CN117648613A - Diabetes foot ulcer prediction method and system based on artificial intelligence - Google Patents

Abstract

The invention discloses an artificial intelligence-based diabetic foot ulcer prediction method and system. The invention relates to the technical field of diabetic foot ulcer survival prediction, in particular to an artificial intelligence-based diabetic foot ulcer prediction method and system, wherein the method collects multi-mode diabetic foot ulcer data while collecting image data; the data enhancement processing flow of image data enhancement, numerical value standardization and missing value elimination is adopted, so that the data quality is improved; the lightweight convolutional neural network is adopted to conduct foot ulcer pre-classification, and necessary pre-classification data optimization and reconstruction assistance are provided for subsequent survival prediction; and predicting the occurrence and survival of foot ulcers by adopting a multi-layer perceptron combined with super-parameter optimization, training two predictor models, and optimizing the models by a super-parameter optimization unit.

Description

Diabetes foot ulcer prediction method and system based on artificial intelligence

Technical Field

The invention relates to the technical field of diabetic foot ulcer survival prediction, in particular to an artificial intelligence-based diabetic foot ulcer prediction method and system.

Background

The artificial intelligence-based diabetic foot ulcer prediction method aims at predicting whether a patient is likely to generate diabetic foot ulcers or not by analyzing clinical data and related characteristics of the patient by utilizing machine learning and data analysis technology, recognizing the severity of the diabetic foot ulcers of the patient, and predicting the influence of the foot ulcers on the survival condition of the patient so as to intervene and manage in advance.

However, in the existing diabetic foot ulcer prediction method, the existing diabetic foot ulcer prediction method is only oriented to simple prediction classification of image data, and lacks a technical problem that the further combination processing of the image classification data is carried out, so that the practicability of the foot ulcer classification prediction is to be improved; in the existing diabetic foot ulcer prediction method, multi-mode diabetic foot ulcer data exist, and aiming at different data structures, data types and data contents, the technical problems of data loss, non-uniform data quality and unbalanced data contents inevitably exist; in the existing diabetic foot ulcer prediction method, the existing image-based foot ulcer classification method mostly carries out two classifications, the classification requirement of the prediction of the numerical value data cannot be met, and meanwhile, the image classification cannot effectively provide the technical problem of supporting the numerical value prediction; in the existing diabetic foot ulcer prediction method, the technical problems that in the existing foot ulcer survival rate prediction method flow, whether the survival condition exists or not is only predicted within a certain time, the survival condition of a patient cannot be properly reflected, and the quantity and the dimension of prediction information are to be improved exist.

Disclosure of Invention

Aiming at the problems that in the existing diabetic foot ulcer prediction method, the existing diabetic foot ulcer prediction method only faces to simple prediction classification of image data, and further combination processing is lack of image classification data, so that the practicability of the classification prediction of the foot ulcer is to be improved, the scheme creatively collects multi-mode diabetic foot ulcer data while collecting the image data, and provides basic data support for follow-up application of the survival prediction of the diabetic foot ulcer; aiming at the technical problems of data deficiency, non-uniform data quality and unbalanced data content which are inevitably caused by different data structures, data types and data contents in the existing diabetic foot ulcer prediction method, the scheme creatively adopts the data enhancement processing flow of image data enhancement, numerical value standardization and missing value elimination, improves the data quality and provides good data resources for the following various prediction tasks; aiming at the technical problems that in the existing diabetic foot ulcer prediction method, the existing image-based foot ulcer classification method mostly carries out two classifications, the type requirement of prediction of logarithmic data cannot be met, and meanwhile, the image classification cannot effectively provide support for numerical value prediction, the scheme creatively adopts a lightweight convolutional neural network to carry out foot ulcer pre-classification, divides the foot ulcer image into 4 severity degrees, improves the classification depth, provides necessary pre-classification data optimization and reconstruction assistance for subsequent survival prediction, and improves the usability and fluency of the whole method and system; aiming at the technical problems that in the existing diabetic foot ulcer prediction method, whether the existing foot ulcer survival rate prediction method processes exist, the survival condition of a patient cannot be properly reflected, the prediction information quantity and the dimension are required to be improved, the scheme creatively adopts a multi-layer sensor combined with super-parameter optimization to predict the disease survival of the foot ulcer, and the comprehensive usability, accuracy and interpretability of the diabetic foot ulcer prediction are improved by training two survival rate predictor models facing different time limits and performing model optimization through a super-parameter optimization unit.

The technical scheme adopted by the invention is as follows: the invention provides an artificial intelligence-based diabetic foot ulcer prediction method, which comprises the following steps of:

step S1: collecting data;

step S2: data enhancement processing;

step S3: pre-classification of foot ulcers;

step S4: predicting the occurrence and survival of foot ulcers;

step S5: prediction of diabetic foot ulcers.

Further, in step S1, the data collection is used for collecting a basic data set of diabetic foot ulcer prediction, specifically, from clinical data, obtaining diabetic foot ulcer prediction original data through collection;

the data types of the diabetic foot ulcer original data comprise foot ulcer pre-classification data, foot ulcer classification variable data and foot ulcer onset survival prediction numerical data;

the data structure of the original data of the diabetic foot ulcer specifically comprises image data, classification tag data and numerical data;

further, in step S2, the data enhancement processing is configured to perform optimization enhancement processing on raw data, specifically, perform data enhancement on image data in the diabetic foot ulcer prediction raw data to obtain optimized foot ulcer classification data, and perform normalization and missing value elimination processing on classification tag data and numerical data in the diabetic foot ulcer prediction raw data to obtain optimized foot ulcer onset survival prediction data, and includes the following steps:

Step S21: the method comprises the steps of enhancing image data, wherein the image data is used for preventing a foot ulcer pre-classification model from being excessively fitted, and particularly enhancing the image data of the foot ulcer pre-classification data to obtain optimized foot ulcer classification data;

the step of enhancing the image data comprises the following steps:

step S211: data oversampling, specifically, data oversampling by rotating, inverting, and employing multiple color models;

step S212: contrast improvement, specifically, randomly adjusting the image contrast of the foot ulcer pre-classification data to perform contrast improvement;

step S213: randomly scaling, namely, randomly adjusting the image size of the foot ulcer pre-classification data, wherein the random upper limit threshold of the image size is 224 multiplied by 24 pixels, and performing random scaling;

step S214: reconstructing enhanced data, namely randomly extracting 50% of image data from the pre-classified data of the foot ulcers, and randomly extracting 50% of image data from the enhanced image data obtained through data oversampling, contrast improvement and random scaling to obtain optimized classified data of the foot ulcers, wherein the optimized classified data of the foot ulcers specifically comprise abnormal images of the foot ulcers and images of healthy feet;

Step S22: data normalization, specifically, normalizing variables in the numerical data within the numerical range of [0,1 ];

step S23: deletion value elimination, namely deleting abnormal data with deletion value;

step S24: and data enhancement processing, namely carrying out data enhancement processing through the image data enhancement, the data standardization and the deletion value elimination to obtain optimized foot ulcer classification data and optimized foot ulcer onset survival prediction data.

Further, in step S3, the pre-classification of foot ulcers is used for performing pre-classification prediction of foot ulcers according to the image, specifically, performing pre-classification of foot ulcers according to the optimized foot ulcer classification data by using a lightweight convolutional neural network, so as to obtain intermediate data of pre-classification of foot ulcers;

the lightweight convolutional neural network comprises a convolutional layer, a batch standardization layer, a maximum pooling layer, a random inactivation layer, a full connection layer and a classification output layer;

the convolution layer is used for calculating the characteristic diagram output;

the batch normalization layer is used for normalizing the output characteristic diagrams of the convolution layers in batches and reducing covariate drift;

the maximum pooling layer is used for constructing maximum pooling for feature detection;

The random inactivation layer is used for minimizing excessive fitting and realizing regularization of the lightweight convolutional neural network;

the full connection layer is used for connecting information transmission and feature combination;

the classification output layer is used for calculating and outputting a final classification result;

the step of adopting a lightweight convolutional neural network to conduct foot ulcer pre-classification to obtain foot ulcer pre-classification intermediate data comprises the following steps:

step S31: constructing a convolution layer, specifically constructing six convolution layers, activating by adopting a ReLU activation function, and extracting features of image edges and corner points by calculating the size of an output feature map of the convolution layer, wherein the calculation formula of the size of the output feature map of the convolution layer is as follows:

；

in which W is _out Is a convolution layer output characteristic diagramWidth of H _out Is the height of the convolution layer output feature map, m is the width of the input image sample, n is the height of the input image sample, w is the width of the convolution kernel, h is the height of the convolution kernel, d _w Is the width expansion factor, d _h Is a high expansion factor, p _w Is the horizontal fill size of the input dimension, p _h Is the vertical end fill size of the input dimension, S _w Is a convolution kernel moving step in the horizontal direction, S _h Is the vertical convolution kernel movement stride;

Step S32: constructing a batch normalization layer, namely constructing the batch normalization layer after the convolution layer, carrying out batch normalization operation on the output feature images of the convolution layer, reducing covariate drift through regularization to obtain normalized feature images, and constructing the batch normalization layer;

step S33: constructing a maximum pooling layer, specifically constructing four maximum pooling layers with the sizes of 2 multiplied by 2 and 3 multiplied by 3 and the step length of 2, and constructing the maximum pooling layer;

step S34: constructing a random inactivation layer, namely adding two random inactivation layers after the maximum pooling layer, and immediately deleting part of neurons of the model through the random inactivation layers to construct the random inactivation layer;

step S35: constructing a full-connection layer, specifically after a random inactivation layer;

step S36: constructing a classification output layer, namely calculating classification probability through a softmax classifier to obtain foot ulcer pre-classification output, and constructing the classification output layer, wherein the foot ulcer pre-classification output comprises healthy foot, mild foot ulcer foot, moderate foot ulcer foot and severe foot ulcer foot;

step S37: training a foot ulcer pre-classification Model, namely training the foot ulcer pre-classification Model through the construction convolution layer, the construction batch standardization layer, the construction maximum pooling layer, the construction random inactivation layer, the construction full connection layer and the construction classification output layer to obtain a foot ulcer pre-classification Model _FC ；

Step S38: pre-classification of foot ulcers, particularly usingThe foot ulcer pre-classification Model _FC The method comprises the steps of performing foot ulcer pre-classification to obtain foot ulcer pre-classification intermediate data, wherein the foot ulcer pre-classification intermediate data is used for carrying out classification description on severity of foot ulcers of patients and assisting in predicting survival of the onset of the foot ulcers, and the foot ulcer pre-classification intermediate data specifically comprises foot health types, mild foot ulcers, moderate foot ulcers and severe foot ulcers.

Further, in step S4, the prediction of the survival of the foot ulcer is used for predicting the survival of the foot ulcer, specifically, according to the optimized prediction data of the survival of the foot ulcer, the prediction of the survival of the foot ulcer is performed by adopting a multi-layer sensor combined with super-parameter optimization, so as to obtain prediction reference information of the survival of the foot ulcer;

the multilayer perceptron combining the super-parameter optimization comprises an input layer, a hidden layer, an output layer and a super-parameter optimization unit;

the input layer is used for receiving input characteristics and converting the characteristics into representing vectors;

the hidden layer is used for receiving the representation vector and performing feature combination learning;

the output layer is used for generating prediction output data;

The super-parameter optimizing unit is used for optimizing the model super-parameters of the multi-layer perceptron;

the step of predicting the survival of the foot ulcer by adopting the multilayer perceptron combined with super-parameter optimization to obtain the prediction information of the survival of the foot ulcer comprises the following steps:

step S41: the construction of the input layer specifically comprises the following steps:

step S411: the data reconstruction comprises the steps of combining the foot ulcer pre-classification intermediate data with the optimized foot ulcer onset survival prediction data, eliminating data in the foot ulcer pre-classification intermediate data, which are obtained by the foot ulcer pre-classification, of the foot health type disease data and the optimized foot ulcer onset survival prediction data corresponding to the foot health type disease, and carrying out data weighting reconstruction according to the foot ulcer severity classification condition in the foot ulcer pre-classification intermediate data to obtain reconstructed survival prediction data;

step S412: the characteristic engineering is characterized in that the pearson correlation analysis is adopted to conduct characteristic extraction on the foot ulcer classification variable data in the reconstruction survival prediction data to obtain foot ulcer classification characteristics, and the principal component analysis method is adopted to conduct characteristic extraction on the foot ulcer morbidity survival prediction numerical data in the reconstruction survival prediction data to obtain foot ulcer generation characteristics;

Step S413: the input characteristic conversion is specifically that the classification characteristic of the foot ulcer and the survival characteristic of the foot ulcer are used as input characteristics, and are converted into a representation vector through an embedded layer to obtain an input representation vector, and the input layer is constructed;

step S42: constructing a hidden layer, namely constructing and connecting a three-layer multi-layer classification sensor network, and adopting a ReLU activation function to activate, so as to construct the hidden layer;

vector dimensions of the three-layer multi-layer classification sensor network are 1024, 512 and 256 respectively;

step S43: constructing an output layer, specifically combining the output of the hidden layer to obtain model prediction output data;

step S44: the super parameter optimizing unit is constructed, and specifically comprises the following steps:

step S441: the method comprises the steps of constructing a loss function, specifically constructing a cross entropy loss function, and calculating the following formula:

；

wherein L is _BCE Is a cross entropy loss function, N is the total number of samples, i is the sample index, y _i Is a true class of the classification that is,is a predictive classification;

step S442: constructing an objective function for optimizing the hyper-parameters, in particular applying a Hyperopt library, in combination with the cross entropy loss function L _BCE Constructing and obtaining the target functionA number;

step S443: defining a search space, which is used for designating a search range of each super parameter, specifically, applying a Hyperopt library and defining a super parameter search space;

Step S444: selecting a search algorithm, specifically adopting a random optimization algorithm to search;

step S445: executing a hyper-parameter optimization unit, namely obtaining a hyper-parameter combination of an optimization objective function by executing the hyper-opt library function;

step S446: the super-parameter searching is specifically carried out by executing a random optimization algorithm to obtain an optimized super-parameter combination;

step S45: training the foot ulcer onset survival prediction Model, specifically, training the foot ulcer onset survival prediction Model through the construction input layer, the construction hidden layer, the construction output layer and the construction super-parameter optimization unit to obtain a foot ulcer onset survival prediction Model _DP Model of the predictive Model for the survival of foot ulcers _DP Specifically including a first sub-Model for predicting foot ulcer onset survival _{DP_1} And foot ulcer onset survival prediction second sub-Model _{DP_2} ；

The first sub-Model for predicting the disease and survival of foot ulcers _{DP_1} The method is used for predicting the probability of the critical death of the patient suffering from the foot ulcer within 1-5 years of the onset of the patient;

the second sub-Model for predicting the onset and survival of foot ulcers _{DP_2} The method is used for predicting the probability of the critical death of the patient suffering from the foot ulcer within 6-10 years of the onset of the patient;

Step S46: prediction of foot ulcer onset survival, in particular to a Model adopting the foot ulcer onset survival prediction Model _DP And predicting the occurrence and survival of the foot ulcer, and obtaining the prediction reference information of the occurrence and survival of the foot ulcer.

Further, in step S5, the diabetic foot ulcer prediction is used for comprehensively predicting the survival situation of the foot ulcer of the diabetic patient, specifically, the pre-classification of the foot ulcer is performed to obtain pre-classified intermediate data of the foot ulcer, the reconstruction of the pre-classified intermediate data of the disease occurrence survival of the foot ulcer is assisted, and the prediction of the diabetic foot ulcer is performed by combining the pre-classified intermediate data of the foot ulcer to obtain the prediction information of the comprehensive survival data of the diabetic foot ulcer.

The invention provides an artificial intelligence-based diabetic foot ulcer prediction system, which comprises a data collection module, a data enhancement processing module, a foot ulcer pre-classification module, a foot ulcer onset survival prediction module and a diabetic foot ulcer prediction module;

the data collection module is used for collecting data, obtaining diabetic foot ulcer prediction original data through data collection, and sending the diabetic foot ulcer prediction original data to the data enhancement processing module;

The data enhancement processing module is used for data enhancement processing, obtaining optimized foot ulcer classification data and optimized foot ulcer disease survival prediction data through the data enhancement processing, sending the optimized foot ulcer classification data to the foot ulcer pre-classification module, and sending the optimized foot ulcer disease survival prediction data to the foot ulcer disease survival prediction module;

the foot ulcer pre-classification module is used for pre-classifying foot ulcers, obtaining foot ulcer pre-classification intermediate data through the foot ulcer pre-classification, and sending the foot ulcer pre-classification intermediate data to the foot ulcer onset survival prediction module and the diabetes foot ulcer prediction module;

the foot ulcer disease survival prediction module is used for predicting foot ulcer disease survival, obtaining foot ulcer disease survival prediction reference information through foot ulcer disease survival prediction, and sending the foot ulcer disease survival prediction reference information to the diabetic foot ulcer prediction module;

the diabetic foot ulcer prediction module is used for predicting the diabetic foot ulcer, and the comprehensive data prediction information of the survival of the diabetic foot ulcer is obtained through the prediction of the diabetic foot ulcer.

By adopting the scheme, the beneficial effects obtained by the invention are as follows:

(1) Aiming at the technical problems that in the existing diabetic foot ulcer prediction method, the existing diabetic foot ulcer prediction method only faces to the simple prediction classification of image data, and the image classification data is lack of further combined treatment, so that the practicability of the foot ulcer classification prediction is to be improved;

(2) Aiming at the technical problems of data deficiency, non-uniform data quality and unbalanced data content which are inevitably caused by different data structures, data types and data contents in the existing diabetic foot ulcer prediction method, the scheme creatively adopts the data enhancement processing flow of image data enhancement, numerical value standardization and missing value elimination, improves the data quality and provides good data resources for the following various prediction tasks;

(3) Aiming at the technical problems that in the existing diabetic foot ulcer prediction method, the existing image-based foot ulcer classification method mostly carries out two classifications, the type requirement of prediction of logarithmic data cannot be met, and meanwhile, the image classification cannot effectively provide support for numerical value prediction, the scheme creatively adopts a lightweight convolutional neural network to carry out foot ulcer pre-classification, divides the foot ulcer image into 4 severity degrees, improves the classification depth, provides necessary pre-classification data optimization and reconstruction assistance for subsequent survival prediction, and improves the usability and fluency of the whole method and system;

(4) Aiming at the technical problems that in the existing diabetic foot ulcer prediction method, whether the existing foot ulcer survival rate prediction method processes exist, the survival condition of a patient cannot be properly reflected, the prediction information quantity and the dimension are required to be improved, the scheme creatively adopts a multi-layer sensor combined with super-parameter optimization to predict the disease survival of the foot ulcer, and the comprehensive usability, accuracy and interpretability of the diabetic foot ulcer prediction are improved by training two survival rate predictor models facing different time limits and performing model optimization through a super-parameter optimization unit.

Drawings

FIG. 1 is a schematic flow chart of an artificial intelligence based method for predicting diabetic foot ulcers;

FIG. 2 is a schematic diagram of an artificial intelligence based diabetic foot ulcer prediction system provided by the invention;

FIG. 3 is a flowchart of the data enhancement process in step S2;

FIG. 4 is a schematic flow chart of the pre-classification of foot ulcers in step S3;

fig. 5 is a schematic flow chart of step S4 for predicting survival of foot ulcers.

The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention.

Detailed Description

The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are only some, but not all embodiments of the invention; all other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

In the description of the present invention, it should be understood that the terms "upper," "lower," "front," "rear," "left," "right," "top," "bottom," "inner," "outer," and the like indicate orientation or positional relationships based on those shown in the drawings, merely to facilitate description of the invention and simplify the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and thus should not be construed as limiting the invention.

Referring to fig. 1, the method for predicting diabetic foot ulcer based on artificial intelligence provided by the invention comprises the following steps:

step S1: collecting data;

step S2: data enhancement processing;

step S3: pre-classification of foot ulcers;

step S4: predicting the occurrence and survival of foot ulcers;

step S5: prediction of diabetic foot ulcers.

In a second embodiment, referring to fig. 1 and fig. 2, in step S1, the data collection is used to collect a basic data set of diabetic foot ulcer prediction, specifically, from clinical data, the basic data of diabetic foot ulcer prediction is obtained through collection;

the data types of the diabetic foot ulcer original data comprise foot ulcer pre-classification data, foot ulcer classification variable data and foot ulcer onset survival prediction numerical data;

the data structure of the original data of the diabetic foot ulcer specifically comprises image data, classification tag data and numerical data;

the foot ulcer pre-classification data specifically refers to raw data of foot ulcer pre-classification images, and is used for performing foot ulcer classification prediction;

the foot ulcer classification variable data and the foot ulcer onset survival prediction numerical data are used for predicting foot ulcer onset survival;

The foot ulcer classification variable data specifically comprises sex, diabetic retinopathy, diabetic nephropathy, chronic nephropathy, cardiovascular diseases, peripheral arterial diseases, blood pressure, dyslipidemia, hyperuricemia, obesity, metabolism at the time of admission, treatment with metformin or insulin or other oral or injectable antidiabetic drugs prior to admission, ischemia, infection classification, edema, neuropathy, lesion area, lesion depth, wound healing, fever and wound secretion;

the predictive numerical data for the onset and survival of foot ulcers specifically comprises patient age, patient course, death date of a deceased patient, patient admission year, patient hospitalization time, patient readmission information, hemoglobin information, hematocrit information, leukocyte information, neutrophil information, platelet information, fibrinogen information, C-reactive protein information, blood glucose level information, glycosylated hemoglobin information, urea information, creatinine information, alkali reserve information, sodium potassium information, aspartate aminotransferase information, alanine aminotransferase information, total protein information, albumin information, iron blood disorder information, ferritin information), total cholesterol information, high density lipoprotein information, low density lipoprotein information, triglyceride information, uric acid information and glomerular filtration rate information.

By executing the above operation, aiming at the technical problems that in the existing diabetic foot ulcer prediction method, the existing diabetic foot ulcer prediction method only faces to the simple prediction classification of image data, and the further combination processing of the image classification data is lacking, so that the practicability of the foot ulcer classification prediction is to be improved.

Referring to fig. 1, 2 and 3, in the third embodiment, in step S2, the data enhancement process is used for performing an optimization enhancement process on original data, specifically, performing a data enhancement on image data in the diabetic foot ulcer prediction original data to obtain optimized foot ulcer classification data, and performing a normalization and missing value elimination process on classification tag data and numerical data in the diabetic foot ulcer prediction original data to obtain optimized foot ulcer onset survival prediction data, and the method includes the following steps:

step S21: the method comprises the steps of enhancing image data, wherein the image data is used for preventing a foot ulcer pre-classification model from being excessively fitted, and particularly enhancing the image data of the foot ulcer pre-classification data to obtain optimized foot ulcer classification data;

The step of enhancing the image data comprises the following steps:

step S211: data oversampling, specifically, data oversampling by rotating, inverting, and employing multiple color models;

step S212: contrast improvement, specifically, randomly adjusting the image contrast of the foot ulcer pre-classification data to perform contrast improvement;

step S213: randomly scaling, namely, randomly adjusting the image size of the foot ulcer pre-classification data, wherein the random upper limit threshold of the image size is 224 multiplied by 24 pixels, and performing random scaling;

step S214: reconstructing enhanced data, namely randomly extracting 50% of image data from the pre-classified data of the foot ulcers, and randomly extracting 50% of image data from the enhanced image data obtained through data oversampling, contrast improvement and random scaling to obtain optimized classified data of the foot ulcers, wherein the optimized classified data of the foot ulcers specifically comprise abnormal images of the foot ulcers and images of healthy feet;

step S22: data normalization, specifically, normalizing variables in the numerical data within the numerical range of [0,1 ];

step S23: deletion value elimination, namely deleting abnormal data with deletion value;

Step S24: and data enhancement processing, namely carrying out data enhancement processing through the image data enhancement, the data standardization and the deletion value elimination to obtain optimized foot ulcer classification data and optimized foot ulcer onset survival prediction data.

By executing the above operation, aiming at the problems that in the existing diabetic foot ulcer predicting method, the multi-mode diabetic foot ulcer data exists, and aiming at different data structures, data types and data contents, the technical problems of data deficiency, non-uniform data quality and unbalanced data content inevitably exist, the scheme creatively adopts the data enhancement processing flow of image data enhancement, numerical value standardization and missing value elimination, improves the data quality, and provides good data resources for the following various predicting tasks.

Referring to fig. 1, fig. 2 and fig. 4, in the embodiment, in step S3, the pre-classification of foot ulcers is used for performing pre-classification prediction of foot ulcers according to images, specifically, performing pre-classification of foot ulcers according to the optimized foot ulcer classification data, and obtaining intermediate pre-classification data of foot ulcers by using a lightweight convolutional neural network;

The lightweight convolutional neural network comprises a convolutional layer, a batch standardization layer, a maximum pooling layer, a random inactivation layer, a full connection layer and a classification output layer;

the convolution layer is used for calculating the characteristic diagram output;

the batch normalization layer is used for normalizing the output characteristic diagrams of the convolution layers in batches and reducing covariate drift;

the maximum pooling layer is used for constructing maximum pooling for feature detection;

the random inactivation layer is used for minimizing excessive fitting and realizing regularization of the lightweight convolutional neural network;

the full connection layer is used for connecting information transmission and feature combination;

the classification output layer is used for calculating and outputting a final classification result;

the step of adopting a lightweight convolutional neural network to conduct foot ulcer pre-classification to obtain foot ulcer pre-classification intermediate data comprises the following steps:

step S31: constructing a convolution layer, specifically constructing six convolution layers, activating by adopting a ReLU activation function, and extracting features of image edges and corner points by calculating the size of an output feature map of the convolution layer, wherein the calculation formula of the size of the output feature map of the convolution layer is as follows:

；

in which W is _out Is the width of the output characteristic diagram of the convolution layer, H _out Is the height of the convolution layer output feature map, m is the width of the input image sample, n is the height of the input image sample, w is the width of the convolution kernel, h is the height of the convolution kernel, d _w Is the width expansion factor, d _h Is a high expansion factor, p _w Is a horizontal fill of input dimensionsSize, p _h Is the vertical end fill size of the input dimension, S _w Is a convolution kernel moving step in the horizontal direction, S _h Is the vertical convolution kernel movement stride;

step S32: constructing a batch normalization layer, namely constructing the batch normalization layer after the convolution layer, carrying out batch normalization operation on the output feature images of the convolution layer, reducing covariate drift through regularization to obtain normalized feature images, and constructing the batch normalization layer;

step S33: constructing a maximum pooling layer, specifically constructing four maximum pooling layers with the sizes of 2 multiplied by 2 and 3 multiplied by 3 and the step length of 2, and constructing the maximum pooling layer;

step S34: constructing a random inactivation layer, namely adding two random inactivation layers after the maximum pooling layer, and immediately deleting part of neurons of the model through the random inactivation layers to construct the random inactivation layer;

step S35: constructing a full-connection layer, specifically after a random inactivation layer;

Step S36: constructing a classification output layer, namely calculating classification probability through a softmax classifier to obtain foot ulcer pre-classification output, and constructing the classification output layer, wherein the foot ulcer pre-classification output comprises healthy foot, mild foot ulcer foot, moderate foot ulcer foot and severe foot ulcer foot;

step S37: training a foot ulcer pre-classification Model, namely training the foot ulcer pre-classification Model through the construction convolution layer, the construction batch standardization layer, the construction maximum pooling layer, the construction random inactivation layer, the construction full connection layer and the construction classification output layer to obtain a foot ulcer pre-classification Model _FC ；

Step S38: foot ulcers pre-classification, in particular using the Model of the foot ulcers pre-classification Model _FC Performing foot ulcer pre-classification to obtain foot ulcer pre-classification intermediate data, wherein the foot ulcer pre-classification intermediate data is used for carrying out classification description on severity of foot ulcers of patients and assisting in prediction of occurrence and survival of the foot ulcers, and the foot ulcer pre-classification intermediate data comprises the following specific componentsIncluding foot health, mild foot ulcers, moderate foot ulcers and severe foot ulcers.

By executing the above operations, the existing image-based foot ulcer classification method in the existing diabetes foot ulcer prediction method is mostly subjected to two classifications, the type requirement of the prediction of the numerical data cannot be met, and meanwhile, the image classification cannot effectively provide the technical problem of support for the numerical prediction.

Referring to fig. 1, fig. 2, fig. 3, and fig. 5, in this embodiment, in step S4, the prediction of the survival of the foot ulcer is used for predicting the survival of the foot ulcer, specifically, according to the optimized prediction data of the survival of the foot ulcer, the prediction of the survival of the foot ulcer is performed by using a multi-layer sensor combined with super-parameter optimization, so as to obtain prediction reference information of the survival of the foot ulcer;

the multilayer perceptron combining the super-parameter optimization comprises an input layer, a hidden layer, an output layer and a super-parameter optimization unit;

the input layer is used for receiving input characteristics and converting the characteristics into representing vectors;

the hidden layer is used for receiving the representation vector and performing feature combination learning;

the output layer is used for generating prediction output data;

the super-parameter optimizing unit is used for optimizing the model super-parameters of the multi-layer perceptron;

the step of predicting the survival of the foot ulcer by adopting the multilayer perceptron combined with super-parameter optimization to obtain the prediction information of the survival of the foot ulcer comprises the following steps:

step S41: the construction of the input layer specifically comprises the following steps:

step S411: the data reconstruction comprises the steps of combining the foot ulcer pre-classification intermediate data with the optimized foot ulcer onset survival prediction data, eliminating data in the foot ulcer pre-classification intermediate data, which are obtained by the foot ulcer pre-classification, of the foot health type disease data and the optimized foot ulcer onset survival prediction data corresponding to the foot health type disease, and carrying out data weighting reconstruction according to the foot ulcer severity classification condition in the foot ulcer pre-classification intermediate data to obtain reconstructed survival prediction data;

Step S412: the characteristic engineering is characterized in that the pearson correlation analysis is adopted to conduct characteristic extraction on the foot ulcer classification variable data in the reconstruction survival prediction data to obtain foot ulcer classification characteristics, and the principal component analysis method is adopted to conduct characteristic extraction on the foot ulcer morbidity survival prediction numerical data in the reconstruction survival prediction data to obtain foot ulcer generation characteristics;

the foot ulcer classification characteristics specifically comprise severity classification characteristics of foot ulcers of patients, other disease characteristics of patients, foot ulcer characteristics of patients, medication characteristics of patients and treatment condition characteristics of patients;

the survival characteristics of the foot ulcers specifically comprise medical time sequence characteristics, blood characteristics and medical index characteristics;

step S413: the input characteristic conversion is specifically that the classification characteristic of the foot ulcer and the survival characteristic of the foot ulcer are used as input characteristics, and are converted into a representation vector through an embedded layer to obtain an input representation vector, and the input layer is constructed;

step S42: constructing a hidden layer, namely constructing and connecting a three-layer multi-layer classification sensor network, and adopting a ReLU activation function to activate, so as to construct the hidden layer;

vector dimensions of the three-layer multi-layer classification sensor network are 1024, 512 and 256 respectively;

Step S43: constructing an output layer, specifically combining the output of the hidden layer to obtain model prediction output data;

step S44: the super parameter optimizing unit is constructed, and specifically comprises the following steps:

step S441: the method comprises the steps of constructing a loss function, specifically constructing a cross entropy loss function, and calculating the following formula:

；

wherein L is _BCE Is a cross entropy loss function, N is the total number of samples, i is the sample index, y _i Is a true class of the classification that is,is a predictive classification;

step S442: constructing an objective function for optimizing the hyper-parameters, in particular applying a Hyperopt library, in combination with the cross entropy loss function L _BCE Constructing and obtaining the objective function;

the Hyperopt library, in particular to a Python super-parameter optimization library, constructs the objective function by adopting the fmin (-) function of the Hyperopt library;

step S443: defining a search space, which is used for designating a search range of each super parameter, specifically, applying a Hyperopt library and defining a super parameter search space;

the definition hyper-parameter search space is specifically defined by adopting space parameters of the Hyperopt library;

step S444: selecting a search algorithm, specifically adopting a random optimization algorithm to search;

the random optimization algorithm is specifically defined by adopting the algo parameters of the Hyperopt library, and a search algorithm is selected as the random optimization algorithm by defining the numerical value of the algo parameters as tpe.

Step S445: executing a hyper-parameter optimization unit, namely obtaining a hyper-parameter combination of an optimization objective function by executing the hyper-opt library function;

the super-parameter combination of the optimized objective function specifically comprises a hidden layer number parameter set, a hidden unit number parameter set, a random inactivation parameter set, a learning rate parameter set and a training batch scale parameter set;

the hidden layer number parameter group is specifically [3,4,5,6,7];

the hidden unit number parameter group is specifically [64,128,256,512];

the random inactivation parameter set is specifically [0,0.05,0.1,0.3,0.5];

the learning rate parameter group is specifically [0.00001,0.0001,0.001,0.01,0.1]

The training batch size parameter set is specifically [64,128,256,512];

step S446: the super-parameter searching is specifically carried out by executing a random optimization algorithm to obtain an optimized super-parameter combination;

the optimized super-parameter combination specifically refers to an optimized super-parameter combination with the number of hidden layers set to 4, the number of hidden layer units set to 256, the random inactivation parameter set to 0.3, the learning rate set to 0.0001 and the training batch size set to 128;

step S45: training the foot ulcer onset survival prediction Model, specifically, training the foot ulcer onset survival prediction Model through the construction input layer, the construction hidden layer, the construction output layer and the construction super-parameter optimization unit to obtain a foot ulcer onset survival prediction Model _DP Model of the predictive Model for the survival of foot ulcers _DP Specifically including a first sub-Model for predicting foot ulcer onset survival _{DP_1} And foot ulcer onset survival prediction second sub-Model _{DP_2} ；

The first sub-Model for predicting the disease and survival of foot ulcers _{DP_1} The method is used for predicting the probability of the critical death of the patient suffering from the foot ulcer within 1-5 years of the onset of the patient;

the second sub-Model for predicting the onset and survival of foot ulcers _{DP_2} The method is used for predicting the probability of the critical death of the patient suffering from the foot ulcer within 6-10 years of the onset of the patient;

step S46: prediction of foot ulcer onset survival, in particular to a Model adopting the foot ulcer onset survival prediction Model _DP And predicting the occurrence and survival of the foot ulcer, and obtaining the prediction reference information of the occurrence and survival of the foot ulcer.

By executing the above operations, aiming at the technical problems that in the existing diabetic foot ulcer predicting method, whether the existing foot ulcer survival rate predicting method only predicts survival conditions within a certain time and cannot properly reflect survival conditions of patients, the quantity and dimension of predicted information are required to be improved, the scheme creatively adopts a multi-layer sensor combined with super-parameter optimization to predict the occurrence survival of the foot ulcer, and improves comprehensive usability, accuracy and interpretability of the diabetic foot ulcer prediction by training two survival rate predictor models facing different time limits and performing model optimization through a super-parameter optimizing unit.

In step S5, the diabetic foot ulcer prediction is used for comprehensively predicting the survival situation of the foot ulcer of the diabetic patient, specifically, the pre-classification of the foot ulcer is performed to obtain pre-classified intermediate data of the foot ulcer, the reconstruction of the pre-classified data of the disease occurrence survival prediction of the foot ulcer is assisted, and the prediction of the diabetic foot ulcer is performed by combining the pre-classified intermediate data of the foot ulcer to obtain comprehensive prediction information of the survival data of the diabetic foot ulcer.

An embodiment seven, referring to fig. 1 and fig. 2, based on the above embodiment, the invention provides an artificial intelligence-based diabetic foot ulcer prediction system, which comprises a data collection module, a data enhancement processing module, a foot ulcer pre-classification module, a foot ulcer onset survival prediction module and a diabetic foot ulcer prediction module;

the data collection module is used for collecting data, obtaining diabetic foot ulcer prediction original data through data collection, and sending the diabetic foot ulcer prediction original data to the data enhancement processing module;

the data enhancement processing module is used for data enhancement processing, obtaining optimized foot ulcer classification data and optimized foot ulcer disease survival prediction data through the data enhancement processing, sending the optimized foot ulcer classification data to the foot ulcer pre-classification module, and sending the optimized foot ulcer disease survival prediction data to the foot ulcer disease survival prediction module;

The foot ulcer pre-classification module is used for pre-classifying foot ulcers, obtaining foot ulcer pre-classification intermediate data through the foot ulcer pre-classification, and sending the foot ulcer pre-classification intermediate data to the foot ulcer onset survival prediction module and the diabetes foot ulcer prediction module;

the foot ulcer disease survival prediction module is used for predicting foot ulcer disease survival, obtaining foot ulcer disease survival prediction reference information through foot ulcer disease survival prediction, and sending the foot ulcer disease survival prediction reference information to the diabetic foot ulcer prediction module;

the diabetic foot ulcer prediction module is used for predicting the diabetic foot ulcer, and the comprehensive data prediction information of the survival of the diabetic foot ulcer is obtained through the prediction of the diabetic foot ulcer.

It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.

Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.

The invention and its embodiments have been described above with no limitation, and the actual construction is not limited to the embodiments of the invention as shown in the drawings. In summary, if one of ordinary skill in the art is informed by this disclosure, a structural manner and an embodiment similar to the technical solution should not be creatively devised without departing from the gist of the present invention.

Claims (9)

1. An artificial intelligence-based diabetic foot ulcer prediction method is characterized by comprising the following steps of: the method comprises the following steps:

step S1: collecting data;

step S2: data enhancement processing;

step S3: pre-classification of foot ulcers;

step S4: predicting the occurrence and survival of foot ulcers;

step S5: predicting diabetic foot ulcers;

in step S1, the data collection is used for collecting a basic data set of diabetic foot ulcer prediction, specifically, obtaining diabetic foot ulcer prediction original data from clinical data through collection; the data types of the diabetic foot ulcer original data comprise foot ulcer pre-classification data, foot ulcer classification variable data and foot ulcer onset survival prediction numerical data; the data structure of the original data of the diabetic foot ulcer specifically comprises image data, classification tag data and numerical data;

In step S2, the data enhancement processing is configured to perform optimization enhancement processing on the original data, specifically, perform data enhancement on the image data in the diabetic foot ulcer prediction original data to obtain optimized foot ulcer classification data, and perform normalization and missing value elimination processing on the classification tag data and the numerical data in the diabetic foot ulcer prediction original data to obtain optimized foot ulcer onset survival prediction data;

in step S3, the pre-classification of foot ulcers is used for performing pre-classification prediction of foot ulcers according to images, specifically, optimizing the classification data of foot ulcers, and performing pre-classification of foot ulcers by using a lightweight convolutional neural network to obtain intermediate data of pre-classification of foot ulcers; the lightweight convolutional neural network comprises a convolutional layer, a batch standardization layer, a maximum pooling layer, a random inactivation layer, a full connection layer and a classification output layer;

in step S4, the prediction of the survival of foot ulcer is used for predicting the survival of foot ulcer, specifically, according to the optimized prediction data of the survival of foot ulcer, the prediction of the survival of foot ulcer is performed by adopting a multi-layer sensor combined with super-parameter optimization, so as to obtain prediction reference information of the survival of foot ulcer; the multilayer perceptron combining the super-parameter optimization comprises an input layer, a hidden layer, an output layer and a super-parameter optimization unit;

In step S5, the prediction of the diabetic foot ulcer is used for comprehensively predicting the survival condition of the diabetic foot ulcer, so as to obtain comprehensive prediction information of survival data of the diabetic foot ulcer.

2. The artificial intelligence based diabetic foot ulcer prediction method according to claim 1 wherein: in step S2, the data enhancement process specifically includes the following steps:

step S21: the method comprises the steps of enhancing image data, wherein the image data is used for preventing a foot ulcer pre-classification model from being excessively fitted, and particularly enhancing the image data of the foot ulcer pre-classification data to obtain optimized foot ulcer classification data;

the step of enhancing the image data comprises the following steps:

step S211: data oversampling, specifically, data oversampling by rotating, inverting, and employing multiple color models;

step S212: contrast improvement, specifically, randomly adjusting the image contrast of the foot ulcer pre-classification data to perform contrast improvement;

step S213: randomly scaling, namely, randomly adjusting the image size of the foot ulcer pre-classification data, wherein the random upper limit threshold of the image size is 224 multiplied by 24 pixels, and performing random scaling;

step S214: reconstructing enhanced data, namely randomly extracting 50% of image data from the pre-classified data of the foot ulcers, and randomly extracting 50% of image data from the enhanced image data obtained through data oversampling, contrast improvement and random scaling to obtain optimized classified data of the foot ulcers, wherein the optimized classified data of the foot ulcers specifically comprise abnormal images of the foot ulcers and images of healthy feet;

Step S22: data normalization, specifically, normalizing variables in the numerical data within the numerical range of [0,1 ];

step S23: deletion value elimination, namely deleting abnormal data with deletion value;

step S24: and data enhancement processing, namely carrying out data enhancement processing through the image data enhancement, the data standardization and the deletion value elimination to obtain optimized foot ulcer classification data and optimized foot ulcer onset survival prediction data.

3. The artificial intelligence based diabetic foot ulcer prediction method according to claim 2 wherein: the lightweight convolutional neural network comprises a convolutional layer, a batch standardization layer, a maximum pooling layer, a random inactivation layer, a full connection layer and a classification output layer;

the convolution layer is used for calculating the characteristic diagram output;

the batch normalization layer is used for normalizing the output characteristic diagrams of the convolution layers in batches and reducing covariate drift;

the maximum pooling layer is used for constructing maximum pooling for feature detection;

the random inactivation layer is used for minimizing excessive fitting and realizing regularization of the lightweight convolutional neural network;

the full connection layer is used for connecting information transmission and feature combination;

And the classification output layer is used for calculating a final classification result and outputting the final classification result.

4. The artificial intelligence based diabetic foot ulcer prediction method according to claim 3 wherein: in step S3, the step of performing foot ulcer pre-classification by using a lightweight convolutional neural network to obtain intermediate data of foot ulcer pre-classification includes:

step S31: constructing a convolution layer, specifically constructing six convolution layers, activating by adopting a ReLU activation function, and extracting features of image edges and corner points by calculating the size of an output feature map of the convolution layer, wherein the calculation formula of the size of the output feature map of the convolution layer is as follows:

；

in which W is _out Is the width of the output characteristic diagram of the convolution layer, H _out Is the height of the convolution layer output feature map, m is the width of the input image sample, n is the height of the input image sample, w is the width of the convolution kernel, h is the height of the convolution kernel, d _w Is the width expansion factor, d _h Is a high expansion factor, p _w Is the horizontal fill size of the input dimension, p _h Is the vertical end fill size of the input dimension, S _w Is a convolution kernel moving step in the horizontal direction, S _h Is the vertical convolution kernel movement stride;

step S32: constructing a batch normalization layer, namely constructing the batch normalization layer after the convolution layer, carrying out batch normalization operation on the output feature images of the convolution layer, reducing covariate drift through regularization to obtain normalized feature images, and constructing the batch normalization layer;

Step S33: constructing a maximum pooling layer, specifically constructing four maximum pooling layers with the sizes of 2 multiplied by 2 and 3 multiplied by 3 and the step length of 2, and constructing the maximum pooling layer;

step S34: constructing a random inactivation layer, namely adding two random inactivation layers after the maximum pooling layer, and immediately deleting part of neurons of the model through the random inactivation layers to construct the random inactivation layer;

step S35: constructing a full-connection layer, specifically after a random inactivation layer;

step S36: constructing a classification output layer, namely calculating classification probability through a softmax classifier to obtain foot ulcer pre-classification output, and constructing the classification output layer, wherein the foot ulcer pre-classification output comprises healthy foot, mild foot ulcer foot, moderate foot ulcer foot and severe foot ulcer foot;

step S37: training a foot ulcer pre-classification Model, namely training the foot ulcer pre-classification Model through the construction convolution layer, the construction batch standardization layer, the construction maximum pooling layer, the construction random inactivation layer, the construction full connection layer and the construction classification output layer to obtain a foot ulcer pre-classification Model _FC ；

Step S38: foot ulcers pre-classification, in particular using the Model of the foot ulcers pre-classification Model _FC The method comprises the steps of performing foot ulcer pre-classification to obtain foot ulcer pre-classification intermediate data, wherein the foot ulcer pre-classification intermediate data is used for carrying out classification description on severity of foot ulcers of patients and assisting in predicting survival of the onset of the foot ulcers, and the foot ulcer pre-classification intermediate data specifically comprises foot health types, mild foot ulcers, moderate foot ulcers and severe foot ulcers.

5. The artificial intelligence based diabetic foot ulcer prediction method according to claim 4 wherein: the multilayer perceptron combining the super-parameter optimization comprises an input layer, a hidden layer, an output layer and a super-parameter optimization unit;

the input layer is used for receiving input characteristics and converting the characteristics into representing vectors;

the hidden layer is used for receiving the representation vector and performing feature combination learning;

the output layer is used for generating prediction output data;

the super-parameter optimizing unit is used for optimizing the model super-parameters of the multi-layer perceptron.

6. The artificial intelligence based diabetic foot ulcer prediction method according to claim 5 wherein: in step S4, the step of predicting the survival of the foot ulcer by using the multi-layer sensor combined with super-parameter optimization to obtain the prediction information of the survival of the foot ulcer comprises the following steps:

Step S41: the construction of the input layer specifically comprises the following steps:

step S411: the data reconstruction comprises the steps of combining the foot ulcer pre-classification intermediate data with the optimized foot ulcer onset survival prediction data, eliminating data in the foot ulcer pre-classification intermediate data, which are obtained by the foot ulcer pre-classification, of the foot health type disease data and the optimized foot ulcer onset survival prediction data corresponding to the foot health type disease, and carrying out data weighting reconstruction according to the foot ulcer severity classification condition in the foot ulcer pre-classification intermediate data to obtain reconstructed survival prediction data;

step S412: the characteristic engineering is characterized in that the pearson correlation analysis is adopted to conduct characteristic extraction on the foot ulcer classification variable data in the reconstruction survival prediction data to obtain foot ulcer classification characteristics, and the principal component analysis method is adopted to conduct characteristic extraction on the foot ulcer morbidity survival prediction numerical data in the reconstruction survival prediction data to obtain foot ulcer generation characteristics;

step S413: the input characteristic conversion is specifically that the classification characteristic of the foot ulcer and the survival characteristic of the foot ulcer are used as input characteristics, and are converted into a representation vector through an embedded layer to obtain an input representation vector, and the input layer is constructed;

Step S42: constructing a hidden layer, namely constructing and connecting a three-layer multi-layer classification sensor network, and adopting a ReLU activation function to activate, so as to construct the hidden layer;

step S43: constructing an output layer, specifically combining the output of the hidden layer to obtain model prediction output data;

step S44: the super parameter optimizing unit is constructed, and specifically comprises the following steps:

step S441: the method comprises the steps of constructing a loss function, specifically constructing a cross entropy loss function, and calculating the following formula:

；

wherein L is _BCE Is a cross entropy loss function, N is the total number of samples, i is the sample index, y _i Is a true class of the classification that is,is a predictive classification;

step S442: constructing an objective function for optimizing the hyper-parameters, in particular applying a Hyperopt library, in combination with the cross entropy loss function L _BCE Constructing and obtaining the objective function;

step S443: defining a search space, which is used for designating a search range of each super parameter, specifically, applying a Hyperopt library and defining a super parameter search space;

step S444: selecting a search algorithm, specifically adopting a random optimization algorithm to search;

step S445: executing a hyper-parameter optimization unit, namely obtaining a hyper-parameter combination of an optimization objective function by executing the hyper-opt library function;

Step S446: the super-parameter searching is specifically carried out by executing a random optimization algorithm to obtain an optimized super-parameter combination;

step S45: training the foot ulcer onset survival prediction Model, specifically, training the foot ulcer onset survival prediction Model through the construction input layer, the construction hidden layer, the construction output layer and the construction super-parameter optimization unit to obtain a foot ulcer onset survival prediction Model _DP Model of the predictive Model for the survival of foot ulcers _DP Specifically including a first sub-Model for predicting foot ulcer onset survival _{DP_1} And foot ulcer onset survival prediction second sub-Model _{DP_2} ；

The first sub-Model for predicting the disease and survival of foot ulcers _{DP_1} The method is used for predicting the probability of the critical death of the patient suffering from the foot ulcer within 1-5 years of the onset of the patient;

the second sub-Model for predicting the onset and survival of foot ulcers _{DP_2} The method is used for predicting the probability of the critical death of the patient suffering from the foot ulcer within 6-10 years of the onset of the patient;

step S46: prediction of foot ulcer onset survival, in particular to a Model adopting the foot ulcer onset survival prediction Model _DP And predicting the occurrence and survival of the foot ulcer, and obtaining the prediction reference information of the occurrence and survival of the foot ulcer.

7. The artificial intelligence based diabetic foot ulcer prediction method according to claim 6 wherein: in step S5, the diabetic foot ulcer prediction is used for comprehensively predicting the survival situation of the foot ulcer of the diabetic patient, specifically, the pre-classification of the foot ulcer is performed to obtain pre-classified intermediate data of the foot ulcer, the data reconstruction of the disease occurrence survival prediction of the foot ulcer is assisted, and the diabetic foot ulcer prediction is performed by combining the pre-classified intermediate data of the foot ulcer to obtain comprehensive prediction information of the survival data of the diabetic foot ulcer.

8. An artificial intelligence based diabetic foot ulcer prediction system for implementing an artificial intelligence based diabetic foot ulcer prediction method according to any one of claims 1 to 7 wherein: the system comprises a data collection module, a data enhancement processing module, a foot ulcer pre-classification module, a foot ulcer onset survival prediction module and a diabetic foot ulcer prediction module.

9. The artificial intelligence based diabetic foot ulcer prediction system according to claim 8 wherein: the data collection module is used for collecting data, obtaining diabetic foot ulcer prediction original data through data collection, and sending the diabetic foot ulcer prediction original data to the data enhancement processing module;

the data enhancement processing module is used for data enhancement processing, obtaining optimized foot ulcer classification data and optimized foot ulcer disease survival prediction data through the data enhancement processing, sending the optimized foot ulcer classification data to the foot ulcer pre-classification module, and sending the optimized foot ulcer disease survival prediction data to the foot ulcer disease survival prediction module;

the foot ulcer pre-classification module is used for pre-classifying foot ulcers, obtaining foot ulcer pre-classification intermediate data through the foot ulcer pre-classification, and sending the foot ulcer pre-classification intermediate data to the foot ulcer onset survival prediction module and the diabetes foot ulcer prediction module;

The foot ulcer disease survival prediction module is used for predicting foot ulcer disease survival, obtaining foot ulcer disease survival prediction reference information through foot ulcer disease survival prediction, and sending the foot ulcer disease survival prediction reference information to the diabetic foot ulcer prediction module;

the diabetic foot ulcer prediction module is used for predicting the diabetic foot ulcer, and the comprehensive data prediction information of the survival of the diabetic foot ulcer is obtained through the prediction of the diabetic foot ulcer.