Disclosure of Invention
Based on the above, the present invention is needed to provide a platform and a method for predicting and analyzing the therapeutic effect of femoral head necrosis based on big data, so as to solve at least one of the above technical problems.
In order to achieve the above purpose, the femoral head necrosis treatment effect prediction and analysis platform based on big data comprises the following modules:
The femoral head necrosis file generation module is used for acquiring multi-source medical data for femoral head necrosis treatment, wherein the multi-source medical data comprises basic information of a patient, clinical symptoms of the femoral head, femoral head medical images and a femoral head necrosis treatment scheme, and performing data standard cleaning and medical file integration on the multi-source medical data for femoral head necrosis treatment to generate a femoral head necrosis treatment data file;
the femoral head image feature analysis module is used for carrying out femoral head feature analysis on corresponding femoral head medical images in the femoral head necrosis treatment data file so as to obtain femoral head necrosis tissue image features, wherein the femoral head necrosis tissue features comprise a femoral head necrosis tissue structure, a femoral head necrosis tissue form and a femoral head necrosis tissue density;
The femoral head treatment characteristic analysis module is used for carrying out femoral head treatment stage on the corresponding femoral head treatment scheme based on the corresponding femoral head clinical symptoms in the femoral head necrosis treatment data file so as to obtain a femoral head necrosis treatment sub-scheme corresponding to each femoral head symptom stage;
The treatment effect prediction analysis module is used for screening treatment effect influence characteristics of the femoral head necrosis tissue image characteristics and the femoral head necrosis treatment basic characteristics to obtain significant femoral head necrosis treatment effect influence characteristics, constructing a femoral head necrosis treatment effect prediction model, inputting the significant femoral head necrosis treatment effect influence characteristics into the femoral head necrosis treatment effect prediction model for treatment effect prediction analysis so as to predict and output the corresponding femoral head necrosis treatment effects under different treatment schemes.
Further, the femoral head necrosis file generation module comprises the following functions:
acquiring multi-source medical data for treating femoral head necrosis, wherein the multi-source medical data comprise basic information of a patient, clinical symptoms of the femoral head, a femoral head medical image and a femoral head necrosis treatment scheme, the basic information of the patient comprises age, sex, past medical history of the femoral head and medical use history corresponding to the patient, the clinical symptoms of the femoral head comprise pain degree of the femoral head and limited range of movement of the femoral head, the medical image of the femoral head comprises an X-ray image, a CT image and an MRI image corresponding to the femoral head, and the femoral head necrosis treatment scheme comprises a treatment method, a medicine use dosage and an operation record;
Carrying out data standard cleaning on the multi-source medical data for treating the femoral head necrosis to remove corresponding noise data, repeated data and abnormal data, correcting labeling errors corresponding to the image for the image data, carrying out image coding standardization by adopting a unified medical image transmission standard DICOM, and carrying out standardization and deleting repeated medical examination records for the text data by adopting natural language processing to obtain multi-source standard data for treating the femoral head necrosis;
And performing association matching and integration file integration on the multi-source data of the femoral head necrosis treatment corresponding to the same patient in the multi-source standard data of the femoral head necrosis treatment to generate a femoral head necrosis treatment data file.
Further, the femoral head image feature analysis module comprises the following functions:
Performing image physical characteristic deep analysis on corresponding femoral head medical images in a femoral head necrosis treatment data file so as to research attenuation rules corresponding to X-rays when the X-rays pass through femoral head tissues for the X-ray images, analyzing the relation between gray values and tissue densities in the images according to the difference of the X-ray absorption degrees of different tissues, deeply analyzing the femoral head tissue characteristics reflected by voxel values of different layers for CT images, and researching hydrogen proton densities and relaxation times corresponding to magnetic resonance signals and the femoral head tissues for MRI images to obtain a femoral head medical image physical characteristic data set;
Performing femoral head necrosis tissue region segmentation on the femoral head medical image to generate a femoral head necrosis tissue region segmentation image;
Performing multi-scale texture feature analysis on the segmented image of the femoral head necrosis tissue region under different scales based on wavelet transformation to obtain a texture feature set of the femoral head necrosis tissue region;
And carrying out femoral head feature analysis on the femoral head necrosis tissue region texture feature set based on the femoral head medical image physical feature data set so as to invert the tissue structure corresponding to the femoral head necrosis tissue according to the corresponding gray scale and the signal intensity in the image, calculating the density distribution corresponding to the femoral head necrosis tissue by utilizing the known relation between the X-ray attenuation coefficient and the tissue density and combining the image voxel value, and simultaneously, inverting the morphological features corresponding to the femoral head necrosis tissue according to the relation between the T1 and T2 relaxation time and the tissue morphology so as to obtain the femoral head necrosis tissue image feature, wherein the femoral head necrosis tissue image feature comprises the femoral head necrosis tissue structure, the femoral head necrosis tissue morphology and the femoral head necrosis tissue density.
Further, the multi-scale texture feature analysis is performed on the segmented image of the femoral head necrosis tissue region under different scales based on wavelet transformation, specifically, fine texture features corresponding to the femoral head tissue are extracted through high-frequency subband coefficients corresponding to wavelet transformation under small scales, the fine texture features comprise microstructure textures corresponding to bone trabeculae, and the texture roughness and self-similarity corresponding to the whole femoral head are calculated and analyzed by fractal dimension under large scales.
Further, the femoral head treatment characteristic analysis module comprises the following functions:
Performing femoral head treatment staging on the corresponding femoral head necrosis treatment plan based on the corresponding femoral head pain degree and the limited range of femoral head movement in the femoral head necrosis treatment data file to obtain corresponding femoral head necrosis treatment sub-plans under each femoral head symptom staging;
Based on the corresponding femoral head past medical history and medicine use history in the basic information of the patient, carrying out necrosis treatment and allergy probability analysis on the corresponding femoral head necrosis treatment sub-cases under each femoral head symptom stage so as to obtain the femoral head necrosis treatment probability and the femoral head necrosis treatment allergy probability;
Performing necrotic tissue attenuation evaluation analysis on the corresponding femoral head necrosis treatment sub-cases under each femoral head symptom stage based on the femoral head necrosis tissue density to obtain the femoral head necrosis tissue treatment attenuation efficiency;
the femoral head necrosis treatment probability, the femoral head necrosis treatment allergy probability and the femoral head necrosis tissue treatment attenuation efficiency are used as basic characteristics to obtain the basic characteristics of the femoral head necrosis treatment.
Further, the symptom stage of the femoral head is specifically an early stage with the pain degree of the femoral head between 1 and 3 minutes and the limited range of the movement of the femoral head between 10 and 20 percent, comprising the reduction of the forward flexion movement angle of the hip joint from a normal range of 120 to 140 degrees to a range of 100 to 120 degrees, the reduction of the abduction movement angle of the hip joint from a normal range of 30 to 45 degrees to 25 to 35 degrees, and the reduction of the internal rotation movement angle from a normal range of 30 to 40 degrees to 20 to 30 degrees, an intermediate stage with the pain degree of the femoral head between 4 and 6 minutes and the expansion of the limited range of the femoral head between 20 and 40 percent, comprising the reduction of the forward flexion movement angle of the hip joint to 80 to 100 degrees, the reduction of the abduction movement angle of the hip joint to 20 to 25 degrees, and the reduction of the internal rotation movement angle to 10 to 20 degrees, and an late stage with the pain degree of the femoral head between 7 and 10 minutes and the expansion of the limited range of the movement of the femoral head exceeding 40 percent, comprising the forward flexion movement angle of the hip joint less than 80 degrees, the abduction movement angle of the femoral head less than 20 degrees, and the internal rotation angle of the internal rotation angle less than 10 degrees.
Further, the necrosis tissue attenuation evaluation analysis of the femoral head necrosis treatment sub-cases corresponding to each femoral head symptom stage based on the femoral head necrosis tissue density comprises:
performing tissue density treatment simulation on the corresponding femoral head necrosis tissue density based on the corresponding femoral head necrosis treatment sub-plan under each femoral head symptom stage so as to generate a corresponding density attenuation process of the femoral head necrosis tissue under each treatment effect;
Carrying out actual attenuation statistics on the density attenuation process corresponding to the femoral head necrosis tissue under each treatment effect to obtain the actual attenuation of the tissue density corresponding to the femoral head necrosis tissue under each treatment effect;
Obtaining the theoretical maximum attenuation of the femoral head necrosis tissue, and carrying out necrosis tissue attenuation quantification calculation on the actual attenuation of the tissue density corresponding to the femoral head necrosis tissue under each treatment effect based on the theoretical maximum attenuation of the femoral head necrosis tissue, thereby obtaining the treatment attenuation efficiency of the femoral head necrosis tissue.
Further, the treatment effect prediction analysis module comprises the following functions:
Performing treatment effect association mining analysis on the femoral head necrosis tissue image characteristics and the femoral head necrosis characteristic factors in the femoral head necrosis treatment basic characteristics to obtain association relations between the femoral head necrosis characteristic factors and the treatment effects;
performing relevance evaluation on the image features of the femoral head necrosis tissues and the femoral head necrosis feature factors in the basic femoral head necrosis treatment features based on the association relation between the femoral head necrosis feature factors and the treatment effects to obtain feature relevance factors between the femoral head necrosis feature factors and the treatment effects;
Carrying out causal deducing on the treatment effect of each femoral head necrosis characteristic factor in the femoral head necrosis tissue image characteristics and the femoral head necrosis treatment basic characteristics to obtain a potential causal relation coefficient between each femoral head necrosis characteristic factor and the treatment effect;
based on characteristic correlation factors and potential causal relation coefficients between the characteristic factors and the treatment effects of the femoral head necrosis, carrying out treatment effect influence characteristic screening on the image characteristics and the basic treatment characteristics of the femoral head necrosis so as to obtain significant treatment effect influence characteristics of the femoral head necrosis;
And constructing a femoral head necrosis treatment effect prediction model by using a convolutional neural network, inputting significant characteristics of the femoral head necrosis treatment effect influence into the femoral head necrosis treatment effect prediction model for treatment effect prediction analysis so as to predict and output the corresponding femoral head necrosis treatment effects under different treatment schemes.
Further, the femoral head necrosis treatment effect prediction model is specifically a pyramid network architecture for extracting treatment effect characteristics on multiple scales, the architecture extracts detail characteristics corresponding to the femoral head by using convolution kernels corresponding to 3x3 and pooling windows of 1x1 in a shallow network, including micro trabecular changes and early necrosis area boundaries, extracts integral characteristics corresponding to the femoral head by using convolution kernels corresponding to 5x5 and pooling windows of 1x1 in a deep network, including joint morphology and necrosis area macroscopic distribution, and takes activation energy and information transfer energy of each neuron in the network into an optimization target through introducing energy constraint to pay attention to prediction errors corresponding to the model, and simultaneously improves prediction performance corresponding to the model by using cross verification so as to predict and output corresponding femoral head necrosis treatment effects at different network layers.
Furthermore, the invention also provides a femoral head necrosis treatment effect prediction and analysis method based on big data, which is implemented based on the femoral head necrosis treatment effect prediction and analysis platform based on big data, and comprises the following steps:
Acquiring multi-source medical data for treating femoral head necrosis, wherein the multi-source medical data comprises basic information of a patient, clinical symptoms of the femoral head, medical images of the femoral head and a femoral head necrosis treatment scheme, and performing data standard cleaning and medical file integration on the multi-source medical data for treating femoral head necrosis to generate a femoral head necrosis treatment data file;
Performing femoral head feature analysis on corresponding femoral head medical images in the femoral head necrosis treatment data file to obtain femoral head necrosis tissue image features including a femoral head necrosis tissue structure, a femoral head necrosis tissue form and a femoral head necrosis tissue density;
Performing femoral head necrosis treatment stage on the corresponding femoral head necrosis treatment scheme based on the corresponding femoral head clinical symptoms in the femoral head necrosis treatment data file to obtain a femoral head necrosis treatment sub-scheme corresponding to each femoral head symptom stage;
The method comprises the steps of carrying out treatment effect influence characteristic screening on femoral head necrosis tissue image characteristics and basic femoral head necrosis treatment characteristics to obtain significant femoral head necrosis treatment effect influence characteristics, constructing a femoral head necrosis treatment effect prediction model, inputting the significant femoral head necrosis treatment effect influence characteristics into the femoral head necrosis treatment effect prediction model for treatment effect prediction analysis so as to predict and output the corresponding femoral head necrosis treatment effects under different treatment schemes.
The invention has the beneficial effects that:
compared with the prior art, the application has the beneficial effects that the multi-source medical data for treating the femoral head necrosis is collected and arranged to cover basic information, clinical symptoms, medical images and treatment schemes of patients, and the data are subjected to standardized cleaning, wherein the diversity of the data comprises the basic information such as age, sex, medical history, life habit and the like of the patients, and the clinical symptoms such as pain, the method has the advantages that the redundant data can be removed, the missing data can be repaired and the data formats of different sources can be standardized through standard cleaning, so that the method has consistency and comparability, the integration of medical files can ensure the complete record of the treatment process and related information of each patient, a solid foundation is established for the subsequent data analysis and model in the process, more accurate and comprehensive patient information can be provided through the integrated and cleaned data, and a real and effective basis is provided for the subsequent analysis of the femoral head necrosis treatment effect, so that the accuracy and reliability of the final treatment prediction result are ensured. Secondly, through carrying out detailed characteristic analysis on the femoral head medical images in the femoral head necrosis treatment data file, the purpose is to extract the image characteristics of the femoral head necrosis, wherein the image characteristics comprise the structure, the shape, the density and the like of the femoral head necrosis tissue, and quantitative description of the femoral head pathological changes can be realized through an imaging technology, such as the shape change of the femoral head necrosis area, the damage condition of the tissue and the density change of the femoral head necrosis area. The analysis can reveal the image representation of the femoral head necrosis at different stages, and can provide quantitative support for the subsequent treatment effect prediction. Then, through analysis of clinical symptoms of the patient, the treatment scheme of the femoral head necrosis can be divided into different stages or sub-schemes, for example, early femoral head necrosis only needs conservation treatment, and later femoral head necrosis only needs surgical intervention, through further analysis of basic information (such as age, sex and the like) and image characteristics (such as tissue density) of the patient, the most suitable treatment scheme can be selected for each treatment stage, the treatment strategy is individually adjusted, and the characteristic analysis is not only helpful for formulating a staged treatment scheme, but also reflecting treatment effects under different symptom stages, and the analysis process can help doctors to accurately evaluate the current corresponding treatment state, so that individual treatment schemes are provided for patients in different stages, and errors of treatment effect estimation of different patients in the subsequent process can be reduced. Finally, by analyzing the factors affecting the treatment of the femoral head necrosis, the characteristics which have obvious influence on the treatment effect are screened out, and a prediction model of the treatment effect of the femoral head necrosis is established, wherein the treatment effect of the femoral head necrosis is affected by various factors, such as age, sex, treatment scheme, disease progression stage, image characteristics and the like of a patient, and the characteristics which obviously affect the treatment effect are screened out, so that the accuracy of the treatment effect prediction is improved, the prediction model can be trained based on historical data, and an efficient prediction system is established by using a machine learning or statistical method, and the effects of different treatment schemes in a specific patient group can be predicted in advance through the model.
Detailed Description
The following description of the technical platform of the present invention, taken in conjunction with the accompanying drawings, will be clearly and fully described, given by way of illustration of 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 present invention without making any inventive effort, are intended to fall within the scope of the present invention.
Furthermore, the drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus a repetitive description thereof will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. The functional entities may be implemented in software or in one or more hardware modules or integrated circuits or in different networks and/or processor platforms and/or microcontroller platforms.
It will be understood that, although the terms "first," "second," etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another element. For example, a first element could be termed a second element, and, similarly, a second element could be termed a first element, without departing from the scope of example embodiments. The term "and/or" as used herein includes any and all combinations of one or more of the associated listed items.
In order to achieve the above objective, referring to fig. 1 to 3, the present invention provides a femoral head necrosis treatment effect prediction and analysis platform based on big data, the platform comprises the following modules:
The femoral head necrosis file generation module is used for acquiring multi-source medical data for femoral head necrosis treatment, wherein the multi-source medical data comprises basic information of a patient, clinical symptoms of the femoral head, femoral head medical images and a femoral head necrosis treatment scheme, and performing data standard cleaning and medical file integration on the multi-source medical data for femoral head necrosis treatment to generate a femoral head necrosis treatment data file;
the femoral head image feature analysis module is used for carrying out femoral head feature analysis on corresponding femoral head medical images in the femoral head necrosis treatment data file so as to obtain femoral head necrosis tissue image features, wherein the femoral head necrosis tissue features comprise a femoral head necrosis tissue structure, a femoral head necrosis tissue form and a femoral head necrosis tissue density;
The femoral head treatment characteristic analysis module is used for carrying out femoral head treatment stage on the corresponding femoral head treatment scheme based on the corresponding femoral head clinical symptoms in the femoral head necrosis treatment data file so as to obtain a femoral head necrosis treatment sub-scheme corresponding to each femoral head symptom stage;
The treatment effect prediction analysis module is used for screening treatment effect influence characteristics of the femoral head necrosis tissue image characteristics and the femoral head necrosis treatment basic characteristics to obtain significant femoral head necrosis treatment effect influence characteristics, constructing a femoral head necrosis treatment effect prediction model, inputting the significant femoral head necrosis treatment effect influence characteristics into the femoral head necrosis treatment effect prediction model for treatment effect prediction analysis so as to predict and output the corresponding femoral head necrosis treatment effects under different treatment schemes.
In the embodiment of the present invention, please refer to fig. 1, which is a schematic block diagram of a femoral head necrosis treatment effect prediction and analysis platform based on big data according to the present invention, in this example, the femoral head necrosis treatment effect prediction and analysis platform based on big data includes the following blocks:
The system comprises a femoral head necrosis file generation module, a femoral head necrosis treatment data file generation module and a femoral head necrosis treatment data file generation module, wherein the femoral head necrosis file generation module is used for acquiring multi-source femoral head necrosis treatment medical data, including basic patient information, femoral head clinical symptoms, femoral head medical images and femoral head necrosis treatment schemes, and performing data standard cleaning and medical file integration on the multi-source femoral head necrosis treatment medical data to generate a femoral head necrosis treatment data file;
In the embodiment of the invention, by collecting multi-source medical data for femoral head necrosis treatment from a plurality of channels, basic patient information is acquired from a hospital registration system and a medical history system, including age, gender, past medical history and medical use history, and femoral head clinical symptoms are recorded by doctors through inquiry and physical examination, including pain degree and limited range of movement. The femoral head medical image is obtained from an archiving system of an image department, and comprises X-ray, CT and MRI images, a femoral head necrosis treatment scheme is obtained from a treatment record database, the treatment method, the drug dosage and the operation record are included, the collected data are identified and noise, repetition and abnormal data are removed by adopting a statistical analysis method, the image data are coded according to the DICOM standard, marking errors are corrected, text data are normalized by using a natural language processing technology, repeated check records are deleted, finally, various data of the same patient are associated and integrated by taking a unique patient identifier as an index, and the data are stored in a special database, so that a femoral head necrosis treatment data file is finally formed.
S2, a femoral head image feature analysis module, which is used for carrying out femoral head feature analysis on the corresponding femoral head medical images in the femoral head necrosis treatment data file so as to obtain femoral head necrosis tissue image features, wherein the femoral head necrosis tissue features comprise a femoral head necrosis tissue structure, a femoral head necrosis tissue form and a femoral head necrosis tissue density;
In the embodiment of the invention, through analyzing X-ray, CT and MRI images in femoral head necrosis treatment data files, establishing a relation between gray values and tissue densities according to attenuation rules of X-rays passing through different tissues for the X-ray images, analyzing tissue structures, analyzing voxel values of different layers in CT image aspect, determining tissue characteristics by combining medical knowledge, researching the relation between magnetic resonance signals, hydrogen proton density and relaxation time by MRI images, carrying out necrotic tissue region segmentation on the images by adopting threshold segmentation and morphological operation, extracting bone trabecular microstructure textures by wavelet transformation high-frequency subband coefficients under small scale, analyzing overall texture roughness and self-similarity by fractal dimension under large scale, inverting tissue structures, calculating density distribution and deducing morphological characteristics according to physical characteristics and texture characteristics of the images, and finally obtaining the femoral head necrosis tissue image characteristics.
S3, a femoral head treatment characteristic analysis module, which is used for carrying out femoral head treatment stage on the corresponding femoral head necrosis treatment scheme based on the corresponding femoral head clinical symptoms in the femoral head necrosis treatment data file so as to obtain a femoral head necrosis treatment sub-scheme corresponding to each femoral head symptom stage;
In the embodiment of the invention, the treatment scheme is divided into early stage, middle stage and late stage according to the clinical symptoms of the femoral head in the femoral head necrosis treatment data file, such as pain degree and limited range of movement, wherein the early stage is 1-3 minutes of pain degree and the limited range of movement is 10-20%, the middle stage is 4-6 minutes of pain degree and the limited range of movement is 20-40%, the late stage is 7-10 minutes of pain degree and the limited range of movement exceeds 40%, treatment sub-cases are formulated for each stage, basic information of patients such as age, past medical history and medicine use history and femoral head necrosis tissue density are combined, the treatment probability and allergy probability of the treatment sub-cases are analyzed by using a big data analysis and probability statistical model, tissue initial and final density are obtained through medical images, and the treatment probability, allergy probability and attenuation efficiency are used as basic characteristics, so that the basic characteristics of femoral head necrosis treatment are finally obtained.
And S4, a treatment effect prediction analysis module is used for screening treatment effect influence characteristics of the femoral head necrosis tissue image characteristics and the femoral head necrosis treatment basic characteristics to obtain significant femoral head necrosis treatment effect influence characteristics, constructing a femoral head necrosis treatment effect prediction model, and inputting the significant femoral head necrosis treatment effect influence characteristics into the femoral head necrosis treatment effect prediction model for treatment effect prediction analysis so as to predict and output the corresponding femoral head necrosis treatment effects under different treatment schemes.
In the embodiment of the invention, screening standards are set for image features (tissue structure, morphology and density) of femoral head necrosis tissues and basic features (treatment probability, allergy probability and attenuation efficiency) of femoral head necrosis treatment, such as feature correlation factors with absolute values larger than 0.5 and potential causality coefficients larger than 2, significant features are screened out, a convolutional neural network with a pyramid network architecture is constructed by using a Python and a deep learning framework as a prediction model, detail features are extracted from a shallow layer by using a 3x3 convolution kernel and a 1x1 pooling window, integral features are extracted from a deep layer by using a 5x5 convolution kernel and a 1x1 pooling window, neuron activation and information transmission energy are brought into an optimization target by introducing energy constraint, cross verification is adopted, a score dataset training and verification model is adopted, the screened significant features are input into the model, and femoral head necrosis treatment effects under different treatment schemes such as pain relieving degree and joint function recovery conditions are predicted.
Further, the femoral head necrosis file generation module comprises the following functions:
acquiring multi-source medical data for treating femoral head necrosis, wherein the multi-source medical data comprise basic information of a patient, clinical symptoms of the femoral head, a femoral head medical image and a femoral head necrosis treatment scheme, the basic information of the patient comprises age, sex, past medical history of the femoral head and medical use history corresponding to the patient, the clinical symptoms of the femoral head comprise pain degree of the femoral head and limited range of movement of the femoral head, the medical image of the femoral head comprises an X-ray image, a CT image and an MRI image corresponding to the femoral head, and the femoral head necrosis treatment scheme comprises a treatment method, a medicine use dosage and an operation record;
Carrying out data standard cleaning on the multi-source medical data for treating the femoral head necrosis to remove corresponding noise data, repeated data and abnormal data, correcting labeling errors corresponding to the image for the image data, carrying out image coding standardization by adopting a unified medical image transmission standard DICOM, and carrying out standardization and deleting repeated medical examination records for the text data by adopting natural language processing to obtain multi-source standard data for treating the femoral head necrosis;
And performing association matching and integration file integration on the multi-source data of the femoral head necrosis treatment corresponding to the same patient in the multi-source standard data of the femoral head necrosis treatment to generate a femoral head necrosis treatment data file.
As an embodiment of the present invention, referring to fig. 2, a functional flow diagram of the femoral head necrosis file generating module in fig. 1 is shown, and in this embodiment, the femoral head necrosis file generating module includes the following functions:
S11, acquiring multi-source medical data for treating femoral head necrosis, wherein the multi-source medical data comprise basic information of a patient, clinical symptoms of the femoral head, a femoral head medical image and a femoral head necrosis treatment scheme, the basic information of the patient comprises age, sex, past medical history of the femoral head and medical use history corresponding to the patient, the clinical symptoms of the femoral head comprise the pain degree of the femoral head and the limited range of movement of the femoral head, the medical image of the femoral head comprises an X-ray image, a CT image and an MRI image corresponding to the femoral head, and the femoral head necrosis treatment scheme comprises a treatment method, a medicine use dosage and an operation record;
In the embodiment of the invention, the related data of femoral head necrosis treatment is collected from a plurality of medical data sources, basic information of a patient is obtained from an electronic medical record system of a hospital, the system can record the age, sex, past medical history of the femoral head and medical use history of the patient in detail, the clinical symptoms of the femoral head are recorded through doctor's inquiry and physical examination of the patient, the pain degree of the femoral head is determined by using a pain score scale, the limited range of femoral head movement is defined through measuring the movement angle of the hip joint, the femoral head medical image data is obtained from an image archiving and communication system (PACS) of the hospital, the system stores X-ray images, CT images and MRI images of the patient, the data of a femoral head necrosis treatment scheme is obtained from a treatment record database of the hospital, the information including treatment methods (such as conservative treatment, surgical treatment and the like), the dosage of medicines and the surgical record and the like, and the data obtained from different data sources are integrated, and finally the multi-source medical data of femoral head necrosis treatment is formed.
S12, carrying out data standard cleaning on multi-source medical data for femoral head necrosis treatment to remove corresponding noise data, repeated data and abnormal data, correcting marking errors corresponding to images and carrying out image coding standardization by adopting a unified medical image transmission standard DICOM for image data, and carrying out standardization and deleting repeated medical examination records by adopting natural language processing for text data to obtain multi-source standard data for femoral head necrosis treatment;
In the embodiment of the invention, the cleaning operation is carried out on the multi-source medical data for femoral head necrosis treatment, the numerical data such as patient age, medicine dosage and the like are identified and removed by setting a reasonable value range, meanwhile, the data record is compared, repeated data are deleted, the noise data is processed by adopting a filtering algorithm such as median filtering, random interference in the data is removed, the image data is arranged, professional medical image personnel are arranged for auditing the image annotation, the annotation error is corrected, then all the image data are encoded according to a medical image transmission standard DICOM, the uniform format of the image data is ensured, the text is normalized by adopting a natural language processing technology such as word segmentation, part-of-speech annotation, named entity identification and the like, meanwhile, the repeated medical examination record is deleted by comparing the text content, and the standard data for femoral head necrosis treatment is finally obtained through the processing steps.
And S13, performing association matching and integration file integration on the multi-source data of the femoral head necrosis treatment corresponding to the same patient in the multi-source standard data of the femoral head necrosis treatment to generate a femoral head necrosis treatment data file.
In the embodiment of the invention, the data are associated and matched by utilizing the unique identification (such as an identification card number, a medical record number and the like) of a patient in multi-source standard data for femoral head necrosis treatment, the basic information, the clinical symptoms of the femoral head, the femoral head medical images and the femoral head necrosis treatment scheme data of the same patient are integrated, firstly, the patient identification is used as an index, all relevant records of the patient are searched in different types of data tables, then, the records are organized according to a certain structure, for example, a comprehensive data structure comprising a basic information table, a clinical symptom table, an image record table and a treatment scheme table of the patient is created, the storage path of the image data is associated with the corresponding patient record, and finally, the integrated data are stored in a special database to form a complete femoral head necrosis treatment data file, so that the data can be conveniently analyzed and utilized later.
Further, the femoral head image feature analysis module comprises the following functions:
Performing image physical characteristic deep analysis on corresponding femoral head medical images in a femoral head necrosis treatment data file so as to research attenuation rules corresponding to X-rays when the X-rays pass through femoral head tissues for the X-ray images, analyzing the relation between gray values and tissue densities in the images according to the difference of the X-ray absorption degrees of different tissues, deeply analyzing the femoral head tissue characteristics reflected by voxel values of different layers for CT images, and researching hydrogen proton densities and relaxation times corresponding to magnetic resonance signals and the femoral head tissues for MRI images to obtain a femoral head medical image physical characteristic data set;
Performing femoral head necrosis tissue region segmentation on the femoral head medical image to generate a femoral head necrosis tissue region segmentation image;
Performing multi-scale texture feature analysis on the segmented image of the femoral head necrosis tissue region under different scales based on wavelet transformation to obtain a texture feature set of the femoral head necrosis tissue region;
And carrying out femoral head feature analysis on the femoral head necrosis tissue region texture feature set based on the femoral head medical image physical feature data set so as to invert the tissue structure corresponding to the femoral head necrosis tissue according to the corresponding gray scale and the signal intensity in the image, calculating the density distribution corresponding to the femoral head necrosis tissue by utilizing the known relation between the X-ray attenuation coefficient and the tissue density and combining the image voxel value, and simultaneously, inverting the morphological features corresponding to the femoral head necrosis tissue according to the relation between the T1 and T2 relaxation time and the tissue morphology so as to obtain the femoral head necrosis tissue image feature, wherein the femoral head necrosis tissue image feature comprises the femoral head necrosis tissue structure, the femoral head necrosis tissue morphology and the femoral head necrosis tissue density.
As an embodiment of the present invention, referring to fig. 3, a functional flow diagram of the femoral head image feature analysis module in fig. 1 is shown, where in this embodiment, the femoral head image feature analysis module includes the following functions:
S21, performing image physical characteristic depth analysis on corresponding femoral head medical images in a femoral head necrosis treatment data file to research a corresponding attenuation rule of X-rays when the X-rays pass through femoral head tissues for the X-ray images, analyzing the relation between gray values and tissue densities in the images according to the difference of the absorption degree of the X-rays by different tissues, performing depth analysis on femoral head tissue characteristics reflected by voxel values of different layers for CT images, and researching hydrogen proton densities and relaxation times corresponding to magnetic resonance signals and the femoral head tissues for MRI images to obtain a femoral head medical image physical characteristic data set;
In the embodiment of the invention, three femoral head medical images of X-rays, CT and MRI are extracted from a femoral head necrosis treatment data file, for the X-ray images, intensity changes before and after X-rays pass through femoral head tissues are measured by utilizing an X-ray physics principle, attenuation rules of the X-rays are analyzed, mathematical models of gray values and tissue densities in the images are established by experimental comparison of X-ray absorption conditions of tissues with different densities (such as bones, muscles and fats), voxel values of each layer are analyzed in detail for CT images, the voxel values are related to electron densities of tissues according to the CT imaging principle, femoral head tissue characteristics such as normal bone tissues, necrotic bone tissues and the like corresponding to different voxel value ranges are determined by referring to medical data and databases, for the MRI images, professional magnetic resonance imaging analysis software is adopted, magnetic resonance signal intensity is measured, and the relations between signals and the femoral head are studied by combining known hydrogen proton density and relaxation time (T1 and T2) theory, and all data obtained through analysis are arranged and summarized, and finally a medical image physical characteristic data set is formed.
S22, dividing the femoral head necrosis tissue region of the femoral head medical image to generate a femoral head necrosis tissue region division image;
In the embodiment of the invention, the femoral head medical image is processed by adopting a method based on the combination of threshold segmentation and morphological operation, for X-ray, CT and MRI images, the approximate gray value or the signal intensity range of the femoral head necrosis tissue in the image is firstly determined according to the image physical characteristic data set, and is used as a threshold value for preliminary segmentation, for example, in the CT image, the voxel value of the necrosis tissue is usually positioned in a specific section, the section is used as the threshold value for binarization processing of the image, so as to obtain a preliminary necrosis tissue area, then morphological operation such as expansion and corrosion is applied, small noise and holes in the segmentation result are removed, blank parts in the necrosis tissue area are filled, the segmentation result is more accurate and complete, and finally, the boundary correction is carried out on the segmented image, so that the clear and accurate boundary of the necrosis tissue area is ensured, and the femoral head necrosis tissue area segmentation image is generated.
S23, performing multi-scale texture feature analysis on the segmented image of the femoral head necrosis tissue region under different scales based on wavelet transformation to obtain a texture feature set of the femoral head necrosis tissue region;
In the embodiment of the invention, the image is decomposed into sub-bands with different scales by carrying out wavelet transformation on the image of the femoral head necrosis tissue region, the high-frequency sub-band coefficients after wavelet transformation are focused on under the small scale, the coefficients reflect detailed information in the image, the microstructure texture features corresponding to the bone trabeculae in the femoral head tissue are obtained by carrying out characteristic extraction on the high-frequency sub-band coefficients, such as calculation of texture energy, contrast and the like, for example, the arrangement mode and the density of the bone trabeculae can embody specific modes in the high-frequency sub-band coefficients, the fine texture features can be extracted by analyzing the modes, the whole femoral head necrosis tissue region can be analyzed by using a fractal dimension calculation method under the large scale, the roughness degree and the self-similarity of textures can be measured by calculating the fractal dimension, the quantitative features of the whole femoral head texture can be obtained, and the texture features extracted under the small scale and the large scale are summarized, so that the texture feature set of the femoral head necrosis tissue region can be finally obtained.
S24, performing femoral head feature analysis on the texture feature set of the femoral head necrosis tissue area based on the femoral head medical image physical feature data set to invert the tissue structure corresponding to the femoral head necrosis tissue according to the corresponding gray scale and the signal intensity in the image, calculating the density distribution corresponding to the femoral head necrosis tissue by combining the known relation between the X-ray attenuation coefficient and the tissue density and the image voxel value, and simultaneously inverting the morphological feature corresponding to the femoral head necrosis tissue according to the relation between the T1 and T2 relaxation time and the tissue morphology to obtain the femoral head necrosis tissue image feature, wherein the femoral head necrosis tissue feature comprises the femoral head necrosis tissue structure, the femoral head necrosis tissue morphology and the femoral head necrosis tissue density.
In the embodiment of the invention, the physical characteristic data set of the femoral head medical image and the texture characteristic set of the femoral head necrosis tissue area are combined for analysis, the gray scale and the signal intensity in the image are calculated, the tissue density corresponding to each voxel is calculated according to the attenuation coefficient formula and the voxel value, the density distribution diagram is drawn, the T1 and T2 relaxation time in the MRI image are inverted according to the corresponding relation between the T1 and T2 relaxation time and the tissue form in the medical research, the morphological characteristics of the necrosis tissue such as the size and the shape of the necrosis area can be obtained, the known relation between the X-ray attenuation coefficient and the tissue density is utilized, the voxel value in the CT image is combined for calculation, and the density distribution diagram of the femoral head necrosis tissue is obtained, for example, the tissue density corresponding to each voxel is calculated according to the attenuation coefficient formula and the voxel value, and the density distribution diagram is drawn, and the morphological characteristics of the necrosis tissue are inverted according to the corresponding relation between the T1 and T2 relaxation time and the tissue form in the medical research.
Further, the multi-scale texture feature analysis is performed on the segmented image of the femoral head necrosis tissue region under different scales based on wavelet transformation, specifically, fine texture features corresponding to the femoral head tissue are extracted through high-frequency subband coefficients corresponding to wavelet transformation under small scales, the fine texture features comprise microstructure textures corresponding to bone trabeculae, and the texture roughness and self-similarity corresponding to the whole femoral head are calculated and analyzed by fractal dimension under large scales.
Further, the femoral head treatment characteristic analysis module comprises the following functions:
Performing femoral head treatment staging on the corresponding femoral head necrosis treatment plan based on the corresponding femoral head pain degree and the limited range of femoral head movement in the femoral head necrosis treatment data file to obtain corresponding femoral head necrosis treatment sub-plans under each femoral head symptom staging;
In the embodiment of the invention, the pain degree score and the limited range of femoral head movement data of each patient are extracted from the femoral head necrosis treatment data file, the pain degree score is divided by 1-10 degrees, the limited range of movement is obtained by measuring the anteflexion, abduction and internal rotation movement angles of the hip joint according to subjective evaluation of the patient, and is calculated by comparing the limited range of movement with the normal range of movement, for example, the limited range of movement is reduced to 10 degrees at the early stage when the pain degree of the femoral head is between 1-3 minutes and the limited range of movement is between 10% -20%, the limited range of movement comprises the reduction of the anterior flexion angle of the hip joint from the normal range of 120 degrees to 140 degrees to 100 degrees, the limited range of movement is reduced to 25 degrees to 35 degrees from the normal range of 30 degrees to 40 degrees, the limited range of internal rotation angle is reduced to 20 degrees from the normal range of 30 degrees to 30 degrees, the limited range of movement is enlarged to 20 degrees from the limited range of the femoral head of the middle stage when the pain degree of the femoral head is between 4-6 minutes and the limited range of movement is enlarged to 20% -40%, the limited range of movement comprises the anteflexion angle is reduced to 80 degrees to 20 degrees, the limited range of the internal rotation angle is reduced to 10 degrees, the limited to the final stage when the limited range of movement is reduced to the femoral head is reduced to the limited to the angle of the corresponding to the small range of the anteversion of the angle is 10 degrees.
Preferably, based on the corresponding femoral head past medical history and medicine use history in the basic information of the patient, necrosis treatment and allergy probability analysis are carried out on the corresponding femoral head necrosis treatment sub-cases under each femoral head symptom stage so as to obtain the femoral head necrosis treatment probability and the femoral head necrosis treatment allergy probability;
In the embodiment of the invention, the past medical history of the femoral head is collected from basic information of a patient, including the conditions of fracture, dislocation and the like, and the medical use history, such as whether a corticosteroid and other medicines possibly causing femoral head necrosis are used, for each treatment sub-table under the symptom stage of the femoral head, a big data analysis method and a probability statistical model are used for analysis, for example, a database containing a large amount of patient data is established, the treatment effect and the allergy condition of the patient with the past specific medical history or the specific medicine when receiving different treatment sub-tables are analyzed, for the early treatment sub-table, the ratio of the number of patients with the past fracture medical history and the corticosteroid medicine when receiving the treatment sub-table is counted to obtain the femoral head necrosis treatment probability, the ratio of the number of people who have anaphylactic reaction to the total number of people is counted to obtain the femoral head necrosis treatment allergy probability, and the necrosis treatment probability corresponding to the treatment sub-tables under the middle stage and the late stage of the symptom stage are also analyzed, and the necrosis treatment probability and the necrosis allergy probability corresponding to the femoral head necrosis treatment sub-table under each symptom stage are finally obtained.
Preferably, based on the femoral head necrosis tissue density, performing necrosis tissue attenuation evaluation analysis on the femoral head necrosis treatment sub-cases corresponding to each femoral head symptom stage to obtain the femoral head necrosis tissue treatment attenuation efficiency;
In an embodiment of the present invention, initial density data of necrotic tissue of a femoral head of a patient is obtained by using a medical imaging technique (e.g., CT scan). For each treatment sub-regimen at each femoral head symptom stage, the initial density of necrotic tissue was recorded prior to initiation of treatmentIn the treatment process, CT scanning is carried out according to a certain time interval (such as once a month), real-time density data of necrotic tissue is obtained, and after treatment is finished, the final necrotic tissue density is recordedBy calculating density differencesThe actual attenuation of necrotic tissue is obtained. Meanwhile, through analysis of a large amount of clinical data and medical research, the theoretical maximum attenuation of the femoral head necrosis tissue under different symptom stages is determinedUsing the formulaAnd calculating the treatment attenuation efficiency of the femoral head necrosis tissue corresponding to each treatment sub-pattern. For example, under the early treatment sub-plan, the initial density is 1.2g/cm3, the final density is 1.0g/cm3, the theoretical maximum attenuation is 0.3g/cm3, the treatment attenuation efficiency is 1.2-1.0/0.3=66.7, and finally the treatment attenuation efficiency of the femoral head necrosis tissue is obtained.
Preferably, the femoral head necrosis treatment probability, the femoral head necrosis treatment allergy probability and the femoral head necrosis tissue treatment attenuation efficiency are taken as basic characteristics to obtain the femoral head necrosis treatment basic characteristics.
In the embodiment of the invention, after the previous calculation is completed, the obtained femoral head necrosis treatment probability, femoral head necrosis treatment allergy probability and femoral head necrosis tissue treatment attenuation efficiency corresponding to each femoral head symptom stage treatment sub-plan are arranged, and the three characteristic values of each treatment sub-plan are recorded in a table form to form a data set. For example, the early treatment case 1 has a treatment probability of 80%, an allergy probability of 5% and a treatment attenuation efficiency of 70%, the middle treatment case 2 has a treatment probability of 60%, an allergy probability of 10% and a treatment attenuation efficiency of 50%, and the like, and these data are combined together to form basic characteristics of femoral head necrosis treatment, which can be used for subsequent treatment effect prediction and analysis, and finally the basic characteristics of femoral head necrosis treatment are obtained.
Further, the symptom stage of the femoral head is specifically an early stage with the pain degree of the femoral head between 1 and 3 minutes and the limited range of the movement of the femoral head between 10 and 20 percent, comprising the reduction of the forward flexion movement angle of the hip joint from a normal range of 120 to 140 degrees to a range of 100 to 120 degrees, the reduction of the abduction movement angle of the hip joint from a normal range of 30 to 45 degrees to 25 to 35 degrees, and the reduction of the internal rotation movement angle from a normal range of 30 to 40 degrees to 20 to 30 degrees, an intermediate stage with the pain degree of the femoral head between 4 and 6 minutes and the expansion of the limited range of the femoral head between 20 and 40 percent, comprising the reduction of the forward flexion movement angle of the hip joint to 80 to 100 degrees, the reduction of the abduction movement angle of the hip joint to 20 to 25 degrees, and the reduction of the internal rotation movement angle to 10 to 20 degrees, and an late stage with the pain degree of the femoral head between 7 and 10 minutes and the expansion of the limited range of the movement of the femoral head exceeding 40 percent, comprising the forward flexion movement angle of the hip joint less than 80 degrees, the abduction movement angle of the femoral head less than 20 degrees, and the internal rotation angle of the internal rotation angle less than 10 degrees.
Further, the necrosis tissue attenuation evaluation analysis of the femoral head necrosis treatment sub-cases corresponding to each femoral head symptom stage based on the femoral head necrosis tissue density comprises:
performing tissue density treatment simulation on the corresponding femoral head necrosis tissue density based on the corresponding femoral head necrosis treatment sub-plan under each femoral head symptom stage so as to generate a corresponding density attenuation process of the femoral head necrosis tissue under each treatment effect;
in the embodiment of the invention, the corresponding femoral head necrosis treatment sub-cases are respectively determined according to different femoral head symptom stages, such as early stage, medium stage and late stage, and the treatment sub-cases comprise specific treatment modes, treatment periods, medicament doses and other information. The method comprises the steps of obtaining initial density data of femoral head necrosis tissue under each stage by utilizing a medical imaging technology (such as CT scanning), constructing a three-dimensional femoral head necrosis tissue model, simulating the action mechanism of a medicine on the necrosis tissue, the influence of physical treatment on the tissue and the like according to treatment parameters in a treatment sub-table in a computer simulation environment, for example, for medicine treatment, simulating the interaction of medicine molecules and necrosis tissue cells according to the components and action principles of the medicine to cause apoptosis and decomposition of tissue cells, thereby causing the change of tissue density, dividing the whole treatment process into a plurality of time steps according to a treatment period, updating the tissue density data in each time step, and finally generating a density attenuation process corresponding to the femoral head necrosis tissue under each treatment effect.
Preferably, the actual attenuation amount statistics is carried out on the density attenuation process corresponding to the femoral head necrosis tissue under each treatment effect, so as to obtain the actual attenuation amount of the tissue density corresponding to the femoral head necrosis tissue under each treatment effect;
In the embodiment of the invention, the tissue density data of each time step is analyzed after the previous density attenuation process simulation is completed, and the tissue density value at the initial time is recorded for the density attenuation process corresponding to each treatment sub-caseAnd tissue density value at the end of treatmentBy calculating the difference between the twoThe method comprises obtaining the integral density attenuation of femoral head necrosis tissue under the treatment effect, recording tissue density values at multiple key time points during treatment, calculating density differences between adjacent time points, and accumulating the differences, for example, selecting three key time points in treatment periodThe corresponding tissue density values are respectively 、、Actual attenuation amountSuch statistics is carried out on the density attenuation process under each treatment effect, and the actual attenuation of the tissue density corresponding to the femoral head necrosis tissue under each treatment effect is finally obtained.
Preferably, the theoretical maximum attenuation of the femoral head necrosis tissue is obtained, and the actual attenuation of the tissue density corresponding to the femoral head necrosis tissue under each treatment effect is subjected to necrosis tissue attenuation quantitative calculation based on the theoretical maximum attenuation of the femoral head necrosis tissue, so that the femoral head necrosis tissue treatment attenuation efficiency is obtained.
In the embodiment of the invention, a great amount of medical research and clinical data are analyzed to determine the theoretical maximum attenuation of the femoral head necrosis tissue, wherein the theoretical maximum attenuation refers to the maximum density attenuation value which can be achieved by the femoral head necrosis tissue under ideal treatment conditions, and is generally related to the initial range, the property and other factors of the necrosis tissue, and the theoretical maximum attenuation of the femoral head necrosis tissue under a certain symptom stage is determined by researchActual attenuation of tissue density for each previously obtained treatment effect(Representing different treatment sub-protocols), using the formulaAnd (3) carrying out quantitative calculation on the attenuation of the necrotic tissue, for example, if the actual attenuation of the tissue density under a certain treatment sub-plan is 0.5g/cm < 3 >, and the theoretical maximum attenuation under the symptom stage is 1g/cm < 3 >, the treatment attenuation efficiency of the necrotic tissue of the femoral head corresponding to the treatment sub-plan is 50%, and carrying out such calculation on the actual attenuation under all treatment effects, so as to finally obtain the treatment attenuation efficiency of the necrotic tissue of the femoral head under each treatment effect.
Further, the treatment effect prediction analysis module comprises the following functions:
Performing treatment effect association mining analysis on the femoral head necrosis tissue image characteristics and the femoral head necrosis characteristic factors in the femoral head necrosis treatment basic characteristics to obtain association relations between the femoral head necrosis characteristic factors and the treatment effects;
In the embodiment of the invention, a large amount of data of femoral head necrosis patients are collected, wherein femoral head necrosis tissue image characteristics are acquired through medical image equipment (such as magnetic resonance imaging, MRI, X-rays and the like) and comprise information of bone trabecular structure, necrosis area size, position and the like, the femoral head necrosis treatment basic characteristics cover factors such as femoral head necrosis treatment probability, femoral head necrosis treatment allergy probability, femoral head necrosis tissue treatment attenuation efficiency and the like, a correlation rule mining algorithm such as an Apriori algorithm is used for processing the data of the characteristic factors and treatment effects (such as pain relieving degree, joint function recovery condition and the like), the Apriori algorithm is taken as an example, a minimum support degree and a minimum confidence threshold are set, a frequent item set is found through multiple times of data scanning, and then the correlation relation between each femoral head necrosis characteristic factor and the treatment effect is obtained, for example, the fact that when the age of a patient is less than 50 years and the treatment effect is remarkably recovered by adopting surgery is found out to be a correlation relation, and finally the correlation relation between each femoral head necrosis characteristic factor and the treatment effect is obtained.
Preferably, based on the association relation between the femoral head necrosis characteristic factors and the treatment effect, carrying out correlation evaluation on the femoral head necrosis tissue image characteristics and the femoral head necrosis characteristic factors in the femoral head necrosis treatment basic characteristics to obtain characteristic correlation factors between the femoral head necrosis characteristic factors and the treatment effect;
in the embodiment of the invention, the correlation between the necrosis characteristic factors and the treatment effect of each femoral head is evaluated by adopting a pearson correlation coefficient method according to the correlation obtained previously, the pearson correlation coefficient between each femoral head necrosis characteristic factor (such as the size of the necrosis area) and the treatment effect index (such as the pain relieving degree) is calculated, the pearson correlation coefficient is calculated based on the mean value, standard deviation and covariance of the pearson correlation coefficient and the treatment effect index, if the pearson correlation coefficient between the necrosis area size and the pain relieving degree is calculated, the pearson correlation coefficient is-0.6, the two are indicated to be in negative correlation, namely, the greater the necrosis area is, the lower the pain relieving degree is, the characteristic correlation factors between the necrosis characteristic factors and the treatment effect of each femoral head necrosis characteristic factor are obtained by carrying out such calculation on the characteristic factors in all femoral head necrosis tissue image characteristics and the treatment basic characteristics, and the characteristic correlation factors between the characteristic factors and the treatment effect are quantized, and the characteristic correlation factors between the characteristic factors and the treatment effect are finally obtained.
Preferably, the causal deducing of the treatment effect is carried out on the image characteristics of the femoral head necrosis tissue and the various femoral head necrosis characteristic factors in the basic characteristics of the femoral head necrosis treatment, so as to obtain the potential causal relation coefficient between the various femoral head necrosis characteristic factors and the treatment effect;
In the embodiment of the invention, causal relation between the image characteristics of the femoral head necrosis tissue and each characteristic factor and the treatment effect in the treatment basic characteristics is inferred by applying a causal inference algorithm, such as a glauch causal test, and taking the glauch causal test as an example, time series data of each characteristic factor (such as a treatment period) and the treatment effect (such as joint function recovery condition) are input into a test model, and if the change of the treatment period is found to be statistically prior to the change of the joint function recovery condition and has significance through test analysis, the treatment period is considered to have potential causal relation to the joint function recovery condition, the causal inference is performed on all the characteristic factors, and the specific number of the potential causal relation between each characteristic factor and the treatment effect is counted, so that the potential causal relation coefficient between each femoral head necrosis characteristic factor and the treatment effect is finally obtained.
Preferably, the characteristic of influence of the treatment effect on the image characteristic of the femoral head necrosis tissue and the basic characteristic of the treatment of the femoral head necrosis is screened based on the characteristic correlation factor and the potential causal relation coefficient between the characteristic factors of the femoral head necrosis and the treatment effect, so as to obtain the characteristic of remarkable influence of the treatment effect of the femoral head necrosis;
In the embodiment of the invention, the characteristic correlation factors and the potential causal relation coefficients between the characteristic factors and the treatment effects of the femoral head necrosis are comprehensively considered by setting screening criteria, the larger the absolute value of the characteristic correlation factors is, the stronger the correlation is, the larger the potential causal relation coefficients are, the larger the quantity of the characteristic factors are, for example, the characteristic factors with the absolute value of the characteristic correlation factors being larger than 0.5 and the potential causal relation coefficients being larger than 2 are set as obvious influence characteristics, according to the criteria, the image characteristics of the femoral head necrosis tissues and all the characteristic factors in the treatment basic characteristics are screened, if the characteristic correlation factors of the size of the necrosis areas are-0.7, and the potential causal relation coefficients are 3, the size of the necrosis areas is determined as obvious characteristics of the influence of the treatment effects of the femoral head necrosis, and through the screening, a group of characteristics with obvious influence on the treatment effects of the femoral head necrosis treatment effects are finally obtained.
Preferably, a femoral head necrosis treatment effect prediction model is constructed by utilizing a convolutional neural network, and significant characteristics of the femoral head necrosis treatment effect are input into the femoral head necrosis treatment effect prediction model for treatment effect prediction analysis so as to predict and output the corresponding femoral head necrosis treatment effects under different treatment schemes.
In the embodiment of the invention, a prediction model of femoral head necrosis treatment effect is constructed by using a Python programming language and a deep learning framework (such as TensorFlow or PyTorch), a convolution neural network of a pyramid network architecture is constructed, convolution operation is carried out on input image data and treatment characteristic data by using a convolution kernel of 3x3 in a shallow layer of the network, detail characteristics such as trabecular change of a micro bone and early necrosis area boundary are extracted, then the characteristics are reduced by using a pooling window of 1x1, in the deep layer network, integral characteristics such as joint morphology, macroscopic distribution and the like are extracted by using a convolution kernel of 5x5, the dimension is reduced by using a pooling window of 1x1, and energy constraint is introduced during training of the model, the activation energy and the information transmission energy of each neuron are calculated and are incorporated into an optimization objective function, a prediction error is minimized through an optimizer (such as random gradient descent), meanwhile, a cross-validation method is adopted, a data set is divided into a plurality of subsets, one subset is taken as a validation set in turn, the other subsets are taken as training sets, a model is trained and validated for many times, model parameters with the best performance are selected, the screened femoral head necrosis treatment effect influence significant characteristic data are input into the trained model, the model processes and predicts the input data in different network layers, and finally, prediction results of corresponding femoral head necrosis treatment effects under different treatment schemes such as pain relieving degree, joint function recovery condition and the like are output.
Further, the femoral head necrosis treatment effect prediction model is specifically a pyramid network architecture for extracting treatment effect characteristics on multiple scales, the architecture extracts detail characteristics corresponding to the femoral head by using convolution kernels corresponding to 3x3 and pooling windows of 1x1 in a shallow network, including micro trabecular changes and early necrosis area boundaries, extracts integral characteristics corresponding to the femoral head by using convolution kernels corresponding to 5x5 and pooling windows of 1x1 in a deep network, including joint morphology and necrosis area macroscopic distribution, and takes activation energy and information transfer energy of each neuron in the network into an optimization target through introducing energy constraint to pay attention to prediction errors corresponding to the model, and simultaneously improves prediction performance corresponding to the model by using cross verification so as to predict and output corresponding femoral head necrosis treatment effects at different network layers.
Furthermore, the invention also provides a femoral head necrosis treatment effect prediction and analysis method based on big data, which is implemented based on the femoral head necrosis treatment effect prediction and analysis platform based on big data, and comprises the following steps:
Acquiring multi-source medical data for treating femoral head necrosis, wherein the multi-source medical data comprises basic information of a patient, clinical symptoms of the femoral head, medical images of the femoral head and a femoral head necrosis treatment scheme, and performing data standard cleaning and medical file integration on the multi-source medical data for treating femoral head necrosis to generate a femoral head necrosis treatment data file;
Performing femoral head feature analysis on corresponding femoral head medical images in the femoral head necrosis treatment data file to obtain femoral head necrosis tissue image features including a femoral head necrosis tissue structure, a femoral head necrosis tissue form and a femoral head necrosis tissue density;
Performing femoral head necrosis treatment stage on the corresponding femoral head necrosis treatment scheme based on the corresponding femoral head clinical symptoms in the femoral head necrosis treatment data file to obtain a femoral head necrosis treatment sub-scheme corresponding to each femoral head symptom stage;
The method comprises the steps of carrying out treatment effect influence characteristic screening on femoral head necrosis tissue image characteristics and basic femoral head necrosis treatment characteristics to obtain significant femoral head necrosis treatment effect influence characteristics, constructing a femoral head necrosis treatment effect prediction model, inputting the significant femoral head necrosis treatment effect influence characteristics into the femoral head necrosis treatment effect prediction model for treatment effect prediction analysis so as to predict and output the corresponding femoral head necrosis treatment effects under different treatment schemes.
The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.
The foregoing is only a specific embodiment of the invention to enable those skilled in the art to understand or practice the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.