 2024-08-19 2024-08-19 Lymphedema limb volume measurement data management system and method Active CN118658633B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202411135190.8A CN118658633B (en) 2024-08-19 2024-08-19 Lymphedema limb volume measurement data management system and method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202411135190.8A CN118658633B (en) 2024-08-19 2024-08-19 Lymphedema limb volume measurement data management system and method

Publications (2)

Publication Number Publication Date
CN118658633A true CN118658633A (en) 2024-09-17
CN118658633B CN118658633B (en) 2024-11-15

Family

ID=92700591

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202411135190.8A Active CN118658633B (en) 2024-08-19 2024-08-19 Lymphedema limb volume measurement data management system and method

Country Status (1)

Country Link
CN (1) CN118658633B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN119344746A (en) * 2024-12-30 2025-01-24 杭州六极加科技有限公司 Method and system for analyzing abnormal ECG data based on artificial intelligence

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115732068A (en) * 2022-11-23 2023-03-03 上海中医药大学 Lymphedema limb volume measurement data management system and method based on Internet of things
CN117524489A (en) * 2024-01-08 2024-02-06 吉林大学 Patient health monitoring management system and method
CN117694864A (en) * 2023-12-14 2024-03-15 江苏大学附属医院 An intelligent measurement method for lymphedema volume

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115732068A (en) * 2022-11-23 2023-03-03 上海中医药大学 Lymphedema limb volume measurement data management system and method based on Internet of things
CN117694864A (en) * 2023-12-14 2024-03-15 江苏大学附属医院 An intelligent measurement method for lymphedema volume
CN117524489A (en) * 2024-01-08 2024-02-06 吉林大学 Patient health monitoring management system and method

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
温洋: "《基于深度学习的医学影像语义分割技术研究》", 《CNKI硕士学位论文全文库》, 15 March 2024 (2024-03-15) *
赵赛;孟繁洁;: "乳腺癌术后淋巴水肿的测量和评估研究现状", 中华护理杂志, no. 06, 5 June 2013 (2013-06-05), pages 558 - 561 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN119344746A (en) * 2024-12-30 2025-01-24 杭州六极加科技有限公司 Method and system for analyzing abnormal ECG data based on artificial intelligence

Also Published As

Publication number Publication date
CN118658633B (en) 2024-11-15

Similar Documents

Publication Publication Date Title
US20170147777A1 (en) Method and apparatus for predicting health data value through generation of health data pattern
JP2020513615A (en) Decentralized diagnostic workflow training of deep learning neural networks
US20070179769A1 (en) Medical risk stratifying method and system
CN118658633B (en) Lymphedema limb volume measurement data management system and method
KR20180038251A (en) Measuring system machine learning body composition
CN109949280B (en) Image processing method, image processing apparatus, device storage medium, and growth evaluation system
CN111524598A (en) Perioperative complication prediction method and system
JP2022549051A (en) Diagnostic imaging device and method using deep learning model
US12087444B2 (en) Population-level gaussian processes for clinical time series forecasting
KR20190086345A (en) Time series data processing device, health predicting system including the same, and method for operating time series data processing device
Sriram et al. [Retracted] A Smart Solution for Cancer Patient Monitoring Based on Internet of Medical Things Using Machine Learning Approach
CN114343585A (en) Early warning method, device, equipment and storage medium for cognitive and behavioral disorders
CN117582222A (en) Informationized blood glucose monitoring system and informationized blood glucose monitoring method
CN114202772B (en) Reference information generation system and method based on artificial intelligence and intelligent medical treatment
WO2023110477A1 (en) A computer implemented method and a system
CN118571501B (en) A medical emergency warning method based on digital twin technology
Tabarisaadi et al. A deep bayesian ensembling framework for COVID-19 detection using chest CT images
WO2019131255A1 (en) Data processing device, data processing method, and data processing program
CN116895381A (en) Methods, devices, media, and equipment for constructing Alzheimer's disease prediction models
Pivithuru et al. E-patient Card: An Integrated Electronic Health Recording System for Patient
Seshadri et al. Early prediction of Alzheimer’s disease with a multimodal multitask deep learning model
CN114155933A (en) User information prompting system and method based on deep learning analysis and intelligent medical treatment
CN109003680B (en) Epilepsy data statistical method and device
CN111243697A (en) Method and system for judging target object data based on neural network
Harshini et al. Human Stress Detection in and through Sleep using Machine Learning

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant