CN116805536A

CN116805536A - Data processing method and system based on tumor case follow-up

Info

Publication number: CN116805536A
Application number: CN202311058447.XA
Authority: CN
Inventors: 郑良杰
Original assignee: Leling People's Hospital
Current assignee: Leling People's Hospital
Priority date: 2023-08-22
Filing date: 2023-08-22
Publication date: 2023-09-26

Abstract

The invention relates to the technical field of medical information processing, in particular to a data processing method and system based on tumor case follow-up, wherein the method comprises the following steps: acquiring first tumor case follow-up data from a hospital case system; medical detection is carried out on a patient by using tumor detection medical equipment, and latest tumor case follow-up data are obtained; performing visual projection on the follow-up data of the second tumor case by using a matrix decomposition method to generate a tumor information feature matrix projection diagram; carrying out trend prediction analysis on the tumor feature matrix projection graph by using a random forest algorithm; performing expansion convolution and multi-scale sampling on the tumor lesion trend prediction graph by using a cyclic convolution algorithm; binding a tumor characteristic trend prediction model in a time sequence; carrying out data fusion on the follow-up data of the second tumor case and the tumor characteristic trend prediction model; the tumor follow-up encryption data are uploaded to a hospital case system, and the method realizes ordered and efficient processing of the tumor case follow-up data.

Description

Data processing method and system based on tumor case follow-up

Technical Field

The invention relates to the technical field of medical information processing, in particular to a data processing method and system for follow-up tumor case.

Background

The modern medical technology greatly improves the survival rate of tumor patients, so that more and more cases need continuous follow-up and monitoring, in the traditional case follow-up data processing method, the scale of the follow-up data is huge, the processing efficiency of the manually recorded case follow-up data is low, and the data is chaotic, therefore, a data processing method and a system for tumor case follow-up are needed to manage and process the data, a data processing method and a system for tumor case follow-up use modern artificial intelligence technology, data mining technology, big data analysis technology and the like to analyze and process the tumor case follow-up data, the modern data mining technology is used to analyze the tumor case follow-up data to obtain key information and observation conclusion, support is provided for researching tumor conditions, and the system needs to integrate information of doctors, patients and related parties of cases, and builds a unified data management system, so that the data can be shared, transferred and exchanged.

Disclosure of Invention

The invention provides a data processing method and system for follow-up tumor case access to solve at least one technical problem.

In order to achieve the above purpose, the present invention provides a data processing method for tumor case follow-up, comprising the following steps:

step S1: acquiring first tumor case follow-up data from a hospital case system; medical detection is carried out on a patient by using tumor detection medical equipment, and latest tumor case follow-up data are obtained; updating the first tumor case follow-up data in real time based on the latest tumor case follow-up data to generate second tumor case follow-up data;

step S2: performing visual projection on the follow-up data of the second tumor case by using a matrix decomposition method to generate a tumor information feature matrix projection diagram;

step S3: carrying out trend prediction analysis on the tumor feature matrix projection graph by using a random forest algorithm to generate a tumor lesion trend prediction graph;

step S4: performing expansion convolution and multi-scale sampling on the tumor lesion trend prediction graph by using a cyclic convolution algorithm, and constructing a tumor feature trend prediction model;

step S5: binding a tumor characteristic trend prediction model in a time sequence; carrying out data fusion on the second tumor case follow-up data and the tumor characteristic trend prediction model by using a neural network algorithm to generate third tumor case follow-up data;

Step S6: carrying out affine data encryption on the follow-up data of the third tumor case by utilizing an affine encryption algorithm to generate affine encryption data of the tumor case; and uploading the tumor follow-up encrypted data to a hospital case system by using a distributed network slicing algorithm so as to realize real-time management of the follow-up data.

The invention provides a data processing method based on tumor case follow-up, the timeliness and accuracy of data can be ensured by updating first tumor case follow-up data in real time, more complete data support is provided for follow-up and treatment of patients, latest tumor case follow-up data can be acquired by using tumor detection medical equipment, more comprehensive treatment advice is provided for doctors, more visual data visualization effect can be obtained by using matrix decomposition method to carry out visual projection on tumor case follow-up data, better data foundation can be provided for follow-up data analysis and predictive modeling by using visual projection, trend prediction analysis can be carried out on tumor feature matrix projection graph by using random forest algorithm, better understanding of disease trend by using doctor, more specific advice is provided for treatment, the analysis result can be used as an important reference of tumor case data, expansion convolution and multi-scale sampling can be carried out on tumor lesion prediction graph by using cyclic convolution algorithm, tumor feature trend prediction model can be more accurately predicted, more visual data can be obtained by using more visual projection method, more visual data understanding can be better understanding of doctors, more visual data can be provided for patients, more privacy-relevant data can be encrypted by using the following-up data of the following-up data, the clinical data can be provided for the clinical data, the clinical data can be more complete data can be encrypted by using the clinical data, the clinical data can be more complete data, the clinical data can be better provided for the clinical data has a specific effect can be better estimated by using the clinical data, the clinical data has been provided for treatment, the distributed network slicing algorithm is used for uploading the encrypted data to the hospital case system, so that real-time management and sharing of follow-up data can be realized, and a more convenient and efficient data sharing mode is provided on the premise of ensuring data safety.

Preferably, step S1 comprises the steps of:

step S11: acquiring first tumor case follow-up data from a hospital case system, wherein the first tumor case follow-up data comprises patient information data, patient follow-up data, patient medical record data and tumor chemotherapy scheme data;

step S12: medical detection is carried out on a patient by using tumor detection medical equipment, and latest tumor case follow-up data are obtained; the tumor detection medical equipment comprises a nuclear magnetic resonance imager, an X-ray instrument, a blood analyzer, a biochemical analyzer and a nucleic acid analyzer;

step S13: extracting characteristic information of the latest tumor case follow-up data to generate tumor case marking data;

step S14: and updating the first tumor case follow-up data in real time based on the tumor case marking data, and generating second tumor case follow-up data.

The invention obtains the first tumor case follow-up data from the medical hospital case system and obtains the latest tumor case follow-up data by using the tumor detection medical equipment to carry out medical detection on the patient, can update the tumor case data of the patient in real time, extracts the characteristic information of the data to generate the tumor case marking data, can promote doctors to know the tumor condition of the patient more timely, and better makes a treatment scheme for the patient, the first tumor case follow-up data comprises the patient information data, the patient follow-up data, the patient medical history data and the tumor chemotherapy scheme data, the data is very critical for the doctors to carry out the disease analysis of the patient, but the latest tumor case follow-up data can not reflect the latest disease change of the patient due to the slow updating speed of the data, can obtain the latest tumor case follow-up data of the patient by using the tumor detection medical equipment, comprises a nuclear magnetic resonance imaging instrument, an X-ray instrument, a blood analyzer, a biochemical analyzer, a nucleic acid analyzer and the like, can provide more accurate disease data of the patient, greatly helps the doctors to analyze and diagnose the tumor, extracts the characteristic information of the latest tumor case follow-up data, can generate the tumor case data, can better monitor the patient condition quality by optimizing the patient case data, can better test and make the real-time tumor case data, can better read the patient condition quality, and make the patient data by making the real-time tumor case follow-up data, and better patient data can be better read by the data, and the data, and improves the therapeutic effect of the patient.

Preferably, step S2 comprises the steps of:

step S21: performing data preprocessing on the second tumor case follow-up data to generate a tumor case follow-up data preprocessing pipeline, wherein the data preprocessing comprises cleaning, integration and standardization;

step S22: dividing a matrix of a tumor case follow-up data preprocessing pipeline to obtain a plurality of tumor case follow-up data submatrices;

step S23: performing matrix decomposition on the tumor case follow-up data submatrices by using a non-negative matrix decomposition method to generate a main component matrix and a minimum error matrix;

step S24: performing feature extraction according to the principal component matrix and the minimum error matrix to generate a tumor follow-up feature weight matrix;

step S25: and performing visual projection on the tumor follow-up feature weight matrix by using an application visual projection method to generate a tumor feature matrix projection map.

The invention carries out data preprocessing through the second tumor case follow-up data, including cleaning, integrating and standardization processing, generates a tumor case follow-up data preprocessing pipeline, the process can reduce noise and redundant parts of the data, improve the quality and reliability of the data, carries out matrix division on the tumor case follow-up data preprocessing pipeline to obtain a plurality of tumor case follow-up data submatrices, carries out matrix division on the data, can decompose large-scale data into small-scale matrixes and carries out processing on data dimension, thereby improving the calculation efficiency and precision, carries out matrix decomposition on the tumor case follow-up data submatrices by utilizing a non-negative matrix decomposition method, generates a principal component matrix and a minimum error matrix, can extract the hidden characteristics of the matrix by carrying out matrix decomposition on the data, and divides the matrix into the principal component matrix and the minimum error matrix, the method has the advantages that the main component matrix can better represent similarity and difference between data, the feature extraction is carried out according to the main component matrix and the minimum error matrix, the tumor follow-up feature weight matrix is generated, the important features of tumor cases can be extracted through the feature extraction of the main component matrix, the tumor follow-up feature weight matrix is generated, the feature weights can be used for further analysis and comparison of the tumor cases, the tumor follow-up feature weight matrix is subjected to visual projection by using a visual projection method, a tumor follow-up feature matrix projection graph is generated, the high-dimensional data can be visualized into two-dimensional or three-dimensional data through the visual projection of the tumor follow-up feature weight matrix, the difference and the similarity of the data are maintained in the visualization process, and the efficiency of data interpretation and analysis is improved.

Preferably, step S3 comprises the steps of:

step S31: cutting tumor lesion parts by using an image geometry method to the tumor feature matrix projection graph to generate a tumor lesion feature graph and a tumor non-lesion feature graph;

step S32: carrying out image gray level homogenization distribution treatment on the tumor lesion feature map by using a histogram equalization algorithm to generate a high-definition tumor lesion feature map;

step S33: marking tumor lesion feature points on the high-definition tumor lesion feature map by using a feature point detection algorithm to generate coordinates of the tumor lesion feature points;

step S34: gridding the tumor lesion feature point coordinates by using a random forest algorithm to generate a tumor lesion feature point grid;

step S35: trend calculation is carried out on the tumor lesion feature point grids by using a tumor lesion trend prediction formula, and a tumor lesion trend prediction graph is generated;

according to the invention, a tumor feature matrix projection image is cut through an image geometric structure method, a tumor feature part is separated from a non-tumor part, so that tumor feature is better extracted, a histogram equalization algorithm is utilized to perform image gray level homogenization distribution treatment on the tumor feature image, the contrast and definition of the tumor feature image are improved, subsequent operation and analysis are facilitated, a feature point detection algorithm is utilized to perform tumor feature point marking on a high-definition tumor feature image, specifically, local maximum points with feature properties are searched in the image, the feature points can reflect tumor feature positions and features, a random forest algorithm is utilized to perform gridding treatment on tumor feature point coordinates, the feature points are distributed on the image, rectangular grids with a certain rule are divided, data analysis and prediction on the tumor feature part are better realized, finally, a tumor feature point grid is subjected to trend calculation through a tumor feature point trend prediction formula, and clinical treatment and intervention schemes are further guided according to the trend and distribution rule of the feature point grids.

Preferably, the tumor lesion trend prediction formula in step S35 is specifically:

；

wherein ,diffusion range size value predicted for tumor lesion trend,/-for>Weight of the early tumor lesion in the normal part>For early growth rate of tumor->For the currently measured tumor diffusion coordinate range, +.>Coordinate range for early tumor spread +_>Weight of the current tumor lesion in normal part>For the current growth rate of the tumor, < > is->Tumor diffusion coordinate range for last time point detection, +.>Is the degree of periodic concussion of tumor development, +.>Is the oscillation period of the tumor development period, +.>For the extent of the influence of time on the tumor growth rate, < + >>Weight of trend of slowing down tumor growth rate with time, ++>Is the upper limit of the tumor size.

The invention is realized byComparing the early tumor diffusion range with healthy parts, regulating the influence degree of early tumor diffusion by weight, mapping the input value to an output value ranging from 0 to 1, and having the characteristics of smoothness, continuity, nonlinearity and the like, by the piecewise fitting, the size of the diffusion range of the tumor can be more accurately predicted, the influence of limiting the early tumor diffusion by taking A as a limit can be calculated, and the influence of limiting the early tumor diffusion by taking A as a limit is calculated >The method comprises the steps of carrying out secondary evaluation and judgment on the current illness state of a patient, evaluating the illness state of the patient, namely the invasion degree of the illness part of the patient, effectively distinguishing early-stage illness from late-stage illness through calculation of the formula, further determining the size and range of illness, being beneficial to diagnosis of tumors, reducing the progress and the diffusion of illness through early-stage diagnosis, improving the treatment effect, avoiding excessive injury in the treatment process, and obtaining the medicine for treating the illness state of the patient by using the formulaThe influence degree of the predicted result is limited according to the tumor size, so that the predicted result is more reliable, the diffusion range of the tumor lesion is gradually increased along with the increase of the lesion part, thereby leading to the increase of the error of the predicted result, and if the tumor size factor is not considered, the predicted result has certain uncertainty, thusInconvenience and excessive financial burden can be caused to patients, meanwhile, reliability and repeatability of a model can be influenced for doctors and researchers, so that the problems can be effectively avoided by limiting the size of the tumor, the formula limits the influence of the tumor in a predicted diffusion range, and the formula suppresses the size and predicted value of the tumor, so that a predicted result is more reliable and accurate.

Preferably, step S33 includes the steps of:

step S331: performing feature point detection on the high-definition tumor lesion feature map by using a feature point detection algorithm to generate initial tumor lesion feature point data;

step S332: filtering and screening the feature point data of the initial tumor lesion feature map by using a feature value threshold limiting method to generate reference tumor lesion feature map feature point data;

step S333: and marking the feature points of the reference tumor lesion feature map by using a space three-dimensional coordinate method to generate coordinates of the feature points of the tumor lesion.

According to the invention, feature point detection is carried out on the high-definition tumor lesion feature map through a feature point detection algorithm, namely, local maximum points with feature properties are searched in the image, the feature points can reflect the features such as brightness and color in the image, subsequent processing and analysis are facilitated, feature point data of the initial tumor lesion feature map are filtered and screened through a feature value threshold limiting method, invalid data such as non-representative or noise points are removed, more representative feature points are extracted, the step is very important for improving the feature point accuracy of tumor lesion parts, a space three-dimensional coordinate method is used for marking feature points of the reference tumor lesion feature map feature point data, the screened feature points are recorded and tumor lesion feature point coordinates are generated, and the coordinates can reflect the specific positions and forms of tumor lesions and are key bases for the trend prediction of the subsequent tumor lesions.

Preferably, step S4 comprises the steps of:

step S41: performing convolution pretreatment on the tumor lesion trend prediction graph by using a super-pixel convolution network to generate a tumor lesion prediction characteristic sample set;

step S42: performing convolution data cutting on the tumor lesion feature sample set by using a cyclic convolution algorithm to generate a tumor lesion prediction convolution data block;

step S43: performing edge feature reinforcement processing on the tumor lesion convolution data block by using an expansion convolution algorithm to generate a tumor lesion prediction convolution feature network;

step S44: carrying out space pyramid pooling multi-layer sampling on the tumor lesion convolution feature network by utilizing a multi-scale sampling algorithm to generate a tumor lesion prediction convolution feature sequence;

step S45: and carrying out data mining modeling on the tumor lesion prediction convolution feature sequence by using a correlation rule algorithm, and constructing a tumor feature trend prediction model.

According to the invention, the characteristic feature samples of the tumor lesion are extracted by carrying out convolution pretreatment on the tumor lesion trend prediction graph through the super-pixel convolution network so as to facilitate subsequent processing and analysis, the tumor lesion feature sample set is subjected to convolution data cutting by utilizing a cyclic convolution algorithm to generate tumor lesion prediction convolution data blocks, the convolution data blocks contain feature information of the tumor lesion, the extraction of processing detail features of the part is facilitated, the edge feature reinforcement processing is carried out on the tumor lesion convolution data blocks by utilizing an expansion convolution algorithm, the feature information of the tumor contour can be further highlighted, the feature information of the tumor contour is very critical for accurately predicting the position, the form, the trend and the like of the tumor lesion, the spatial pyramid pooling multi-layer sampling is carried out on the tumor lesion convolution feature network by utilizing a multi-scale sampling algorithm, the information with different scales and different features is extracted, the tumor lesion prediction convolution feature sequence is generated, the accuracy of the tumor lesion feature is improved, the data mining modeling is carried out on the tumor lesion prediction convolution feature sequence by utilizing a correlation rule algorithm, the tumor feature trend prediction model is established, the accurate prediction and the tumor feature trend is predicted, the tumor is predicted, the position, the tumor is predicted, the important and the tumor lesion position and the important trend is found and has a great significance.

Preferably, step S44 includes the steps of:

step S441: performing spatial pyramid pooling multi-layer sampling on the tumor lesion convolution feature network by using a multi-scale sampling algorithm to generate tumor lesion convolution feature data;

step S442: performing convolution feature mapping on the tumor lesion convolution feature data by using a neuron activation mapping algorithm to generate a tumor lesion convolution feature vector;

step S443: vector stitching is carried out by utilizing the tumor lesion convolution feature vector, and a tumor lesion convolution feature sequence is generated.

According to the invention, a multi-scale sampling algorithm is used for carrying out space pyramid pooling multi-layer sampling on a tumor lesion convolution feature network, feature information with different resolutions and different feature dimensions can be effectively extracted by utilizing sampling pooling operation with different scales, and tumor lesion convolution feature data is generated, so that the richness and the accuracy of features are improved, a neuron activation mapping algorithm is used for carrying out convolution feature mapping on the tumor lesion convolution feature data, high-dimensional feature data can be mapped to a low-dimensional feature space, so that the processing and analysis are easier, meanwhile, the mapping operation can enable the feature data to be denser, more effective information is provided in a smaller space, redundancy and noise in original features can be eliminated, vector splicing is carried out by utilizing tumor lesion convolution feature vectors, the processed feature vectors are combined into a sequence, so that the generated feature sequence can retain a large amount of feature information, has strong descriptive and distinguishing properties, can provide abundant feature support for subsequent model prediction, the feature data can be processed by a plurality of scale sampling algorithms, the neuron activation mapping algorithm, vector convolution mapping algorithm and the like, and the feature analysis is carried out on the lesion prediction model, and the feature analysis is more effective, so that the feature prediction stability is improved.

Preferably, step S5 comprises the steps of:

step S51: defining a time step of a tumor feature trend prediction model by using a time sequence algorithm, and generating a tumor prediction time stamp;

step S52: performing time sequence binding on the tumor feature trend prediction model and the tumor feature trend prediction time stamp by using a linear interpolation algorithm to generate a tumor feature trend prediction time sequence model;

step S53: and carrying out data fusion on the second tumor case follow-up data and the tumor characteristic trend prediction model by using a neural network algorithm, and generating third tumor case follow-up data.

According to the invention, a time step is defined for the tumor feature trend prediction model through a time sequence algorithm to generate a tumor prediction time stamp, so that the tumor feature data can be predicted in future according to the preset time step, the development and change of tumors can be better known and processed, the time sequence binding is carried out on the tumor prediction time stamp and the tumor feature trend prediction model through a linear interpolation algorithm to generate a tumor feature trend prediction time sequence model, the generated model can effectively capture the change and trend of the tumor feature data, the future development trend and change can be further accurately predicted, the neural network algorithm is utilized to carry out data fusion on the second tumor case follow-up data and the tumor feature trend prediction model, the third tumor case follow-up data can be generated, the generated data can better reflect the actual situation, and meanwhile, the tumor development trend can be further accurately predicted through the model, so that more accurate support is provided for medical practice, the tumor feature data is processed and analyzed through the time sequence algorithm, the linear interpolation algorithm and the neural network algorithm, and the third tumor case follow-up data is successfully generated, and more accurate and effective support is provided for medical practice.

Preferably, step S6 comprises the steps of:

step S61: carrying out data ciphertext conversion on the third tumor case follow-up data by using an affine encryption algorithm to generate tumor case follow-up ciphertext data;

step S62: affine encryption is carried out on the tumor case follow-up ciphertext data by utilizing an affine encryption algorithm of the tumor case follow-up data, so that affine encryption data of the tumor case are generated;

step S63: performing network scheduling slicing on affine encryption data of tumor cases by using a distributed network slicing algorithm to generate affine encryption data slices of a plurality of tumor cases;

step S64: uploading affine encrypted data slices of a plurality of tumor cases to a tumor information data management system so as to realize real-time management of follow-up data;

according to the invention, the affine encryption algorithm is used for carrying out data ciphertext conversion on the third tumor case follow-up data to generate tumor case follow-up ciphertext data, the privacy and the data safety of the follow-up data can be effectively protected, the affine encryption algorithm is used for carrying out affine encryption on the tumor case follow-up ciphertext data to generate tumor case affine encryption data, the generated data can be provided for users with authority access for analysis and use on the premise of ensuring the data safety, the distributed network slicing algorithm is used for carrying out network scheduling slicing on the tumor case affine encryption data to generate a plurality of tumor case affine encryption data slices, the integrity, the safety and the reliability of the data can be better ensured by processing the generated data, the plurality of tumor case affine encryption data slices can be uploaded to a tumor information data management system, so that the follow-up data can be managed in real time, the medical staff can view, manage and maintain the data at any time, effective support and guarantee are provided for clinical medical practice, and more reliable support and safety are provided for medical practice slicing by utilizing the encryption algorithm.

Preferably, the affine encryption algorithm in step S61 is specifically:

；

wherein ,representing the use of the public key pk for the input data>Encryption result obtained by affine encryption, +.>Ciphertext for data of input model, ++>Encryption key generation element for affine encryption algorithm, < ->Randomly selected cardinality for affine algorithm, +.>Modulus being a specific power of power, < >>For hash function value, converting plaintext data into hash value with fixed length,/for hash function value>Random number for hash function, < >>Hash value obtained by inputting a hash function for model plaintext data, < >>First section of ciphertext taken for encryption, < >>Constructing a generator of an encryption key for the first section of ciphertext, ">Hash function value with radix for random selection, < +.>First section of ciphertext taken for encryption, < >>Weight coefficient for generating element of first section ciphertext and second section ciphertext, ++>The hash function value based on the randomly selected cardinality for the second ciphertext.

The invention is realized byEncrypting the plaintext data to protect the privacy of the data, the hash function h (x) has the function of converting the plaintext data into a hash value with a fixed length to realize an irreversible conversion process, the hash function generates a unique hash value by processing and operating the plaintext data, the privacy of the data can be effectively protected, the hash function random number r0 has the function of increasing the safety and the unpredictability of the hash function, and the statistical analysis and the cracking of the plaintext data by adversaries can be avoided by selecting an unpredictable random number >For the second section of ciphertext based on the hash function value with the base selected randomly, the formula has an affine encryption scheme with the characteristics of random property, hash confusion, modulus protection and multi-section encryption, can effectively ensure the safety and privacy of data, supports various processing and computing operations of encrypted data, and is characterized in that the formula is provided with the affine encryption scheme with the characteristics of random property, hash confusion, modulus protection and multi-section encryption, and is provided with the functions of processing and computing the encrypted data>Performing power operation and affine encryption on the first section of ciphertext and the hash value, multiplying the power of the first section of ciphertext by the second section of ciphertext m2, and finally performing square root processing to mask the characteristics and values of the original data, wherein the first section of ciphertext and the hash value have reversibility and masking property with an encryption algorithm in the process, can process the encrypted data, protect the data privacy, simultaneously does not influence the integrity and accuracy of the data, and protects the data privacyIn the aspect of security assurance, the affine encryption algorithm can realize various applications such as secret calculation, secure storage, trusted sharing and the like, and has wide application prospect and practical significance.

In one embodiment of the present specification, a method and a system for processing data based on tumor case follow-up are provided, including

At least one processor;

a memory communicatively coupled to the at least one processor;

the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the data processing method based on tumor case follow-up as described above.

The invention can update the tumor case data of the patient in real time and extract the characteristic information of the data to generate tumor case mark data by establishing a data processing system based on tumor case follow-up, acquiring first tumor case follow-up data from a hospital case system and carrying out medical detection on the patient by utilizing tumor detection medical equipment, the process can promote doctors to know the tumor condition of the patient more timely and better formulate a treatment scheme for the patient, carry out matrix division on a pretreatment pipeline of the tumor case follow-up data to acquire a plurality of tumor case follow-up data submatrices, carry out matrix division on the data, decompose large-scale data into small-scale matrixes and carry out treatment on the data dimension, thereby improving the calculation efficiency and the accuracy, the tumor lesion characteristics are processed and analyzed by adopting a multiscale sampling algorithm, a neuron activation mapping algorithm, a vector splicing and other technologies to generate a tumor lesion characteristic convolution sequence, the prediction accuracy and the model stability are improved, the tumor characteristic data are processed and analyzed by utilizing a time sequence algorithm and a neural network algorithm to generate third tumor case follow-up data, more accurate and effective support is provided for medical practice, the tumor case follow-up data are protected and managed by utilizing an encryption algorithm and a distributed network slicing algorithm, an effective data security and privacy maintenance mechanism is provided, the system realizes the efficient processing and analysis of the tumor case follow-up data, the tumor prediction accuracy and the model stability are improved, the data security and privacy are ensured,

Provides comprehensive and high-quality medical care service for hospitals.

Drawings

FIG. 1 is a flow chart showing steps of a data processing method based on tumor follow-up;

FIG. 2 is a detailed implementation step flow diagram of step S1;

FIG. 3 is a detailed implementation step flow diagram of step S2;

fig. 4 is a detailed implementation step flow diagram of step S3.

Detailed Description

It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.

The embodiment of the application provides a data processing method and system based on tumor follow-up. The execution main body of the data processing method and system based on tumor follow-up comprises, but is not limited to, the system: mechanical devices, data processing platforms, cloud server nodes, network uploading devices, etc. may be considered general purpose computing nodes of the present application, including but not limited to: at least one of an audio image management system, an information management system and a cloud data management system.

Referring to fig. 1 to 4, the present application provides a data processing method based on tumor follow-up, the method comprising the following steps:

In the embodiment of the present invention, as described with reference to fig. 1, a schematic flow chart of steps of a data processing method and system based on tumor case follow-up according to the present invention is provided, where in this example, the steps of the data processing method based on tumor case follow-up include:

in the embodiment of the invention, first tumor case follow-up data are acquired from a hospital case system, the acquired data comprise basic information, case numbers, follow-up time, treatment schemes, examination results and other data information of a patient, tumor detection medical equipment is used for detecting the patient, latest tumor case follow-up data comprising data information such as tumor size, position, shape, growth speed and the like are acquired, the latest tumor case follow-up data are compared with the first tumor case follow-up data, the first tumor case follow-up data are updated in real time according to differences, second tumor case follow-up data are generated, the updated data comprise the case follow-up time, treatment schemes, examination results and other information, and finally the second tumor case follow-up data are uploaded to the case system for storage.

Step S2: and performing visual projection on the follow-up data of the second tumor case by using a matrix decomposition method to generate a tumor information feature matrix projection diagram.

In the embodiment of the invention, the second tumor case follow-up data is preprocessed, including operations such as data cleaning, normalization, data conversion and the like, the second tumor case follow-up data is converted into a matrix form, the preprocessed matrix is decomposed by a matrix decomposition method, a tumor information feature matrix is generated by a decomposition reconstruction method, the tumor information feature matrix is subjected to visual projection, and a tumor information feature matrix projection diagram is drawn on a two-dimensional plane, wherein the image reflects the distribution condition and the correlation of tumor information features in the case follow-up data.

Step S3: and carrying out trend prediction analysis on the tumor feature matrix projection graph by using a random forest algorithm to generate a tumor lesion trend prediction graph.

In the embodiment of the invention, the tumor information feature matrix is trained, the correlation between the case follow-up data and the tumor lesion trend is trained by adopting a random forest algorithm, a model is generated, a tumor feature matrix projection image is input into the trained model, the random forest algorithm is utilized to conduct trend prediction analysis on the tumor feature matrix projection image, the tumor lesion trend of a patient is predicted, the predicted tumor lesion trend data is visualized, and a tumor lesion trend prediction image is drawn, so that the medical staff can conveniently carry out follow-up of the patient and adjustment of a treatment scheme.

Step S4: and performing expansion convolution and multi-scale sampling on the tumor lesion trend prediction graph by using a cyclic convolution algorithm, and constructing a tumor feature trend prediction model.

In the embodiment of the invention, the tumor feature trend prediction graph is input into a cyclic convolution algorithm, the cyclic convolution algorithm is carried out, the range of trend change is enlarged, the effect of feature extraction is improved, the image after the cyclic convolution is subjected to multi-scale sampling, features with different sizes are extracted, more comprehensive and accurate tumor feature data are obtained, the tumor feature data obtained by multi-scale sampling are input into a model, and the tumor feature trend prediction model is obtained through training, so that the future tumor feature trend of a patient can be accurately predicted by the model.

Step S5: binding a tumor characteristic trend prediction model in a time sequence; and carrying out data fusion on the second tumor case follow-up data and the tumor characteristic trend prediction model by using a neural network algorithm, and generating third tumor case follow-up data.

In the embodiment of the invention, time sequence binding processing is carried out on the tumor feature trend prediction model, the tumor feature trend prediction model is aligned with a patient follow-up time sequence, so that the model can predict future trend according to current time, second tumor case follow-up data and the tumor feature trend prediction model are subjected to data fusion, a prediction model is established by using a neural network algorithm, future tumor lesion trend is predicted, time sequence binding processing is carried out on the tumor feature trend prediction model, the future trend can be predicted according to current time, data fusion is carried out on the second tumor case follow-up data and the tumor feature trend prediction model by using the neural network algorithm, and a prediction model is established, so that future tumor lesion trend is predicted. And finally, generating follow-up data of the third tumor case according to the prediction result.

In the embodiment of the invention, the encryption processing is carried out on the follow-up data of the third tumor case, the affine encryption algorithm is used for ensuring the privacy and the safety of the data, the affine encryption data of the tumor case is generated according to the encryption algorithm, the data can be transmitted and stored under the condition of not exposing the privacy information, the encryption data is divided into a plurality of fragments by using the distributed network slicing algorithm, and the fragments are uploaded to a hospital case system, so that the real-time management of the follow-up data is realized.

In the embodiment of the present invention, as described with reference to fig. 2, a detailed implementation step flow diagram of the step S1 is shown, and in one embodiment of the present specification, the detailed implementation step of the step S1 includes:

In the embodiment of the invention, first tumor case follow-up data is acquired through a hospital case system, the data comprises patient information data, patient follow-up data, patient medical record data, tumor chemotherapy scheme data and the like, medical detection is carried out by adopting tumor detection medical equipment, such as a nuclear magnetic resonance imager, an X-ray instrument, a blood analyzer, a biochemical analyzer, a nucleic acid analyzer and the like, latest tumor case follow-up data of a patient, such as latest tumor size, lesion degree, metastasis condition and the like, are acquired from the tumor detection medical equipment, characteristic information extraction is carried out on the latest tumor case follow-up data, such as information of tumor size, lesion degree, metastasis condition, chemotherapy scheme and the like, tumor case mark data is generated based on the characteristic information extraction result, namely, the information of each patient is converted into binary code mark, the tumor case mark data is matched with the first tumor case follow-up data, tumor information, a chemotherapy scheme and the like in the patient follow-up data are updated in real time, and second tumor case follow-up data are generated according to the updated first tumor case follow-up data, and the new chemotherapy scheme and the like aiming at the information are contained.

In the embodiment of the present invention, as described with reference to fig. 3, a detailed implementation step flow diagram of step S2 is shown, and in one embodiment of the present specification, the detailed implementation step of step S2 includes:

In the embodiment of the invention, the data preprocessing is carried out on the follow-up data of the second tumor case, the data is cleaned, the missing value, the abnormal value and the repeated value are removed, the data are integrated, and the data from different sources are integrated into a data set and standardized data: and normalizing each feature, for example, subtracting the mean value of each feature and dividing the mean value by the standard deviation of each feature, carrying out matrix division on a follow-up data set subjected to data pretreatment according to a case ID or a time period to obtain a plurality of matrix subsets, carrying out matrix division on the follow-up data set subjected to data pretreatment according to the case ID or the time period, selecting proper topic numbers according to experience or indexes, or carrying out topic number selection by using cross verification and other technologies, randomly initializing a main component matrix W and a coefficient matrix H, alternately updating the W and the H until the minimum error reaches a preset value or the maximum iteration number to obtain a plurality of matrix subsets, extracting main features of the follow-up data according to the main component matrix and the minimum error matrix, carrying out dimension reduction on the feature weight matrix through observing the feature vector of the main component matrix or the coefficient matrix to obtain the extracted feature weight, dividing the feature weight matrix by the maximum value to obtain a normalized feature weight matrix, carrying out transposition on the feature weight matrix to obtain a two-dimensional projection matrix, and carrying out the transposition weight matrix to obtain a feature projection matrix, and carrying out the two-dimensional projection on the feature matrix to obtain a feature projection matrix.

In the embodiment of the present invention, as described with reference to fig. 4, a detailed implementation step flow diagram of step S3 is shown, and in one embodiment of the present invention, the detailed implementation step of step S3 includes:

In the embodiment of the invention, a tumor feature matrix projection image is cut by utilizing an image geometric structure method, a tumor lesion part mark such as a lesion boundary, a lesion size and the like is carried out, geometric transformation such as translation, rotation, scaling and the like is carried out on the tumor feature matrix projection image, the tumor lesion part is aligned to the center of an image, the tumor feature matrix projection image is cut according to the tumor lesion part mark, a tumor lesion feature image and a tumor non-lesion feature image are obtained, a histogram equalization algorithm is utilized to carry out image gray level equalization distribution processing on the tumor feature image, pixel gray values of the tumor feature image are read, a pixel gray level histogram of the tumor feature image is counted, a probability distribution function of each gray value is calculated, gray level equalization distribution processing is carried out according to the probability distribution function, and a high-definition tumor feature image is obtained, selecting a proper feature point detection algorithm, such as SIFT, SURF and the like, carrying out tumor lesion feature point marking on a high-definition tumor lesion feature map by utilizing the feature point detection algorithm, generating tumor lesion feature point coordinates, carrying out gridding treatment on tumor lesion feature points, dividing the area where the tumor lesion feature points are positioned into a plurality of grids, constructing a random forest model, taking the coordinates of the tumor lesion feature points as sample input, carrying out classification prediction on each grid, distributing the tumor lesion feature points into the grids, calculating the lesion degree of each grid, such as lesion area, lesion density and the like, according to the lesion follow-up data, constructing a tumor lesion trend prediction model, such as linear regression, support vector regression and the like, taking the lesion degree of the tumor lesion feature point grids as input, and carrying out trend prediction on the tumor feature point grid by using a prediction model, and carrying out trend calculation on the tumor feature point grid by using a tumor lesion trend prediction formula to generate a tumor lesion trend prediction graph.

In one embodiment of the present disclosure, the tumor lesion trend prediction formula in step S35 is specifically:

；

wherein ,diffusion range size value predicted for tumor lesion trend,/-for>Weight of the early tumor lesion in the normal part>For early growth rate of tumor->Is at presentMeasured tumor spread coordinate range,/->Coordinate range for early tumor spread +_>Weight of the current tumor lesion in normal part>For the current growth rate of the tumor, < > is->Tumor diffusion coordinate range for last time point detection, +.>Is the degree of periodic concussion of tumor development, +.>Is the oscillation period of the tumor development period, +.>For the extent of the influence of time on the tumor growth rate, < + >>Weight of trend of slowing down tumor growth rate with time, ++>Is the upper limit of the tumor size.

The invention is realized byComparing the early tumor diffusion range with healthy parts, regulating the influence degree of early tumor diffusion by weight, mapping the input value to an output value ranging from 0 to 1, and having the characteristics of smoothness, continuity, nonlinearity and the like, by the piecewise fitting, the size of the diffusion range of the tumor can be more accurately predicted, the influence of limiting the early tumor diffusion by taking A as a limit can be calculated, and the influence of limiting the early tumor diffusion by taking A as a limit is calculated >The method comprises the steps of carrying out secondary evaluation and judgment on the current illness state of a patient, evaluating the illness state of the patient, namely the invasion degree of the illness part of the patient, effectively distinguishing early-stage illness from late-stage illness through calculation of the formula, further determining the size and range of illness, being beneficial to diagnosis of tumors, reducing the progress and the diffusion of illness through early-stage diagnosis, improving the treatment effect, avoiding excessive injury in the treatment process, and obtaining the medicine for treating the illness state of the patient by using the formulaThe prediction result is more reliable by limiting the influence degree of the tumor size on the prediction result, the diffusion range of the tumor lesion is gradually increased along with the increase of the lesion part, so that the error of the prediction result is increased, if the tumor size factor is not considered, the prediction result has certain uncertainty, inconvenience and excessive financial burden can be caused to a patient, meanwhile, the reliability and the repeatability of a model can be influenced for doctors and researchers, and therefore, the problems can be effectively avoided by limiting the tumor size, the influence of the tumor on the prediction diffusion range is limited by the formula, the size and the prediction value of the tumor are restrained by the formula, and the prediction result is more reliable and accurate.

In one embodiment of the present specification, the specific steps of step S33 are as follows:

In the embodiment of the invention, a proper characteristic point detection algorithm is selected, such as SIFT, SURF and the like, a characteristic point detection algorithm is utilized to carry out tumor lesion characteristic point marking on a high-definition tumor lesion characteristic map, a tumor lesion characteristic point coordinate is generated, a characteristic value threshold limiting method is used to carry out filtering and screening on initial tumor lesion characteristic map characteristic point data, characteristic value calculation is carried out on the initial tumor lesion characteristic map characteristic point data, such as a Harris corner detection algorithm, a corner response value can be calculated, a proper threshold value is set according to the characteristic value of the characteristic point data, filtering and screening are carried out on points with the characteristic value smaller than the threshold value, the reference tumor lesion characteristic point data is obtained, a space three-dimensional coordinate is calculated according to a grid center point and a grid boundary point, the space three-dimensional coordinate marking is carried out on the reference tumor lesion characteristic point data, and the space three-dimensional coordinate of each characteristic point is recorded, so that the tumor lesion characteristic point coordinate is obtained.

In one embodiment of the present specification, the specific step of step S4 is:

In the embodiment of the invention, a tumor lesion trend prediction graph is subjected to convolution pretreatment through a super-pixel convolution network, the tumor lesion trend prediction graph is input into the super-pixel convolution network for convolution pretreatment, a tumor lesion prediction feature sample set is obtained, the tumor lesion prediction feature sample set is diced according to a certain size to obtain a tumor lesion prediction convolution data block, a cyclic convolution algorithm is utilized to carry out convolution operation on the tumor lesion prediction convolution data block to obtain a convolution feature sequence, the tumor lesion convolution data block is subjected to expansion operation to strengthen edge features, the tumor lesion convolution data block is subjected to convolution operation by utilizing the expansion convolution algorithm to obtain a tumor lesion prediction convolution feature network, a multi-scale sampling algorithm is utilized to carry out spatial pyramid pooling multi-layer sampling on the tumor lesion convolution feature network to obtain feature graphs of a plurality of scales, the feature graphs are combined by utilizing the pooling operation to obtain a tumor lesion prediction convolution feature sequence, a data mining means is used for finding association rules in the tumor lesion prediction convolution feature sequence, such as an Apriori algorithm, an FP-Grow algorithm and the like, the tumor lesion prediction feature sequence is searched by utilizing data mining means, and the association rules in the data mining means such as the Apriori algorithm, the FP-Grow algorithm is used for finding the association rules in the convolution feature sequence, and the association rule in the tumor lesion prediction feature sequence, and the association algorithm is found by utilizing the rule such as the Apriori algorithm.

In one embodiment of the present specification, the specific steps of step S44 are as follows:

In the embodiment of the invention, a multi-scale sampling algorithm is utilized to carry out space pyramid pooling multi-layer sampling on a tumor lesion convolution feature network, for each layer of the tumor lesion convolution feature network, a multi-scale convolution check is used to carry out convolution operation to obtain a plurality of scale feature images, for each feature image, a space pyramid pooling algorithm is used to carry out multi-layer sampling to obtain feature vectors with fixed dimensions, the feature vectors of each layer are spliced together to obtain tumor lesion convolution feature data, the tumor lesion convolution feature data is input into a neuron activation mapping algorithm to obtain nonlinear mapped feature data, the mapped feature data is normalized again to ensure the stability of subsequent processing, for each tumor lesion, the corresponding convolution feature vector can be generated by using the neuron activation mapping algorithm, and the convolution feature vectors of each tumor lesion are spliced according to a certain rule to obtain a tumor lesion convolution feature sequence, such as time sequence arrangement.

In one embodiment of the present specification, the specific steps of step S5 are as follows:

In the embodiment of the invention, a predicted time range is determined, a time sequence algorithm is utilized to define a time step of a tumor feature trend prediction model, a time stamp is generated for each time point in the predicted time range, a tumor feature value of the time point is predicted by the tumor feature trend prediction model for each time stamp, the tumor feature values of each time stamp are connected together to obtain a time sequence model, and a linear interpolation algorithm is utilized to complement the missing time point in the time sequence model. For example, for the missing time point t, linear interpolation is performed by using characteristic values of the previous and next time points, the second tumor case follow-up data is input into a neural network model to obtain corresponding patient characteristic vectors, the patient characteristic vectors and time stamps of the second tumor case follow-up data are spliced, prediction of the tumor characteristic values is performed on each time stamp by using the generated time sequence model, and the predicted tumor characteristic values and the corresponding time stamps are spliced to obtain third tumor case follow-up data.

In one embodiment of the present specification, the specific steps of step S6 are as follows:

In the embodiment of the invention, for the third tumor case follow-up data, pretreatment including data cleaning, duplicate removal, outlier processing and the like is needed, the third tumor case follow-up data is encrypted by utilizing an affine encryption algorithm, the third tumor case follow-up data is converted into tumor case follow-up ciphertext data, the tumor case follow-up ciphertext data is encrypted by utilizing the tumor case follow-up data affine encryption algorithm, tumor case affine encryption data is generated, the tumor case affine encryption data can be subjected to simulated attack treatment so as to verify the security of data encryption, such as standard desensitization, differential privacy and the like, a proper distributed network slicing algorithm is selected for network scheduling slicing, common algorithms including hash slicing, regional slicing, mean slicing and the like are specifically determined according to the characteristics and security requirements of the data, parameters such as the number, the size, the position and the like of slices are set according to actual conditions, the security and the integrity of the tumor case follow-up ciphertext data are uploaded to a tumor information data management system, the follow-up slices are marked for subsequent combination of the security of the data, the signature is needed in the subsequent combination of the recovery data, the security of the data is ensured, the security of the encryption data is ensured, and the security is needed to be transmitted, and the like.

In one embodiment of the present specification, the affine encryption algorithm in step S55 is specifically:

；

The invention is realized byEncrypting the plaintext data to protect the privacy of the data, the hash function h (x) has the function of converting the plaintext data into a hash value with a fixed length to realize an irreversible conversion process, the hash function generates a unique hash value by processing and operating the plaintext data, the privacy of the data can be effectively protected, the hash function random number r0 has the function of increasing the safety and the unpredictability of the hash function, and the statistical analysis and the cracking of the plaintext data by adversaries can be avoided by selecting an unpredictable random number >For the second section of ciphertext based on the hash function value with the base selected randomly, the formula has an affine encryption scheme with the characteristics of random property, hash confusion, modulus protection and multi-section encryption, can effectively ensure the safety and privacy of data, supports various processing and computing operations of encrypted data, and is characterized in that the formula is provided with the affine encryption scheme with the characteristics of random property, hash confusion, modulus protection and multi-section encryption, and is provided with the functions of processing and computing the encrypted data>The method comprises the steps of performing power operation and affine encryption on a first section of ciphertext and a hash value, multiplying the power of the first section of ciphertext by a second section of ciphertext m2, and finally performing square root processing, so that characteristics and values of original data are covered, in the process, the method has reversibility and covering property with an encryption algorithm, can process the encrypted data, protect data privacy, does not influence the integrity and accuracy of the data, and can realize various applications such as secret calculation, safe storage and trusted sharing in the aspects of data privacy protection and safety assurance, and the affine encryption algorithm has wide application prospect and practical significance.

In one embodiment of the present specification, a method and system for data processing based on tumor case follow-up is provided, comprising,

at least one processor;

at least one processor in communication with the memory;

the memory stores a computer program executable by the at least one processor, the computer program being executable by the at least one processor

At least one processor executing to enable the at least one processor to perform the data processing method of any one of the above based on tumor case follow-up.

Provides comprehensive and high-quality medical care service for hospitals.

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.

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.

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.

Claims

1. A data processing method based on tumor case follow-up, which is characterized by comprising the following steps:

step S2: performing visual projection on the follow-up data of the second tumor case by using a matrix decomposition method to generate a tumor feature matrix projection diagram;

2. The method according to claim 1, wherein the specific steps of step S1 are:

3. The method according to claim 2, wherein the specific steps of step S2 are:

4. A method according to claim 3, wherein the specific step of step S3 is:

step S31: image cutting is carried out on the tumor lesion part by utilizing an image geometry method to the tumor feature matrix projection graph, and a tumor lesion feature graph and a tumor non-lesion feature graph are generated;

the tumor lesion trend prediction formula in step S35 specifically includes:

；

5. The method according to claim 4, wherein the specific steps of step S33 are:

6. The method according to claim 5, wherein the specific step of step S4 is:

7. The method according to claim 6, wherein the specific steps of step S44 are:

8. The method according to claim 7, wherein the specific steps of step S5 are:

9. The method according to claim 8, wherein the specific step of step S6 is:

the affine encryption algorithm of the tumor case follow-up data in step S62 specifically includes:

；

wherein ,representing the use of the public key pk for the input data>Encryption result obtained by affine encryption, +.>Ciphertext for data of input model, ++>Encryption key generation element for affine encryption algorithm, < ->Randomly selected cardinality for affine algorithm, +.>Modulus being a specific power of power, < >>For hash function value, converting plaintext data into hash value with fixed length,/for hash function value >Random number for hash function, < >>Hash value obtained by inputting a hash function for model plaintext data, < >>First section of ciphertext taken for encryption, < >>Constructing a generator of an encryption key for the first section of ciphertext, ">Hash function value with radix for random selection, < +.>First section of ciphertext taken for encryption, < >>Weight coefficient for generating element of first section ciphertext and second section ciphertext, ++>The hash function value based on the randomly selected cardinality for the second ciphertext.

10. A data processing system based on tumor case follow-up comprising:

at least one processor;

at least one processor in communication with the memory;

the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the oncology case follow-up-based data processing method of any one of claims 1 to 9.