Data storage method and system, equipment and storage medium

Abstract

The embodiment of the invention provides a data storage method, which is applied to a data storage system, the data storage system comprises a prediction module and a retrieval service module, and the data storage method comprises the following steps: the retrieval service module sends first image data to the prediction module; the prediction module predicts the first image data according to the target prediction model to obtain a first retrieval prediction value corresponding to the first image data; wherein the target prediction model is obtained by performing model training through a training data set; the prediction module determines a storage position corresponding to the first image data according to the first retrieval prediction value, that is, before the storage operation of the first image data is carried out, the storage position of the first image data is determined, so that the retrieval service module stores the first image data to the storage position; therefore, the storage of the first image data is more accurate and reasonable, and the storage performance of the system is improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a data storage method and system, device and storage medium. Background Art

[0002] At present, the mainstream medical imaging data storage solution adopts a hierarchical storage method. There are differences in read-write performance and cost of storage devices at different levels. Generally speaking, the higher the read-write performance, the higher the cost of the device. For example, the hot pool uses storage with high read-write efficiency (cache, memory, etc.), and the cold pool uses ordinary mechanical disk storage. In actual scenarios, a patient's imaging data is highly likely not to be viewed again after being reviewed by a doctor once. The storage cost of the hot pool is much higher than that of the cold pool. Since the hot pool stores a large amount of useless data, it consumes too much hot storage resources, which in turn affects the storage performance of the system. Summary of the invention

[0003] The embodiments of the present application provide a data storage method and system, a device and a storage medium, which can predetermine the storage location corresponding to the image data, so that the image data storage is more accurate and reasonable, thereby reducing the consumption of excessive thermal storage resources, and further improving the storage performance of the system.

[0004] The technical solution of the embodiment of the present application is implemented as follows:

[0005] In a first aspect, an embodiment of the present application provides a data storage method, the method is applied to a data storage system, the data storage system includes a prediction module and a retrieval service module, the method includes:

[0006] The review service module sends the first image data to the prediction module;

[0007] The prediction module predicts the first image data according to the target prediction model to obtain a first reference prediction value corresponding to the first image data; wherein the target prediction model is obtained by model training through the training data set;

[0008] The prediction module determines a storage location corresponding to the first image data according to the first review prediction value, so that the review service module stores the first image data in the storage location.

[0009] In a second aspect, an embodiment of the present application provides a data storage system, the data storage system comprising: a prediction module and a retrieval service module,

[0010] The reading service module is used to send the first image data to the prediction module;

[0011] The prediction module is used to predict the first image data according to the target prediction model to obtain a first reference prediction value corresponding to the first image data; wherein the target prediction model is obtained by model training through the training data set;

[0012] The prediction module is further configured to determine a storage location corresponding to the first image data according to the first reference prediction value, so that the reference service module stores the first image data in the storage location.

[0013] In a third aspect, an embodiment of the present application provides a data storage system, the data storage system comprising: a processor and a memory; wherein:

[0014] The memory is used to store a computer program that can be run on the processor;

[0015] The processor is used to execute the data storage method as described above when running the computer program.

[0016] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, characterized in that a computer program code is stored on the storage medium, and when the computer program code is executed by a computer, the data storage method as described above is implemented.

[0017] An embodiment of the present application provides a data storage method and system, a device and a storage medium. The data storage method is applied to a data storage system. The data storage system includes a prediction module and a review service module. The review service module sends first image data to the prediction module; the prediction module predicts the first image data according to a target prediction model to obtain a first review prediction value corresponding to the first image data; wherein the target prediction model is obtained by model training through a training data set; the prediction module determines a storage location corresponding to the first image data according to the first review prediction value, so that the review service module stores the first image data to the storage location. It can be seen that after the review service module sends the first image data to the prediction module, the prediction module can determine the review prediction value corresponding to the image data according to the target prediction model, and then determine the storage location corresponding to the image data according to the review prediction value corresponding to the image data. Since the prediction module in the present application can use the target prediction model to accurately predict the review value corresponding to the image data, it avoids storing a large amount of useless image data in the hot pool. Instead, before performing the image data storage operation, the storage location of the image data is determined so that the review service module stores the image data according to the storage location corresponding to the image data, making the image data storage more accurate and reasonable, thereby reducing the consumption of excessive thermal storage resources and improving the storage performance of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a schematic diagram of the current system storage process;

[0019] Figure 2 Schematic diagram of the data storage method proposed in the embodiment of the present application Figure 1 ;

[0020] Figure 3 Schematic diagram of the data storage system structure proposed in the embodiment of the present application Figure 1 ;

[0021] Figure 4 Schematic diagram of the data storage system structure proposed in the embodiment of the present application Figure 2 ;

[0022] Figure 5 Schematic diagram of the data storage system structure proposed in the embodiment of the present application Figure 3 ;

[0023] Figure 6 A schematic diagram of the process of constructing a target prediction model proposed in an embodiment of the present application;

[0024] Figure 7 Schematic diagram of the data storage method proposed in the embodiment of the present application Figure 2 ;

[0025] Figure 8 Schematic diagram of the structure of the data storage system proposed in the embodiment of the present application Figure 1 ;

[0026] Fig. 9 Schematic diagram of the structure of the data storage system proposed in the embodiment of the present application Figure 2 . DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. It is understood that the specific embodiments described herein are only used to explain the related applications, rather than to limit the applications. It should also be noted that, for ease of description, only the parts related to the related applications are shown in the drawings.

[0028] The storage and retrieval module of medical imaging data is the core functional module of the medical imaging system. At present, the mainstream medical imaging data storage solution in the industry adopts a hierarchical storage method. There are differences in read and write performance and cost of storage devices at different levels. Generally speaking, the higher the read and write performance of the device, the higher the cost. For example, in a two-level storage device, the hot pool uses storage with higher read and write efficiency, such as solid state disk (SSD), cache, memory, etc., and the cold pool uses ordinary mechanical disk storage. Figure 1 This is a schematic diagram of the current system storage process, such as Figure 1 As shown, the system process includes: (1) the system stores newly created and reviewed image data in the hot pool for a period of time; (2) the scheduled task scans the image data in the hot pool that has been stored for a long time and has not been reviewed for a long time and marks it as pre-cooling data; (3) the scheduled task compresses the pre-cooling data and stores it in the cold pool; (4) when the data needs to be reviewed, it is retrieved from the hot pool for review; if the data to be reviewed does not exist in the hot pool, the data is pre-heated and pulled from the cold pool to the hot pool for review.

[0029] Image data requires a large storage space. When a patient undergoes an examination, such as an MRI, hundreds or even thousands of images will be generated on average. A patient may undergo repeated examinations, so the system needs to maintain a large amount of image data. In actual scenarios, after a patient's image data is reviewed by a doctor once, there is a high probability that it will not be reviewed again. The storage cost of the hot pool is much higher than that of the cold pool. Based on the above situation, the existing image system storage and retrieval module has the following problems: (1) The hot pool stores a large amount of useless data, consumes too much hot storage resources, and greatly increases storage costs. (2) Scheduled tasks need to scan a large amount of hot pool data every day, consuming a lot of computing resources.

[0030] In order to solve the problem that the current thermal pool stores a large amount of useless data, thereby consuming too many thermal storage resources, the embodiments of the present application provide a data storage method and system, a device and a storage medium, the method comprising: a review service module sends first image data to a prediction module; the prediction module predicts the first image data according to a target prediction model to obtain a first review prediction value corresponding to the first image data; wherein the target prediction model is obtained by model training through a training data set; the prediction module determines the storage location corresponding to the first image data according to the first review prediction value, so that the review service module stores the first image data to the storage location. It can be seen that after the review service module sends the first image data to the prediction module, the prediction module can determine the review prediction value corresponding to the image data according to the target prediction model, and then determine the storage location corresponding to the image data according to the review prediction value corresponding to the image data. It is precisely because the prediction module in the present application can determine the storage location corresponding to the image data, thereby avoiding storing a large amount of useless image data in the hot pool. Instead, the storage location of the image data is determined before the image data storage operation is performed, so that the review service module stores the image data according to the storage location corresponding to the image data, making the image data storage more accurate and reasonable, thereby reducing the consumption of excessive thermal storage resources, and thus improving the storage performance of the system.

[0031] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0032] Embodiment 1

[0033] The embodiment of the present application provides a data storage method, which is applied to a data storage system, wherein the data storage system includes a prediction module and a retrieval service module. Figure 2 Schematic diagram of the data storage method proposed in the embodiment of the present application Figure 1 ,like Figure 2 As shown, the data storage system may include the following steps:

[0034] Step 101: The access service module sends first image data to the prediction module.

[0035] It should be noted that, in the embodiment of the present application, the access service module may send the first image data to the prediction module.

[0036] It should be noted that, in the embodiments of the present application, Figure 3 Schematic diagram of the data storage system structure proposed in the embodiment of the present application Figure 1 ,like Figure 3 As shown, the data storage system includes a review service module and a prediction module.

[0037] It should be noted that, in the embodiment of the present application, the prediction module can predict the image data sent by the access service module to facilitate the subsequent determination of the storage location corresponding to the image data.

[0038] It should be noted that, in the embodiments of the present application, the first image data may include an image identity, an image review record, and an inspection result record. The present application does not specifically limit the content included in the first image data.

[0039] Step 102: The prediction module predicts the first image data according to the target prediction model to obtain a first reference prediction value corresponding to the first image data; wherein the target prediction model is obtained by model training through a training data set.

[0040] It should be noted that, in the embodiment of the present application, after the review service module sends the first image data to the prediction module, the prediction module can predict the first image data according to the target prediction model to obtain a first review prediction value corresponding to the first image data.

[0041] It should be noted that, in the embodiments of the present application, the first reference prediction value can be used to represent the probability prediction value of the first image data being referenced, and the present application does not specifically limit the size of the probability prediction value.

[0042] It should be noted that in an embodiment of the present application, the prediction module can obtain a training data set before predicting the first image data according to the target prediction model and obtaining the first reference prediction value corresponding to the first image data; then the initial prediction model can be trained based on the training data set, so that the target prediction model can be determined.

[0043] That is to say, in an embodiment of the present application, the prediction module may first train the initial prediction model to obtain a target prediction model, and then use the target prediction model to predict the first image data, thereby obtaining a first reference prediction value corresponding to the first image data.

[0044] It should be noted that in an embodiment of the present application, when the prediction module obtains a training data set, it can first obtain historical image data and the review data corresponding to the historical image data; then it can perform feature extraction on the historical image data to obtain a feature vector corresponding to the historical image data, and then it can construct a training data set based on the feature vector and the review data.

[0045] It should be noted that in the embodiments of the present application, historical image data may include image identity identification, image retrieval records, examination result records, diagnosis records, and diagnosis conclusions. The present application does not specifically limit the content of the historical image data.

[0046] It should be noted that, in the embodiments of the present application, the retrieved data corresponding to the historical image data may be the actual value of the retrieved historical image data.

[0047] Exemplarily, in an embodiment of the present application, the prediction module performs feature extraction on the historical image data, and can obtain the feature vector x corresponding to the historical image data, and the reference data y corresponding to the historical image data. x and y can be constituted into a training sample, and the set of x and the set of y can be constituted into a training data set, wherein, if the historical image data is referenced within T seconds after it is generated, y=1 can be set; if the historical image data is not referenced within T seconds after it is generated, y=0 can be set, and T is an integer greater than 0.

[0048] It should be noted that, in the embodiments of the present application, after acquiring the training data set, the prediction module can train the initial prediction model based on the training data set, so as to determine the target prediction model.

[0049] Furthermore, in an embodiment of the present application, when the prediction module trains the initial prediction model based on the training data set and determines the target prediction model, the feature vector can be first input into the initial prediction model and the prediction review result can be output; then, the first fitness value can be determined based on the prediction review result, the review data and the loss function; and then, the initial prediction model can be corrected based on the first fitness value to determine the target prediction model.

[0050] It should be noted that, in the embodiment of the present application, the prediction module can input the feature vector x into the initial prediction model, so as to output the prediction result. The initial prediction model can be expressed as the following formula.

[0051]

[0052] Among them, S l represents the number of neurons in the lth layer, represents the weighted sum of the i-th unit in the l-th layer, Represents the model parameters of the neural network (initial prediction model) at the lth layer of the vocabulary i and the vocabulary j in the historical image data, represents the model parameters of the neural network (initial prediction model) at the lth layer of the word i in the historical image data, Represents the number of times word j appears in the historical image data.

[0053] Further, in an embodiment of the present application, after obtaining the prediction review result, the prediction module can determine the first fitness value according to the prediction review result, the review data and the loss function, wherein the loss function can be expressed as the following formula.

[0054]

[0055] Among them, y i represents the actual value (reviewed data) corresponding to word i, represents the predicted reading result corresponding to word i, S l Represents the number of neurons in the lth layer.

[0056] It should be noted that, in the embodiment of the present application, the first fit value is used to represent the fit between the actual value of the query corresponding to the image data and the predicted query result corresponding to the image data.

[0057] Further, in an embodiment of the present application, after determining the first fitness value, the prediction module may correct the initial prediction model based on the first fitness value, and further determine the target prediction model.

[0058] It should be noted that in an embodiment of the present application, when the prediction module corrects the initial prediction model based on the first fitness value and determines the target prediction model, the bias derivative of the model parameter w can be calculated according to the first fitness value to obtain the corrected value of the model parameter w. The calculation formula is as follows.

[0059]

[0060] Among them, J(y,x) represents the first fit value, represents the weighted sum of the i-th unit in the l-th layer.

[0061] It should be noted that, in an embodiment of the present application, the prediction module can also calculate the bias derivative of the model parameter b according to the first fitness value to obtain a corrected value of the model parameter b, and the calculation formula is as follows.

[0062]

[0063] Furthermore, in an embodiment of the present application, after obtaining the corrected value of the model parameter w and the corrected value of the model parameter b, the prediction module can substitute the corrected value of the parameter w and the corrected value of the parameter b into the above formula (2) for repeated training correction. After each training, a reference with a window size of k can be taken for the obtained fitness value, and then the mean square error can be calculated using the value of the window size k at the current moment and the value of the window size k at the previous moment. If the size of the calculated value obtained is less than the preset value, the training is stopped to obtain the final values ​​of the model parameter w and the model parameter b. The mean square error calculation formula is as follows.

[0064]

[0065] Among them, J t+k Indicates the value of the window size k at the current moment, J t+k-1 The window size at the previous moment is the value of k, and β represents the preset value.

[0066] It should be noted that, in the embodiment of the present application, the empirical value of the preset value β is 0.1, and it can also be any other decimal between 0 and 1. The present application does not specifically limit the size of the preset value β.

[0067] It should be noted that in the embodiment of the present application, after the prediction module obtains the final values ​​of the model parameters w and the model parameters b, that is, after determining the model parameters of the initial prediction model, it can obtain the target prediction model, and the target prediction model can be expressed by the following formula.

[0068]

[0069] Among them, f(x ij ) is the xij represents the total number of times word j in the historical image data appears in the context of word i in the historical image data, c i , c j are the word vectors corresponding to word i and word j in the historical image data, respectively, and d i , d j represents the bias term of the i-th word in the historical image data and the j-th word in the historical image data, respectively. f(c) represents the minimum value of the target prediction model. V is all the words in the historical image data in the vocabulary. x is the word co-occurrence matrix in the historical image data. If the word i in the historical image data and the word j in the historical image data have never appeared at the same time, then f(x ij )=0.

[0070] It should be noted that, in the embodiments of the present application, the vector expressions of vocabulary i and vocabulary j can be expressed by the following formula.

[0071]

[0072] Among them, c i represents the word vector corresponding to word i in the historical image data, x i represents the number of times word i appears in the historical image data, x j Represents the number of times word j appears in the historical image data.

[0073] It should be noted that, in the embodiments of the present application, the prediction model can use an activation function to increase the nonlinearity of the entire model and improve the model complexity so that the model can better fit the target. The activation function is shown in the following formula.

[0074]

[0075] That is to say, in an embodiment of the present application, the prediction module can train an initial prediction model based on a training data set to determine a target prediction model, and then the image data input into the prediction module can be predicted according to the obtained target prediction model to obtain a first review prediction value corresponding to the image data, that is, a probability prediction value of the image data being reviewed can be obtained.

[0076] It should be noted that, in the embodiments of the present application, the above-mentioned training data set and target prediction model can be stored in the prediction module, and called and maintained by the prediction module.

[0077] Step 103: The prediction module determines a storage location corresponding to the first image data according to the first reference prediction value, so that the reference service module stores the first image data in the storage location.

[0078] It should be noted that in an embodiment of the present application, after the prediction module predicts the first image data according to the target prediction model and obtains the first review prediction value corresponding to the first image data, the prediction module can determine the storage location corresponding to the first image data according to the first review prediction value, so that the review service module stores the first image data to the storage location.

[0079] It should be noted that, in an embodiment of the present application, when the prediction module determines the storage location corresponding to the first image data according to the first review prediction value, if the first review prediction value is greater than or equal to the preset review threshold, it can be determined that the storage location corresponding to the first image data is a hot pool; if the first review prediction value is less than the preset review threshold, it can be determined that the storage location corresponding to the first image data is a cold pool. The present application does not specifically limit the size of the preset review threshold.

[0080] It should be noted that in the embodiments of the present application, the hot pool includes cache, memory, and solid-state disk (SSD), and the cold pool includes a mechanical disk. The present application does not specifically limit the cache address categories of the hot pool and the cold pool.

[0081] Further, in an embodiment of the present application, after the prediction module determines the storage location corresponding to the first image data according to the first reference prediction value, the storage location may be sent to the reference service module; the reference service module may store the first image data according to the storage location.

[0082] It should be noted that, in the embodiments of the present application, Figure 4 Schematic diagram of the data storage system structure proposed in the embodiment of the present application Figure 2 ,like Figure 4 As shown, the data storage system includes not only a query service module and a prediction module, but also a hot pool and a cold pool.

[0083] That is to say, in an embodiment of the present application, the prediction module can first determine the storage location corresponding to the first image data according to the first review prediction value, that is, pre-determine whether the first image data should be stored in the cold pool or the hot pool, and then send the storage location corresponding to the first image data to the review service module, and the review service module can then store the first image data according to the storage location, thereby reasonably and specifically storing the first image data, avoiding storing a large amount of data in the hot pool and wasting storage resources.

[0084] It should be noted that, in an embodiment of the present application, the data storage system can scan and process the second image data stored in the hot pool according to a preset scanning cycle to obtain third image data; wherein the third image data is part or all of the image data in the second image data; the third image data is stored in the cold pool, wherein the scanning cycle can be an integer greater than 0, and the present application does not specifically limit the value of the scanning cycle.

[0085] Exemplarily, in an embodiment of the present application, assuming that the preset scanning cycle is 24 hours, the data storage system scans the second image data stored in the hot pool according to a 24-hour cycle, and finds that some data in the second image data has not been viewed within 24 hours, then this part of the data can be stored in the cold pool as third data.

[0086] It should be noted that in the embodiments of the present application, compared with the current storage process, a large amount of useless image data is stored in the hot pool, and the scheduled task needs to scan a large amount of hot pool data according to a preset scanning cycle every day, which requires a large amount of computing resources. The review service module in the present application stores the image data according to the storage location corresponding to the image data. There is only a small amount of image data in the hot pool. When performing a scanning task, the storage pressure of the hot pool is reduced, thereby reducing the computing pressure of the scheduled task scanning, improving the computing efficiency, and reducing the computing cost.

[0087] It should be noted that, in the embodiments of the present application, the prediction module may update the training data set according to the third image data to obtain an updated training data set; and update the target prediction model based on the updated training data set.

[0088] Exemplarily, in an embodiment of the present application, the prediction module can update the training data set according to the third image data, that is, can update the reference data in the training data set, and then can update the target prediction model based on the updated training data set.

[0089] To summarize, before the review service module stores the image data, the image data can be sent to the prediction module first. Then the prediction module can predict the image data according to the trained target prediction model to obtain the review prediction probability value corresponding to the image data. According to the review prediction probability value, the storage location (cold pool or hot pool) corresponding to the image data can be further obtained. The prediction module then sends the storage location corresponding to the image data to the review service module, so that the review service module stores the image data according to the storage location. That is, the present application can make a predictive judgment on the image data before storing the image data, so that the image data storage is more accurate and reasonable, thereby reducing the consumption of excessive thermal storage resources and reducing the cost of image storage.

[0090] An embodiment of the present application provides a data storage method, which is applied to a data storage system, wherein the data storage system includes a prediction module and a review service module, and the method includes: the review service module sends first image data to the prediction module; the prediction module predicts the first image data according to a target prediction model to obtain a first review prediction value corresponding to the first image data; wherein the target prediction model is obtained by model training through a training data set; the prediction module determines a storage location corresponding to the first image data according to the first review prediction value, so that the review service module stores the first image data to the storage location. It can be seen that after the review service module sends the first image data to the prediction module, the prediction module can determine the review prediction value corresponding to the image data according to the target prediction model, and then determine the storage location corresponding to the image data according to the review prediction value corresponding to the image data. Since the prediction module in the present application can use the target prediction model to accurately predict the review value corresponding to the image data, it avoids storing a large amount of useless image data in the hot pool. Instead, before performing the image data storage operation, the storage location of the image data is determined so that the review service module stores the image data according to the storage location corresponding to the image data, making the image data storage more accurate and reasonable, thereby reducing the consumption of excessive thermal storage resources and improving the storage performance of the system.

[0091] Embodiment 2

[0092] Based on the above embodiments, another embodiment of the present application provides a data storage method. The present application introduces an image storage prediction module (prediction module). The improved storage system introduces an image storage prediction module (prediction module), which can be used as a control unit for hierarchical storage of medical image data, and interacts with the medical image storage retrieval service module (retrieval service module), hot pool, cold pool and other modules for data, algorithms, etc. The storage mode can be optimized, greatly reducing the storage pressure of the hot pool and the computing pressure of the scheduled tasks. At the same time, a set of intelligent data storage mechanisms can be further formed to make the operation of the entire medical imaging system more efficient; and medical imaging resources can be further calculated and utilized, and an artificial intelligence (AI) model of medical imaging data and text records can be established, which can form a basis for further realizing applications such as AI-assisted diagnosis. The data storage method proposed in this application is further explained below.

[0093] It should be noted that, in the embodiments of the present application, Figure 5 Schematic diagram of the data storage system structure proposed in the embodiment of the present application Figure 3 ,like Figure 5As shown, based on the hierarchical storage method adopted in the traditional medical image data storage solution, an image storage prediction module (prediction module) is introduced. This module can construct a data set based on the text features of medical image data for storage, and iteratively train an AI model (target prediction model) based on this data set to determine whether to store medical image data in a hot pool / cold pool, thereby forming a new medical image data storage mechanism.

[0094] It should be noted that in the embodiment of the present application, the data storage process is as follows: the image storage prediction module (prediction module) can use the existing medical data to construct (historical image data) a medical text record data set (training data set) related to the image data, and then use the above data set for machine training and learning. The data set can be constructed in the following way: the image review record (review data) or its form is converted into a label value, and other data or their form is converted into feature values. On this basis, a prediction model f (target prediction model) is output, and the model (target prediction model) is used to predict the probability of medical image data being reviewed.

[0095] It should be noted that, in the embodiments of the present application, the retrieved data may be the actual value of the retrieved historical image data.

[0096] Exemplarily, in an embodiment of the present application, the prediction module performs feature extraction on the historical image data, and can obtain the feature vector x corresponding to the historical image data, and the reference data y corresponding to the historical image data. x and y can be constituted into a training sample, and the set of x and the set of y can be constituted into a training data set, wherein, if the historical image data is referenced within T seconds after it is generated, y=1 can be set; if the historical image data is not referenced within T seconds after it is generated, y=0 can be set, and T is an integer greater than 0.

[0097] Furthermore, in an embodiment of the present application, before the image data is stored, the storage system submits a medical text record (first image data) containing at least the identity identifier of the image data to a prediction module for subsequent determination of the storage method (storage location) of the image.

[0098] It should be noted that, in the embodiments of the present application, the image data may include the identity identification of the image data, the image review record, and the inspection result record. The present application does not specifically limit the content included in the image data.

[0099] Further, in an embodiment of the present application, the image storage prediction module (prediction module) runs the prediction model f (target prediction model) according to the input image data medical text record (image data), calculates the probability of the image being accessed in the future (first access prediction value), and determines the probability (first access prediction value) to be determined as a storage mode of "high access probability-storage in hot pool" or "low access probability-storage in cold pool".

[0100] It should be noted that, in an embodiment of the present application, when the prediction module determines the storage location corresponding to the image data according to the first review prediction value, if the first review prediction value is greater than or equal to the preset review threshold, it can be determined that the storage location corresponding to the image data is a hot pool; if the first review prediction value is less than the preset review threshold, it can be determined that the storage location corresponding to the image data is a cold pool. The present application does not specifically limit the size of the preset review threshold.

[0101] It should be noted that in the embodiments of the present application, the hot pool includes cache, memory, and solid-state disk (SSD), and the cold pool includes a mechanical disk. The present application does not specifically limit the cache address categories of the hot pool and the cold pool.

[0102] Furthermore, in an embodiment of the present application, the image storage prediction module (prediction module) reports the determined medical image data storage method (storage location) to the medical image storage retrieval service module (retrieval service module), and the medical image storage retrieval service module (retrieval service module) stores the medical image according to the reported storage method (storage location).

[0103] That is to say, in an embodiment of the present application, the prediction module can first determine the storage location corresponding to the medical imaging data based on the first review prediction value, that is, pre-determine whether the medical imaging data should be stored in the cold pool or the hot pool, and then send the storage location corresponding to the medical imaging data to the review service module. The review service module can then store the medical imaging data according to the storage location, thereby storing the medical imaging data reasonably and in a targeted manner, avoiding storing a large amount of data in the hot pool and wasting storage resources.

[0104] It should be noted that, in the embodiment of the present application, the image data that has been stored for a long time and has not been accessed for a long time in the scheduled task scanning thermal pool is marked as pre-cooling data.

[0105] Exemplarily, in an embodiment of the present application, assuming that the preset scanning cycle is 24 hours, the data storage system scans the second image data stored in the hot pool according to a 24-hour cycle, and finds that some data in the second image data has not been viewed within 24 hours, then this part of the data can be stored in the cold pool as third data.

[0106] Furthermore, in an embodiment of the present application, the scheduled task can compress and store the pre-cooling data in the cold pool, and the prediction module can synchronously modify the label value (review record data) in the medical imaging data set in the image storage prediction module (prediction module) for the medical imaging data judged as the cold pool in the scheduled task, so as to update the data set (training data set) and the algorithm model (target prediction model).

[0107] It should be noted that in the embodiments of the present application, compared with the current storage process, a large amount of useless image data is stored in the hot pool, and the scheduled task needs to scan a large amount of hot pool data according to a preset scanning cycle every day, which requires a large amount of computing resources. The review service module in the present application stores the image data according to the storage location corresponding to the image data. There is only a small amount of image data in the hot pool. When performing a scanning task, the storage pressure of the hot pool is reduced, thereby reducing the computing pressure of the scheduled task scanning, improving the computing efficiency, and reducing the computing cost.

[0108] Furthermore, in the embodiments of the present application, Figure 6 The following is a flow chart of the target prediction model construction process proposed in the embodiment of the present application. Figure 6 As shown in the figure, the construction process of the target prediction model is as follows:

[0109] Step S101: feature data extraction.

[0110] It should be noted that, in the embodiment of the present application, the prediction module can convert the relevant medical text record data (historical image data) corresponding to the stock images into a feature vector x to complete the extraction of feature data.

[0111] Step S102: construct a training data set.

[0112] It should be noted that, in the embodiments of the present application, the prediction module can form a training sample with x and the review record (review data) y, and the set of x and the set of y constitute a training sample set (training data set). If the historical image data is reviewed within T seconds after it is generated, y=1 can be set; if the historical image data is not reviewed within T seconds after it is generated, y=0 can be set, and T is an integer greater than 0.

[0113] Step S103: Select a training model algorithm.

[0114] It should be noted that, in the embodiment of the present application, the prediction module can select or construct a hypothesis space F of a model θ (x), θ is the model parameter to be determined, and the set of w and b of each layer of the neural network in the above formula (1) is (w, b) represents the hypothesis space about the set x, that is, F θ (x).

[0115] Step S104: input data for training.

[0116] It should be noted that, in an embodiment of the present application, the prediction module can input the training sample set (training data set) into the loss function L(Y, F(X)) for training. The loss function is as shown in the above formula (2), and the model parameter θ after training is fixed.

[0117] Step S105: output the target prediction model.

[0118] It should be noted that in an embodiment of the present application, after the prediction module inputs the training sample set (training data set) into the loss function L(Y, F(X)) for training, a fixed model parameter θ is obtained, and a prediction model (target prediction model) can be obtained based on the fixed model parameter θ.

[0119] It should be noted that, in the embodiments of the present application, Figure 7 Schematic diagram of the data storage method proposed in the embodiment of the present application Figure 2 ,like Figure 7 As shown, a medical imaging data set (training data set) is constructed based on medical imaging data and its text records. The medical imaging data set (training data set) is used to train an artificial intelligence prediction model (target prediction model) that can identify the probability of medical imaging data being accessed in the future. The above-mentioned medical imaging data set (training data set) and artificial intelligence model (target prediction model) are stored in the imaging storage prediction module (prediction module), and are called and maintained by the imaging storage prediction module (prediction module). Among them, the medical text records used to construct the medical imaging data set include but are not limited to: image identity, image access records, examination result records, diagnosis records, diagnosis conclusions and other data.

[0120] Further, in the embodiments of the present application, Figure 7 As shown, after the newly created or reviewed medical image data (first image data) is reviewed and used, before being stored, the medical image storage review service module reports (review service module) text information containing at least the medical image data identity identifier to the image storage prediction module (prediction module) for determining the storage method (storage location) of the medical image data (first image data).

[0121] Further, in an embodiment of the present application, the medical image prediction module (prediction module) runs a prediction model based on the image data text record input by the medical image storage and retrieval service module (retrieval service module), calculates the probability (P) of the image being retrieved within a future period of time (T), and determines whether the probability (first retrieval prediction value) should be judged as "high retrieval probability-stored in a hot pool" or "low retrieval probability-stored in a cold pool", wherein the method for judging whether the probability (first retrieval prediction value) is high or low should at least include a threshold judgment method, that is, when P>threshold, it is judged as high, otherwise it is low.

[0122] It should be noted that, in an embodiment of the present application, when the prediction module determines the storage location corresponding to the first image data according to the first review prediction value, if the first review prediction value is greater than or equal to the preset review threshold, it can be determined that the storage location corresponding to the first image data is a hot pool; if the first review prediction value is less than the preset review threshold, it can be determined that the storage location corresponding to the first image data is a cold pool. The present application does not specifically limit the size of the preset review threshold.

[0123] Furthermore, in an embodiment of the present application, the image storage prediction module (prediction module) reports the determined storage mode of the medical image data (first image data) to the medical image storage and retrieval service module (retrieval service module); when the reported judgment result is "high retrieval probability-stored in the hot pool", the medical image storage and retrieval service module (retrieval service module) stores the medical image directly in the hot pool; when the reported judgment result is "low retrieval probability-stored in the cold pool", the medical image storage and retrieval service module (retrieval service module) compresses the medical image and stores it in the cold pool.

[0124] It should be noted that in an embodiment of the present application, when the system starts a scheduled task, the medical imaging data (second imaging data) stored in the thermal pool is scanned, and the imaging data that has been stored in the thermal pool for a long time and has not been accessed for a long time is marked as pre-cooling data.

[0125] Furthermore, in an embodiment of the present application, the text information corresponding to the medical image data (third image data) judged as cold pool storage in the scheduled task is synchronously updated to the image storage prediction module (prediction module), so as to modify the label value (review data) corresponding to the medical image data set stored in the image storage prediction module (prediction module) to realize the update and continuous training of the data set (training data set) and the algorithm model (target prediction model).

[0126] Exemplarily, in an embodiment of the present application, the prediction module can update the training data set according to the third image data, that is, can update the reference data in the training data set, and then can update the target prediction model based on the updated training data set.

[0127] To summarize, before the review service module stores the image data, the image data can be sent to the prediction module first. Then the prediction module can predict the image data according to the trained target prediction model to obtain the review prediction probability value corresponding to the image data. According to the review prediction probability value, the storage location (cold pool or hot pool) corresponding to the image data can be further obtained. The prediction module then sends the storage location corresponding to the image data to the review service module, so that the review service module stores the image data according to the storage location. That is, the present application can make a predictive judgment on the image data before storing the image data, so that the image data storage is more accurate and reasonable, thereby reducing the consumption of excessive thermal storage resources and reducing the cost of image storage.

[0128] The embodiment of the present application provides a data storage method, which is applied to a data storage system, the data storage system includes a prediction module and a review service module, the method includes: the review service module sends the first image data to the prediction module; the prediction module predicts the first image data according to the target prediction model to obtain the first review prediction value corresponding to the first image data; wherein the target prediction model is obtained by model training through the training data set; the prediction module determines the storage location corresponding to the first image data according to the first review prediction value, so that the review service module stores the first image data in the storage location. It can be seen that after the review service module sends the first image data to the prediction module, the prediction module can determine the review prediction value corresponding to the image data according to the target prediction model, and then can determine the storage location corresponding to the image data according to the review prediction value corresponding to the image data. It is precisely because the prediction module in the present application can determine the storage location corresponding to the image data, thereby avoiding storing a large amount of useless image data in the hot pool, but before performing the storage operation of the image data, the storage location of the image data is determined, so that the review service module stores according to the storage location corresponding to the image data, making the image data storage more accurate and reasonable, thereby reducing the consumption of excessive thermal storage resources, and thus improving the storage performance of the system.

[0129] Embodiment 3

[0130] Based on the above embodiments, the present application provides a data storage system. Figure 8 Schematic diagram of the structure of the data storage system Figure 1 ,like Figure 8 As shown, the data storage system 10 includes: a prediction module 11 and a query service module 12.

[0131] The access service module 12 is used to send the first image data to the prediction module;

[0132] The prediction module 11 is used to predict the first image data according to the target prediction model to obtain a first reference prediction value corresponding to the first image data; wherein the target prediction model is obtained by model training through the training data set;

[0133] The prediction module 11 is further configured to determine a storage location corresponding to the first image data according to the first reference prediction value, so that the reference service module stores the first image data in the storage location.

[0134] In the embodiments of the present application, further, Fig. 9 Schematic diagram of the structure of the data storage system Figure 2 ,like Fig. 9 As shown, the data storage system 10 proposed in the embodiment of the present application may also include a processor 13, a memory 14 storing executable instructions of the processor 13, and further, the data storage system 10 may also include a communication interface 15, and a bus 16 for connecting the processor 13, the memory 14 and the communication interface 15.

[0135] In the embodiment of the present application, the processor 13 may be at least one of an Application Specific Integrated Circuit (ASIC), a Digital Signal Processor (DSP), a Digital Signal Processing Device (DSPD), a Programmable Logic Device (PLD), a Field Programmable Gate Array (FPGA), a Central Processing Unit (CPU), a controller, a microcontroller, and a microprocessor. It is understandable that for different devices, the electronic device used to implement the function of the processor may also be other, and the embodiment of the present application is not specifically limited. The data storage system 10 may also include a memory 14, which may be connected to the processor 13, wherein the memory 14 is used to store executable program code, the program code includes computer operation instructions, and the memory 14 may include a high-speed RAM memory, and may also include a non-volatile memory, for example, at least two disk memories.

[0136] In the embodiment of the present application, the bus 16 is used to connect the communication interface 15, the processor 13 and the memory 14, and the mutual communication between these devices.

[0137] In the embodiment of the present application, the memory 14 is used to store instructions and data.

[0138] Further, in an embodiment of the present application, the processor 13 is used for the access service module to send the first image data to the prediction module;

[0139] The prediction module predicts the first image data according to the target prediction model to obtain a first reference prediction value corresponding to the first image data; wherein the target prediction model is obtained by model training through the training data set;

[0140] The prediction module determines a storage location corresponding to the first image data according to the first review prediction value, so that the review service module stores the first image data in the storage location.

[0141] In practical applications, the memory 14 may be a volatile memory, such as a random access memory (RAM); or a non-volatile memory, such as a read-only memory (ROM), a flash memory, a hard disk (HDD) or a solid-state drive (SSD); or a combination of the above types of memory, and provide instructions and data to the processor 13.

[0142] The embodiment of the present application provides a data storage system, the data storage system includes a prediction module and a review service module, the review service module sends the first image data to the prediction module; the prediction module predicts the first image data according to the target prediction model to obtain the first review prediction value corresponding to the first image data; wherein the target prediction model is obtained by model training through the training data set; the prediction module determines the storage location corresponding to the first image data according to the first review prediction value, so that the review service module stores the first image data in the storage location. It can be seen that after the review service module sends the first image data to the prediction module, the prediction module can determine the review prediction value corresponding to the image data according to the target prediction model, and then can determine the storage location corresponding to the image data according to the review prediction value corresponding to the image data. Since the prediction module in the present application can accurately predict the review value corresponding to the image data using the target prediction model, it avoids storing a large amount of useless image data in the hot pool, but before performing the storage operation of the image data, the storage location of the image data is determined, so that the review service module stores according to the storage location corresponding to the image data, so that the image data storage is more accurate and reasonable, thereby reducing the consumption of excessive thermal storage resources, thereby improving the storage performance of the system.

[0143] An embodiment of the present application provides a computer-readable storage medium having a program stored thereon, and when the program is executed by a processor, the data storage method described above is implemented.

[0144] Specifically, the program instructions corresponding to a data storage method in this embodiment can be stored in a storage medium such as a CD, a hard disk, a USB flash drive, etc. When the program instructions corresponding to a data storage method in the storage medium are read or executed by an electronic device, the following steps are included:

[0145] The review service module sends the first image data to the prediction module;

[0146] The prediction module predicts the first image data according to the target prediction model to obtain a first reference prediction value corresponding to the first image data; wherein the target prediction model is obtained by model training through the training data set;

[0147] The prediction module determines a storage location corresponding to the first image data according to the first review prediction value, so that the review service module stores the first image data in the storage location.

[0148] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of hardware embodiments, software embodiments, or embodiments in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) that contain computer-usable program code.

[0149] The present application is described with reference to implementation flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the process in the flowchart. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0150] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which is implemented in the implementation flow diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0151] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for executing the steps in the flowchart. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0152] The above description is only a preferred embodiment of the present application and is not intended to limit the protection scope of the present application.

Claims

1. A data storage method, characterized in that: The method is applied to a data storage system, the data storage system includes a prediction module and a retrieval service module, and the method includes: The review service module sends the first image data to the prediction module; The prediction module predicts the first image data according to the target prediction model to obtain a first reference prediction value corresponding to the first image data; wherein the target prediction model is obtained by model training through a training data set; The prediction module determines a storage location corresponding to the first image data according to the first review prediction value, so that the review service module stores the first image data in the storage location.

2. The method according to claim 1, characterized in that Before the prediction module predicts the first image data according to the target prediction model and obtains the first reference prediction value corresponding to the first image data, the method further includes: The prediction module obtains the training data set; The prediction module trains an initial prediction model based on the training data set to determine the target prediction model.

3. The method according to claim 2, characterized in that The prediction module obtains the training data set, including: The prediction module obtains historical image data and the reference data corresponding to the historical image data; The prediction module extracts features from the historical image data to obtain feature vectors corresponding to the historical image data; The prediction module constructs the training data set according to the feature vector and the retrieved data.

4. The method according to claim 3, characterized in that The prediction module trains the initial prediction model based on the training data set to determine the target prediction model, including: The prediction module inputs the feature vector into the initial prediction model and outputs the prediction result; The prediction module determines a first fitness value according to the prediction review result, the review data and the loss function; The prediction module modifies the initial prediction model based on the first fitness value to determine the target prediction model.

5. The method according to claim 1, characterized in that The prediction module determines the storage location corresponding to the first image data according to the first reference prediction value, including: In the case where the first reference prediction value is greater than or equal to a preset reference threshold, the prediction module determines that the storage location corresponding to the first image data is a hot pool; When the first reference prediction value is less than a preset reference threshold, the prediction module determines that the storage location corresponding to the first image data is a cold pool.

6. The method according to claim 5, characterized in that The hot pool includes a cache and a memory, and the cold pool includes a mechanical disk.

7. The method according to claim 6, characterized in that The method further comprises: Scanning and processing the second image data stored in the heat pool according to a preset scanning cycle to obtain third image data; wherein the third image data is part or all of the image data in the second image data; The review service module stores the third image data in the cold pool.

8. The method according to claim 7, characterized in that The method further comprises: The prediction module updates the training data set according to the third image data to obtain an updated training data set; The prediction module updates the target prediction model based on the updated training data set.

9. The method according to claim 1, characterized in that: The method further comprises: The prediction module sends the storage location to the access service module; The access service module stores the first image data according to the storage location.

10. A data storage system, characterized in that: The data storage system includes: a prediction module and a retrieval service module. The reading service module is used to send the first image data to the prediction module; The prediction module is used to predict the first image data according to the target prediction model to obtain a first reference prediction value corresponding to the first image data; wherein the target prediction model is obtained by model training through the training data set; The prediction module is further configured to determine a storage location corresponding to the first image data according to the first reference prediction value, so that the reference service module stores the first image data in the storage location.

11. A data storage system, characterized in that: The data storage system comprises: a processor and a memory; wherein, The memory is used to store a computer program that can be run on the processor; The processor is configured to execute the method according to any one of claims 1 to 9 when running the computer program.

12. A computer-readable storage medium, characterized in that: The storage medium stores computer program codes, and when the computer program codes are executed by a computer, the method according to any one of claims 1 to 9 is executed.

Images