CN114300143A - Hierarchical scheme prediction method, system, device and storage medium - Google Patents
Hierarchical scheme prediction method, system, device and storage medium Download PDFInfo
- Publication number
- CN114300143A CN114300143A CN202111629301.7A CN202111629301A CN114300143A CN 114300143 A CN114300143 A CN 114300143A CN 202111629301 A CN202111629301 A CN 202111629301A CN 114300143 A CN114300143 A CN 114300143A
- Authority
- CN
- China
- Prior art keywords
- target
- data
- model
- fusion
- prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The application relates to a hierarchical scheme prediction method, a system, a device and a storage medium, wherein the method comprises the following steps: acquiring training set data, wherein the training set data comprises multi-source medical characteristic data of a plurality of target objects; establishing an initial fusion model, performing iterative optimization on the initial fusion model based on training set data to obtain a target fusion model, and outputting the fusion medical characteristics of each target object; performing iterative training on a preset initial prediction model based on the fusion medical characteristics of each target object and a preset grading scheme label to obtain a target prediction model; and acquiring multi-source medical characteristic data of the object to be assisted, and obtaining a prediction grading scheme of the object to be assisted based on the target fusion model and the target prediction model. The method and the device have the advantages that the multi-source medical characteristic data of the object to be assisted are subjected to fusion expression, and the prediction grading scheme of the object to be assisted is recommended quickly based on the intelligent prediction model.
Description
Technical Field
The present application relates to the field of smart medical technology, and in particular, to a hierarchical scheme prediction method, system, device, and storage medium.
Background
With the development of artificial intelligence, AI models are applied in more and more fields. In the medical field, most diseases adopt a staged and graded treatment scheme, so that the grading scheme of target diseases can also be predicted by adopting an intelligent model, thereby being beneficial to improving the working efficiency of doctors.
The artificial intelligence technology represented by deep learning can learn massive cases, deeply excavate the commonness among patients in the same grade, and finally give a prediction grading scheme of diseases as an auxiliary opinion of doctors. The existing artificial intelligence algorithm model is relatively fixed and single, for example, a convolutional neural network for processing image data, a long-time memory network for processing time sequence signals and the like exist, but a fusion prediction mode aiming at multi-dimensional and diversified information needs to be further researched.
Disclosure of Invention
In view of the above, the present application provides a classification scheme prediction method, system, device and storage medium, which are used to solve the problem of prediction of the classification scheme based on multi-source disease data fusion in the prior art.
In order to solve the above problem, in a first aspect, the present application provides a hierarchical scheme prediction method, including:
acquiring training set data, wherein the training set data comprises multi-source medical characteristic data of a plurality of target objects;
establishing an initial fusion model, performing cyclic iteration optimization on the initial fusion model based on the training set data to obtain a target fusion model, and outputting the fusion medical characteristics of each target object;
performing iterative training on a preset initial prediction model based on the fusion medical characteristics of each target object and a preset grading scheme label to obtain a target prediction model;
and acquiring multi-source medical characteristic data of the object to be assisted, and obtaining a prediction grading scheme of the object to be assisted based on the target fusion model and the target prediction model.
Optionally, the multi-source medical feature data is structured feature vectors of multiple categories, and the structured feature vectors include medical record feature vectors and image feature vectors.
Optionally, before acquiring the multi-source medical feature data of the target objects/the objects to be assisted, the method further includes:
acquiring medical record data and image data of an object to be processed;
the medical record data of the object to be processed is subjected to standardization processing, and medical record characteristic vectors corresponding to the object to be processed are obtained;
extracting an interested region of image data of an object to be processed and performing feature calculation to obtain an image feature vector corresponding to the object to be processed;
the object to be processed is a target object or an object to be assisted.
Optionally, the initial fusion model includes a plurality of fallback coding network models and a mapping regression network model; then, the iterative optimization of the initial fusion model based on the training set data loop to obtain a target fusion model and output the fusion medical characteristics of each target object includes:
presetting initial implicit expression characteristics, inputting the initial implicit expression characteristics to each backspacing coding network model, and iteratively and optimally training each backspacing coding network model by taking each group of training sample data as an output reference to obtain a backspacing coding network model which is completely trained and target implicit expression characteristics which are correspondingly optimized by each group of training sample data;
iteratively and optimally training the mapping regression network model based on each group of training sample data and the corresponding optimized target implicit expression characteristics to obtain a completely trained mapping regression network model which is used as a target fusion model, and outputting real implicit expression characteristics of each group of training sample data corresponding to conversion based on the target fusion model and used as fusion medical characteristics of a corresponding target object;
wherein each set of training sample data comprises multi-source medical characteristic data of a corresponding target object.
Optionally, the preset initial prediction model includes a support vector machine, a random forest, a logistic regression, a multilayer perceptron, a decision tree, or a neural network.
Optionally, before outputting the fused medical features of each target object based on the target fusion model, the method further comprises:
and acquiring test set data, and testing the target fusion model by using the test set data.
Optionally, medical record data and image data of the target object/the object to be assisted are acquired, wherein the medical record data includes basic information of a patient, symptom information and physical examination information; the image data includes an X-ray film image and an MRI image of the target diseased part.
In a second aspect, the present application provides a hierarchical scheme prediction system, the system comprising:
the data acquisition module is used for acquiring training set data, and the training set data comprises multi-source medical characteristic data of a plurality of target objects;
the training fusion model module is used for establishing an initial fusion model, optimizing the initial fusion model based on the training set data cycle iteration to obtain a target fusion model, and outputting the fusion medical characteristics of each target object;
the training prediction model module is used for carrying out iterative training on a preset initial prediction model based on the fusion medical characteristics of each target object and a preset grading scheme label to obtain a target prediction model;
and the prediction module is used for acquiring multi-source medical characteristic data of the object to be assisted and obtaining a prediction grading scheme of the object to be assisted based on the target fusion model and the target prediction model.
In a third aspect, the present application provides a computer device, which adopts the following technical solution:
a computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the hierarchical scheme prediction method when executing the computer program.
In a fourth aspect, the present application provides a computer-readable storage medium, which adopts the following technical solutions:
a computer-readable storage medium, having stored thereon a computer program which, when being executed by a processor, carries out the steps of the hierarchical scheme prediction method.
The beneficial effects of adopting the above embodiment are: the method obtains training set data, wherein the training set data comprises multi-source medical characteristic data of a plurality of target objects, so that a model can be trained subsequently by using the multi-source medical characteristic data; the initial fusion model is established, the initial fusion model is iteratively optimized based on training set data to obtain a target fusion model, and the fusion medical characteristics of each target object are output, so that fusion expression of multi-source medical characteristic data is realized, and a target fusion model with complete training is obtained; performing iterative training on a preset initial prediction model based on the fusion medical characteristics of each target object and a preset grading scheme label to obtain a target prediction model, so that the mapping relation between the diagnosis information of the target object and the grading scheme can be determined; then multi-source medical characteristic data of the object to be assisted are obtained, the fusion medical characteristic of the object to be assisted can be obtained based on the target fusion model, then the prediction grading scheme of the object to be assisted can be intelligently recommended based on the target prediction model, so that fusion prediction of the multi-source disease data can be completed, and the corresponding prediction grading scheme can be output.
Drawings
FIG. 1 is a flow chart of a method of one embodiment of a hierarchical scheme prediction method provided herein;
FIG. 2 is a flow chart of a method of another embodiment of a hierarchical scheme prediction method provided herein;
FIG. 3 is a flowchart of an embodiment of a hierarchical scheme prediction method provided by the present application in step S102;
FIG. 4 is a functional block diagram of an embodiment of a hierarchical scheme prediction system provided herein;
FIG. 5 is a schematic block diagram of an embodiment of a computer device provided herein.
Detailed Description
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate preferred embodiments of the application and together with the description, serve to explain the principles of the application and not to limit the scope of the application.
In the description of the present application, "a plurality" means two or more unless specifically limited otherwise.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.
Referring to fig. 1, the present application provides a hierarchical scheme prediction method, including the steps of:
s101, acquiring training set data, wherein the training set data comprises multi-source medical characteristic data of a plurality of target objects;
s102, establishing an initial fusion model, iteratively optimizing the initial fusion model based on training set data to obtain a target fusion model, and outputting fusion medical characteristics of each target object;
s103, performing iterative training on a preset initial prediction model based on the fusion medical characteristics of each target object and a preset grading scheme label to obtain a target prediction model;
and S104, acquiring multi-source medical characteristic data of the object to be assisted, and obtaining a prediction grading scheme of the object to be assisted based on the target fusion model and the target prediction model.
In this embodiment, the target subject is a patient who has a target disease and has a high probability of recovery after being treated by the grading scheme. In the present embodiment, the target disease is knee joint bone disease, and in other embodiments, the target disease may be hip degenerative disease, carotid plaque disease, or other diseases. In this embodiment, the medical feature data of the target object/the object to be assisted is a plurality of categories of structured feature vectors, including medical record feature vectors and a plurality of image feature vectors, where the image feature vectors include X-ray image feature vectors and MRI image feature vectors, and in other embodiments, the image feature vectors may also include multi-modal image feature vectors such as X-ray image feature vectors, MRI image feature vectors, and CT image feature vectors. In this embodiment, the rating scheme label for each target object is the rating of the treatment scheme corresponding to the target object.
In the embodiment, training set data is obtained, and the training set data comprises multi-source medical characteristic data of a plurality of target objects, so that a model can be trained subsequently by using the multi-source medical characteristic data; the initial fusion model is established, the initial fusion model is iteratively optimized based on training set data to obtain a target fusion model, and the fusion medical characteristics of each target object are output, so that fusion expression of multi-source medical characteristic data is realized, and a target fusion model with complete training is obtained; performing iterative training on a preset initial prediction model based on the fused medical features of each target object and a preset grading scheme label to obtain a target prediction model, so that the mapping relation between the medical information of the target object and the grading scheme can be determined; then multi-source medical characteristic data of the object to be assisted are obtained, the fusion medical characteristic of the object to be assisted can be obtained based on the target fusion model, then the prediction grading scheme of the object to be assisted can be intelligently recommended based on the target prediction model, so that fusion prediction of the multi-source disease data can be completed, and the corresponding prediction grading scheme can be output.
In an embodiment, referring to fig. 2, before acquiring multi-source medical feature data of a plurality of target objects/objects to be assisted in step S101 or step S104, the hierarchical scheme prediction method of the embodiment further includes:
s201, acquiring medical record data and image data of an object to be processed;
s202, performing standardization processing on medical record data of an object to be processed to obtain a medical record characteristic vector corresponding to the object to be processed; the object to be processed is a target object or an object to be assisted;
s203, extracting the region of interest of the image data of the object to be processed and performing feature calculation to obtain an image feature vector corresponding to the object to be processed.
In this embodiment, the medical record data of the object to be processed includes basic information of the patient, symptom information, physical examination information, and the like; wherein, the basic information of the patient comprises age, sex, height, weight, BMI, daily and working requirements, etc.; symptom information includes pain, limitation of movement, swelling, etc.; the physical examination information comprises the physical examination content of the whole set of knee joints. In other embodiments, the medical record data of the subject to be treated further includes a grade of the target diseased site, such as a knee degenerative KL grade of a knee joint. The image data includes X-ray film image data and MRI image data of the target diseased part.
Considering that the X-ray image data and the MRI image data both belong to image data, and the medical record data is serialized data, the medical record data and the image data of the object to be processed need to be structured and arranged to form an isomorphic characteristic information set and serve as input data of a subsequent intelligent model.
Specifically, in the embodiment, a statistical-based method is adopted to standardize medical record data of an object to be processed, and illustratively, for height information of a patient, the embodiment calculates height distribution (i.e., mean and variance) of a general population from a big data sample, standardizes the height of the object to be processed, i.e., the patient, and maps the height of the object to be processed from dimensional data to a range from-1 to 1 as a characteristic value of one dimension of the patient; by carrying out standardization processing on medical record data of other dimensions, a group of medical record characteristic values of the processed object can be obtained and can be used as medical record characteristic vectors.
Further, in this embodiment, a method based on an anatomical structure is adopted to perform feature extraction on X-ray image data and MRI image data of an object to be processed, in a specific embodiment, for example, for X-ray image data and MRI image data of a knee joint, a doctor performs delineation of an anatomical structure in two medical images of the knee joint, specifically, performs knee joint related anatomical structure segmentation (such as femur, tibia, etc.) on the X-ray image and the MRI image respectively, extracts a region of interest (i.e., a mask image of the anatomical structure) in the image, and further calculates related parameters thereof, exemplarily, the number of voxels included in each region of interest is counted in the mask image and multiplied by an image resolution, so as to calculate an area or a volume of the corresponding structure, the intensity of a signal of the region of interest is counted in an original image, so as to calculate an intensity of the corresponding structure, and the like, as the image feature vector of the corresponding image. And respectively performing feature extraction on medical record data, X-ray film image data and MRI image data of a plurality of target objects to obtain medical record feature vectors, X-ray film feature vectors and MRI image feature vectors of each target object, and using the medical record feature vectors, the X-ray film feature vectors and the MRI image feature vectors as a group of training samples, thereby obtaining training set data and test set data.
Through steps S201 and S202, in the embodiment, structured arrangement can be performed on medical record (clinical) information and image information of diversification of a target object to obtain isomorphic feature vectors, which is helpful for solving the problem that an existing artificial intelligence algorithm model cannot process diversified data.
In one embodiment, the initial fusion model comprises a plurality of fallback coding network models and a mapping regression network model; referring to fig. 3, in step S102, an initial fusion model is iteratively optimized based on a training set data loop to obtain a target fusion model, and a fusion medical characteristic of each target object is output, including the following steps:
s301, presetting initial implicit expression characteristics, inputting the initial implicit expression characteristics to each backspacing coding network model, and iteratively and optimally training each backspacing coding network model by taking each group of training sample data as an output reference to obtain a backspacing coding network model which is completely trained and target implicit expression characteristics which are correspondingly optimized by each group of training sample data; each group of training sample data comprises multi-source medical characteristic data of the corresponding target object.
Specifically, randomly initializing a backspacing network parameter to be learned and an initial hidden representation feature; in this embodiment, each set of training samples includes a medical record feature vector, an X-ray film feature vector, and a true feature vector of an MRI image feature vector in three dimensions, so that the initial implicit expression features are randomly set as three initial feature vectors.
Further, fixing initial implicit expression features, sending the initial implicit expression features into each rollback coding network model to obtain three reconstructed feature vectors, calculating loss errors between the reconstructed feature vectors and corresponding real feature vectors by taking a preset first constraint function as constraint, and optimizing each rollback network parameter through a random gradient descent method.
Furthermore, parameters of each backspacing coding network model are fixed, initial implicit expression features are sent to each backspacing coding network model to obtain three reconstruction feature vectors, a preset first constraint function is used as constraint, loss between the three reconstruction feature vectors and corresponding real feature vectors is calculated, and the initial implicit expression features are optimized through a random gradient descent method.
Further, each rollback network parameter and the initial implicit expression characteristic are optimized through joint iteration until the loss function is converged, and then iteration is finished.
In this embodiment, the first constraint function may adopt the following expression:
wherein V represents the feature vector dimension of the target object corresponding to each group of training samples in the training set data, taking the nth target object as an example,
a vth feature vector representing the nth target object,
representing a backspacing coding network model corresponding to the v-th feature vector, | · | | non-calculation2Denotes variance, χnRepresenting an initial implicit representation feature, hnThe learned implicit representation features are represented. Furthermore, in the present embodiment, for the training loss error, a mean square error may be adopted as the loss function.
Through the step S301, the embodiment can deeply mine the internal relation of the target object among the dimensional features, perform better fusion, and facilitate subsequent effective training of the target fusion model.
S302, iteratively optimizing the training mapping regression network model based on each group of training sample data and the corresponding optimized target hidden representation characteristics to obtain a completely-trained mapping regression network model which is used as a target fusion model, and outputting real hidden representation characteristics of each group of training sample data corresponding to conversion based on the target fusion model and used as fusion medical characteristics of a corresponding target object. It should be noted that the target fusion model needs to be tested with the test set data before outputting the fused medical features of each target object.
In this embodiment, the mapping regression network model learns the mapping from the original multi-class structural features to the implicit expression features with the objective implicit expression features as the objective, so that the conversion from the original multi-class structural features to the fused medical features can be directly obtained in the testing stage.
Specifically, the training process of the regression network model is constrained and mapped by using the second constraint function, which is specifically expressed as follows:
wherein N is the total number of the target objects, and Γ (χ)n；Θe) A model of a mapping regression network is represented,
the reconstruction error of the representation mapping regression network model is minimized, and for the training loss error, the mean square error can be used as a loss function. In this embodiment, the mapping regression network model may employ a multi-level perceptron (MLP) as the network framework.
In an embodiment, in step S103, a preset initial prediction model is iteratively trained based on the fusion medical characteristics of each target object and a preset classification scheme label, so as to obtain a target prediction model. And taking the fused medical features of each target object and a preset grading scheme label as new training set data, training a classifier for realizing disease grading scheme prediction based on the implicit expression features, and realizing the grading scheme of intelligently recommended patients. In this embodiment, the classifier may employ a support vector machine, a random forest, a logistic regression, a multi-layer perceptron, a decision tree, or a neural network.
In a specific embodiment, the knee joint disease is classified by expert consensus into early stage, middle stage and late stage for knee joint disease, and the patient is given a graded treatment for different stages of knee joint disease: first-stage: patient education and life exercise guidance. And (2) second stage: drug therapy is given on a primary basis to alleviate symptoms. Third-stage: and on the basis of the second level, performing arthroscopy cleaning. And (4) fourth stage: and (5) cutting bones and protecting knees. And (5) fifth stage: joint replacement surgery. Therefore, for a target subject suffering from knee joint disease, a preset grading scheme label grades the knee joint treatment scheme of the target subject.
In an embodiment, in step S104, multi-source medical feature data of the object to be assisted is obtained, and a prediction classification scheme of the object to be assisted is obtained based on the target fusion model and the target prediction model. In a specific embodiment, for knee and bone joint diseases, the medical record data and the image data of an object to be assisted are subjected to structured arrangement, namely feature extraction, so as to obtain corresponding feature vectors of all dimensions; then inputting the multiple dimensional feature vectors into a target fusion model which is a mapping regression network model and is completely trained to obtain the final fusion medical features of the model; and inputting the fused medical characteristics into a target prediction model, namely a classifier, so as to obtain a prediction diagnosis and treatment scheme of the object to be assisted.
Different from the prior art, the embodiment acquires training set data, and the training set data comprises multi-source medical characteristic data of a plurality of target objects, so that a model can be trained subsequently by using the multi-source medical characteristic data; the initial fusion model is established, the initial fusion model is iteratively optimized based on training set data to obtain a target fusion model, and the fusion medical characteristics of each target object are output, so that fusion expression of multi-source medical characteristic data is realized, and a target fusion model with complete training is obtained; performing iterative training on a preset initial prediction model based on the fused medical features of each target object and a preset grading scheme label to obtain a target prediction model, so that the mapping relation between the medical information of the target object and the grading scheme can be determined; then multi-source medical characteristic data of the object to be assisted are obtained, the fusion medical characteristic of the object to be assisted can be obtained based on the target fusion model, then a prediction grading scheme of the object to be assisted can be intelligently recommended based on the target prediction model, so that the analysis capability of medical big data is improved, the work of professional doctors is assisted, the influence of artificial subjective factors is avoided, objective grading of the grading scheme is realized, the influence of inter-region medical condition difference on diagnosis and treatment results is solved, and doctors can be assisted to rapidly obtain an accurate grading scheme based on statistical results during diagnosis.
It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.
The present embodiment further provides a hierarchical scheme prediction system, which corresponds to the hierarchical scheme prediction methods in the foregoing embodiments one to one. As shown in FIG. 4, the hierarchical scheme prediction system 400 includes an acquisition data module 401, a training fusion model module 402, a training prediction model module 403, and a prediction module 404. The functional modules are explained in detail as follows:
an obtaining data module 401, configured to obtain training set data, where the training set data includes multi-source medical characteristic data of multiple target objects;
a training fusion model module 402, configured to establish an initial fusion model, iteratively optimize the initial fusion model based on training set data cycle, obtain a target fusion model, and output a fusion medical characteristic of each target object;
a training prediction model module 403, configured to perform iterative training on a preset initial prediction model based on the fusion medical features of each target object and a preset classification scheme label, to obtain a target prediction model;
the prediction module 404 is configured to obtain multi-source medical characteristic data of an object to be assisted, and obtain a prediction grading scheme of the object to be assisted based on the target fusion model and the target prediction model.
For the specific definition of each module of the hierarchical scheme prediction system, reference may be made to the above definition of the hierarchical scheme prediction method, which is not described herein again. The various modules in the hierarchical scheme prediction system described above may be implemented in whole or in part by software, hardware, and combinations thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
Referring to fig. 5, the present embodiment further provides a computer device 500, where the computer device 500 may be a computing device such as a mobile terminal, a desktop computer, a notebook, a palmtop computer, and a server. The computer device 500 includes a processor 501, a memory 502, and a display 503. FIG. 5 shows only some of the components of a computer device, but it is to be understood that not all of the shown components are required to be implemented, and that more or fewer components may be implemented instead.
The memory 502 may be an internal storage unit of the computer device, such as a hard disk or a memory of the computer device, in some embodiments. The memory 502 may also be an external storage device of the computer device in other embodiments, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), etc. provided on the computer device. Further, the memory 502 may also include both internal and external storage units of the computer device. The memory 502 is used for storing application software installed on the computer device and various data, such as program codes for installing the computer device. The memory 502 may also be used to temporarily store data that has been output or is to be output. In one embodiment, the memory 502 has stored thereon a hierarchical scheme prediction program 504.
The display 503 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, or the like in some embodiments. The display 503 is used to display information at the computer device and to display a visual user interface. The components 501 and 503 of the computer device communicate with each other via a system bus.
In one embodiment, the following steps are implemented when the processor 501 executes the hierarchical scheme prediction program 504 in the memory 502:
acquiring training set data, wherein the training set data comprises multi-source medical characteristic data of a plurality of target objects;
establishing an initial fusion model, performing iterative optimization on the initial fusion model based on training set data to obtain a target fusion model, and outputting the fusion medical characteristics of each target object;
performing iterative training on a preset initial prediction model based on the fusion medical characteristics of each target object and a preset grading scheme label to obtain a target prediction model;
and acquiring multi-source medical characteristic data of the object to be assisted, and obtaining a prediction grading scheme of the object to be assisted based on the target fusion model and the target prediction model.
The present embodiment also provides a computer-readable storage medium having stored thereon a hierarchical scheme prediction program which, when executed by a processor, performs the steps of:
acquiring training set data, wherein the training set data comprises multi-source medical characteristic data of a plurality of target objects;
establishing an initial fusion model, performing iterative optimization on the initial fusion model based on training set data to obtain a target fusion model, and outputting the fusion medical characteristics of each target object;
performing iterative training on a preset initial prediction model based on the fusion medical characteristics of each target object and a preset grading scheme label to obtain a target prediction model;
and acquiring multi-source medical characteristic data of the object to be assisted, and obtaining a prediction grading scheme of the object to be assisted based on the target fusion model and the target prediction model.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above.
Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
The above description is only for the preferred embodiment of the present application, but the scope of the present application is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present application should be covered within the scope of the present application.
Claims (10)
1. A hierarchical scheme prediction method, the method comprising:
acquiring training set data, wherein the training set data comprises multi-source medical characteristic data of a plurality of target objects;
establishing an initial fusion model, performing cyclic iteration optimization on the initial fusion model based on the training set data to obtain a target fusion model, and outputting the fusion medical characteristics of each target object;
performing iterative training on a preset initial prediction model based on the fusion medical characteristics of each target object and a preset grading scheme label to obtain a target prediction model;
and acquiring multi-source medical characteristic data of the object to be assisted, and obtaining a prediction grading scheme of the object to be assisted based on the target fusion model and the target prediction model.
2. The hierarchical scheme prediction method of claim 1, wherein the multi-source medical feature data is structured feature vectors of multiple categories, and the structured feature vectors include medical record feature vectors and image feature vectors.
3. The hierarchical scheme prediction method according to claim 2, wherein before acquiring the multi-source medical feature data of the plurality of target objects/the object to be assisted, the method further comprises:
acquiring medical record data and image data of an object to be processed;
the medical record data of the object to be processed is subjected to standardization processing, and medical record characteristic vectors corresponding to the object to be processed are obtained;
extracting an interested region of image data of an object to be processed and performing feature calculation to obtain an image feature vector corresponding to the object to be processed;
the object to be processed is a target object or an object to be assisted.
4. The hierarchical scheme prediction method of claim 1, wherein the initial fusion model includes a plurality of fallback coding network models and a mapping regression network model; then, the iterative optimization of the initial fusion model based on the training set data loop to obtain a target fusion model and output the fusion medical characteristics of each target object includes:
presetting initial implicit expression characteristics, inputting the initial implicit expression characteristics to each backspacing coding network model, and iteratively and optimally training each backspacing coding network model by taking each group of training sample data as an output reference to obtain a backspacing coding network model which is completely trained and target implicit expression characteristics which are correspondingly optimized by each group of training sample data;
iteratively and optimally training the mapping regression network model based on each group of training sample data and the corresponding optimized target implicit expression characteristics to obtain a completely trained mapping regression network model which is used as a target fusion model, and outputting real implicit expression characteristics of each group of training sample data corresponding to conversion based on the target fusion model and used as fusion medical characteristics of a corresponding target object;
wherein each set of training sample data comprises multi-source medical characteristic data of a corresponding target object.
5. The hierarchical scheme prediction method according to claim 4, wherein before outputting the fused medical feature of each target object based on the target fusion model, the method further comprises:
and acquiring test set data, and testing the target fusion model by using the test set data.
6. The hierarchical scheme prediction method according to claim 1, wherein the preset initial prediction model is a support vector machine, a random forest, a logistic regression, a multi-layer perceptron, a decision tree, or a neural network.
7. The hierarchical scheme prediction method according to claim 3, wherein medical record data and image data of a target object/object to be assisted are acquired, wherein the medical record data includes patient basic information, symptom information, and physical examination information; the image data includes an X-ray film image and an MRI image of the target diseased part.
8. A hierarchical scheme prediction system, the system comprising:
the data acquisition module is used for acquiring training set data, and the training set data comprises multi-source medical characteristic data of a plurality of target objects;
the training fusion model module is used for establishing an initial fusion model, optimizing the initial fusion model based on the training set data cycle iteration to obtain a target fusion model, and outputting the fusion medical characteristics of each target object;
the training prediction model module is used for carrying out iterative training on a preset initial prediction model based on the fusion medical characteristics of each target object and a preset grading scheme label to obtain a target prediction model;
and the prediction module is used for acquiring multi-source medical characteristic data of the object to be assisted and obtaining a prediction grading scheme of the object to be assisted based on the target fusion model and the target prediction model.
9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the hierarchical scheme prediction method according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program which, when executed by a processor, implements the steps of the hierarchical scheme prediction method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111629301.7A CN114300143A (en) | 2021-12-28 | 2021-12-28 | Hierarchical scheme prediction method, system, device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111629301.7A CN114300143A (en) | 2021-12-28 | 2021-12-28 | Hierarchical scheme prediction method, system, device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114300143A true CN114300143A (en) | 2022-04-08 |
Family
ID=80971283
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111629301.7A Pending CN114300143A (en) | 2021-12-28 | 2021-12-28 | Hierarchical scheme prediction method, system, device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114300143A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117932544A (en) * | 2024-01-29 | 2024-04-26 | 福州城投新基建集团有限公司 | Prediction method, device and storage medium based on multi-source sensor data fusion |
| CN118053054A (en) * | 2024-01-05 | 2024-05-17 | 中国科学院地理科学与资源研究所 | Precipitation fusion product generation method, system, computer equipment and storage medium |
-
2021
- 2021-12-28 CN CN202111629301.7A patent/CN114300143A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118053054A (en) * | 2024-01-05 | 2024-05-17 | 中国科学院地理科学与资源研究所 | Precipitation fusion product generation method, system, computer equipment and storage medium |
| CN117932544A (en) * | 2024-01-29 | 2024-04-26 | 福州城投新基建集团有限公司 | Prediction method, device and storage medium based on multi-source sensor data fusion |
| CN117932544B (en) * | 2024-01-29 | 2024-08-23 | 福州城投新基建集团有限公司 | Prediction method, device and storage medium based on multi-source sensor data fusion |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Bustos et al. | Padchest: A large chest x-ray image dataset with multi-label annotated reports | |
| Azizi et al. | Robust and data-efficient generalization of self-supervised machine learning for diagnostic imaging | |
| CN109599185B (en) | Disease data processing method and device, electronic equipment and computer readable medium | |
| Zhang et al. | Review of the applications of deep learning in bioinformatics | |
| CN109147940B (en) | Apparatus and system for automatically predicting physiological condition from medical image of patient | |
| JP2020500378A (en) | Deep learning medical systems and methods for medical procedures | |
| Beddiar et al. | Automatic captioning for medical imaging (MIC): a rapid review of literature | |
| CN110797101B (en) | Medical data processing method, medical data processing device, readable storage medium and computer equipment | |
| CN114300143A (en) | Hierarchical scheme prediction method, system, device and storage medium | |
| Barhoom et al. | Deep Learning-Xception Algorithm for Upper Bone Abnormalities Classification | |
| EP2126841A2 (en) | Image processing systems and methods | |
| US11741277B2 (en) | Predictive modeling platform for serial casting to correct orthopedic deformities | |
| CN111192660B (en) | Image report analysis method, device and computer storage medium | |
| Hussain et al. | Deep learning-based diagnosis of disc degenerative diseases using MRI: a comprehensive review | |
| CN113688912A (en) | Confrontation sample generation method, device, equipment and medium based on artificial intelligence | |
| Mandal | Developing new machine learning ensembles for quality spine diagnosis | |
| CN111340794B (en) | Quantification method and device for coronary artery stenosis | |
| Yuan et al. | Efficient symptom inquiring and diagnosis via adaptive alignment of reinforcement learning and classification | |
| WO2023108418A1 (en) | Brain atlas construction and neural circuit detection method and related product | |
| Sahoo et al. | A comparative analysis of PGGAN with other data augmentation technique for brain tumor classification | |
| CN114743647A (en) | Medical data processing method, device, equipment and storage medium | |
| CN117672450A (en) | Personalized medicine recommendation method and system based on knowledge graph | |
| Hamamci et al. | A foundation model utilizing chest CT volumes and radiology reports for supervised-level zero-shot detection of abnormalities | |
| CN116485853A (en) | Medical image registration method and device based on deep learning neural network | |
| CN115762721A (en) | Medical image quality control method and system based on computer vision technology |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |