A method and system for predicting spinal health changes based on big data

Abstract

The present invention provides a method and system for predicting spinal health changes based on big data; wherein, the method includes: obtaining spinal image big data of different health stages, forming a training set based on the spinal image big data, and using the training set to train a deep recognition model; obtaining spinal image data of the current health stage, inputting the spinal image data into the trained deep recognition model, and the deep recognition model outputs a spinal health prediction atlas; and outputting the spinal health prediction atlas according to a preset strategy. The scheme of the present invention can accurately predict the spinal health prediction atlas corresponding to the current health stage of the spine, so that doctors and patients can have a deeper understanding of the true situation of spinal health, which is conducive to the determination of a more reasonable diagnosis and treatment plan.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to a method and system for predicting spinal health changes based on big data. Background Art

[0002] With the pace of life and work in modern society, more and more people are experiencing spinal health problems, such as C-type scoliosis, S-type scoliosis, etc. In order to better assist doctors in the diagnosis and treatment of spinal health problems, relevant technologies have begun to study the automatic recognition and annotation of spinal images, such as:

[0003] Patent document 1 (CN114187320A) discloses a method for segmenting a spinal CT image, comprising: determining an edge between an estimated foreground area and an estimated background area on an enhanced spinal CT image, and drawing a directed graph based on each voxel in the edge; determining a minimum cut of the enhanced spinal CT image based on the directed graph to segment the enhanced spinal CT image into an actual foreground area including the spinal imaging and an actual background area including non-spinal images; based on the minimum cut, determining at least the actual foreground area on the enhanced spinal CT image.

[0004] Patent document 2 (CN114170114A) discloses a method for enhancing a spinal CT image, which comprises: performing morphological dilation processing on the spinal CT image to generate a morphological dilation image; performing morphological erosion processing on the morphological dilation image to generate a morphological closure processing image; and generating an enhanced spinal CT image based on the morphological closure processing image.

[0005] Patent document 3 (CN114081471A) discloses a method for measuring the Cobb angle of scoliosis based on three-dimensional images and multi-layer perception, and the method is as follows: S1, using a depth camera to obtain the original three-dimensional image of the back of the human body to be measured; S2, performing normalized interpolation processing on the original three-dimensional image obtained in step S1, sampling the back image of the human body to be measured at equal intervals and projecting it on a three-dimensional coronal plane coordinate system, and obtaining a three-dimensional image with a regular and fixed number of sampling points; S3, using several planes to cross-section the three-dimensional image obtained in step S2, and obtaining a series of back cross-sectional profiles, each of which corresponds to a sampling point sequence; S4, performing symmetry analysis on the sampling point sequence of each back cross-sectional profile obtained in step S3 and obtaining the midline of the spine on each back cross-sectional profile. candidate points, select the best candidate points as the annotated spinal points; S5, construct a multi-layer perceptron, the multi-layer perceptron performs spinal point annotating training for steps S3 and S4, and the trained multi-layer perceptron can accurately find spinal points in the sampling point sequence of each back cross-section contour; S6, input the three-dimensional image of step S2 through the multi-layer perceptron and output the spinal points on the three-dimensional image; S7, curve fit the spinal points located on the three-dimensional coronal plane coordinate system and obtain the spinal midline, calculate the normal vector of the spinal point located on the three-dimensional coronal plane coordinate system about the spinal midline; S8, use the normal vector obtained in step S7 to construct an n-dimensional real symmetric matrix, calculate the angle between the normal vectors respectively, take the two normal vectors that form the largest angle, and the maximum angle is the scoliosis cobb angle.

[0006] Based on the relevant patent literature in the aforementioned prior art, it can be seen that the prior art for spinal health issues mainly focuses on the detection and labeling of certain parameters (such as the Cobb angle) in the current spinal image, but cannot predict and output changes in spinal health, cannot well assist doctors in related diagnosis and treatment, and is not conducive to patients having an in-depth understanding of their own spinal health problems. Summary of the invention

[0007] In order to at least solve the technical problems existing in the above-mentioned background technology, the present invention provides a method, system, electronic device and computer storage medium for predicting spinal health changes based on big data.

[0008] The first aspect of the present invention provides a method for predicting spinal health changes based on big data, comprising the following steps:

[0009] Acquire big data of spinal images at different health stages, build a training set based on the big data of spinal images, and use the training set to train a deep recognition model;

[0010] Acquire spinal image data of the current health stage, input the spinal image data into the trained deep recognition model, and the deep recognition model outputs a spinal health prediction atlas;

[0011] The spinal health prediction atlas is output according to a preset strategy.

[0012] Further, the spinal image big data includes first spinal images at different health stages;

[0013] Then the step of forming a training set based on the spinal image big data includes:

[0014] Performing first spine parameter extraction processing on each of the first spine images, determining a first attribute of each of the first spine images based on the extracted first spine parameters, and labeling each of the first spine images using the first attribute;

[0015] Performing fitting processing on each of the first spinal images having the same first attribute to obtain a second spinal image and corresponding second spinal parameters, and determining a second attribute of the second spinal image according to the second spinal parameters;

[0016] Calculating a first matching degree between the second attribute of each second spinal column image and a standard attribute stage table, determining a first stage sequence number of each second spinal column image according to the first matching degree, and associating a plurality of first spinal column images corresponding to each second spinal column image with the first stage sequence number;

[0017] Starting from the first stage sequence number of each second spinal column image, traversing backward, and associating a number of the first spinal column images corresponding to the second spinal column image involved in the traversal with the second stage sequence number;

[0018] A plurality of training data are obtained according to the first spinal column image and / or the second spinal column image, and each of the training data is organized into the training set.

[0019] Further, the first spinal image is associated with a user attribute and is associated with a third stage sequence number;

[0020] Then, associating a plurality of the first spinal images corresponding to each of the second spinal images with the first stage sequence number includes:

[0021] Determine at least two of the first spinal images based on the third stage sequence number, and calculate the mutation degree according to the first spinal parameters of the at least two first spinal images;

[0022] If the mutation degree is greater than or equal to a first threshold, the corresponding first spinal column image is deleted, and the remaining first spinal column images are associated with the first stage sequence number.

[0023] Furthermore, after the mutation degree is greater than or equal to the first threshold, the method further includes:

[0024] Calculate whether the mutation degree is less than the second threshold, if so, then:

[0025] Determine a plurality of first mechanical models based on the first spinal parameters of the first spinal image that is adjacent to the first spinal image, and calculate a correlation value between the first spinal parameters of the first spinal image and the corresponding first mechanical models;

[0026] If the correlation value is greater than or equal to the third threshold, the mutation degree is determined to be greater than or equal to the first threshold, and the corresponding first spinal image is deleted; otherwise, the mutation degree is determined to be less than the first threshold, and the corresponding first spinal image is not deleted.

[0027] Furthermore, the third threshold can be determined as follows:

[0028] Determine a plurality of second mechanical models based on the first spinal parameters of the first spinal image that is second precedingly adjacent to the first spinal image, calculate a first similarity between the second mechanical models and the corresponding first mechanical models, and determine the third threshold based on the first similarity;

[0029] The third threshold is negatively correlated with the first similarity.

[0030] Furthermore, the step of inputting the spinal image data into the trained deep recognition model, and the deep recognition model outputting a spinal health prediction atlas, comprises:

[0031] Calculating a third spinal column parameter corresponding to the spinal column image data, inputting the third spinal column parameter into the trained deep recognition model, wherein the deep recognition model obtains an initial prediction atlas, wherein the initial prediction atlas includes a plurality of second spinal column images and corresponding fourth stage serial numbers;

[0032] calculating a second similarity between the third spine parameter and the first spine parameter corresponding to each of the first spine images associated with the second spine image;

[0033] If the second similarity is greater than or equal to a fourth threshold, the first spinal image associated with the second spinal image is used as a target prediction image, and the target prediction image corresponding to each second spinal image is associated with the fourth stage sequence number to obtain the spinal health prediction atlas;

[0034] The deep recognition model outputs the spinal health prediction atlas.

[0035] Furthermore, the outputting of the spinal health prediction atlas according to a preset strategy includes:

[0036] Calculating a second matching degree between a third attribute corresponding to the spinal image data of the current health stage and the standard attribute stage table, determining a fifth stage sequence number based on the second matching degree, and determining the number of images in the spinal health prediction atlas according to the fifth stage sequence number;

[0037] Determine a sequence number of a key stage according to the third attribute, and correct the number of images in the key stage according to the sequence number of the key stage;

[0038] The degree of correction is positively correlated with the criticality value of the critical stage.

[0039] The second aspect of the present invention provides a spinal health change prediction system based on big data, comprising a first acquisition module, a second acquisition module, a processing module, a storage module, and an output module, wherein the processing module is connected to the first acquisition module, the second acquisition module, the storage module, and the output module; wherein:

[0040] The first acquisition module is used to acquire big data of spinal images at different health stages and transmit them to the processing module;

[0041] The second acquisition module is used to acquire the spinal image data of the current health stage and transmit it to the processing module;

[0042] The storage module is used to store executable computer program code;

[0043] The output module is used to output the spine health prediction atlas according to a preset strategy;

[0044] The processing module is used to execute the method as described in any of the preceding items by calling the executable computer program code in the storage module.

[0045] A third aspect of the present invention provides an electronic device, comprising:

[0046] A memory storing executable program code;

[0047] a processor coupled to the memory;

[0048] The processor calls the executable program code stored in the memory to execute the method as described in any of the preceding items.

[0049] A fourth aspect of the present invention provides a computer storage medium having a computer program stored thereon, and the computer program, when executed by a processor, executes any of the methods described above.

[0050] The solution of the present invention obtains big data of spinal images at different health stages, forms a training set based on the big data of spinal images, and uses the training set to train a deep recognition model; obtains spinal image data at the current health stage, inputs the spinal image data into the trained deep recognition model, and the deep recognition model outputs a spinal health prediction atlas; and outputs the spinal health prediction atlas according to a preset strategy. The solution of the present invention can accurately predict the spinal health prediction atlas corresponding to the current health stage of the spine, so that doctors and patients can have a deeper understanding of the true condition of spinal health, which is conducive to the determination of a more reasonable diagnosis and treatment plan. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0052] Figure 1 It is a flowchart of a method for predicting spinal health changes based on big data disclosed in an embodiment of the present invention;

[0053] Figure 2 is a schematic diagram of obtaining a virtual and equivalent second spinal image based on fitting of a first spinal image according to an embodiment of the present invention;

[0054] Figure 3 It is a schematic diagram of the transformation relationship between the first stage sequence number and the second stage sequence number disclosed in the embodiment of the present invention;

[0055] Figure 4 It is a schematic diagram of an output method of the spine health prediction atlas disclosed in an embodiment of the present invention;

[0056] Figure 5 It is a structural schematic diagram of a spinal health change prediction system based on big data disclosed in an embodiment of the present invention;

[0057] Figure 6 It is a structural schematic diagram of an electronic device disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] The words "first, second, third, etc." or module A, module B, module C and the like in the specification and claims are used only to distinguish similar objects and do not represent a specific ordering of the objects. It is understandable that the specific order or sequence can be interchanged where permitted so that the embodiments of the invention described herein can be implemented in an order other than that illustrated or described herein.

[0060] In the following description, the numbers representing the steps, such as S110, S120, etc., do not necessarily mean that the steps must be executed in this manner. If permitted, the order of the previous and next steps can be interchanged, or they can be executed simultaneously.

[0061] The term "comprising" as used in the description and claims should not be interpreted as being limited to what is listed thereafter; it does not exclude other elements or steps. Therefore, it should be interpreted as specifying the presence of the features, integers, steps or components mentioned, but does not exclude the presence or addition of one or more other features, integers, steps or components and groups thereof. Therefore, the expression "a device comprising means A and B" should not be limited to a device consisting of components A and B only.

[0062] References to "one embodiment" or "an embodiment" in this specification mean that a particular feature, structure, or characteristic described in conjunction with the embodiment is included in at least one embodiment of the present invention. Therefore, the phrases "in one embodiment" or "in an embodiment" appearing in various places in this specification do not necessarily refer to the same embodiment, but may refer to the same embodiment. In addition, in one or more embodiments, the particular features, structures, or characteristics can be combined in any appropriate manner, as would be apparent to one of ordinary skill in the art from this disclosure.

[0063] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. In the event of any inconsistency, the meanings described in this specification or derived from the contents recorded in this specification shall prevail. In addition, the terms used herein are only for the purpose of describing the embodiments of the present invention and are not intended to limit the present invention.

[0064] Embodiment 1

[0065] See also Figure 1 , Figure 11 is a flow chart of a method for predicting spinal health changes based on big data disclosed in an embodiment of the present invention. Figure 1 As shown, a method for predicting spinal health changes based on big data in an embodiment of the present invention includes the following steps:

[0066] Acquire big data of spinal images at different health stages, build a training set based on the big data of spinal images, and use the training set to train a deep recognition model;

[0067] Acquire spinal image data of the current health stage, input the spinal image data into the trained deep recognition model, and the deep recognition model outputs a spinal health prediction atlas;

[0068] The spinal health prediction atlas is output according to a preset strategy.

[0069] In this step, as described in the background technology, in actual operation, both doctors and patients have a need for the subsequent development of the current spinal health condition, and the existing technology can only detect the current spinal detection image (such as X-ray image), and the result is also the current spinal health condition. Obviously, the existing technology cannot currently make accurate predictions on the development and changes of spinal health conditions, resulting in the inability to assist doctors in obtaining better diagnosis and treatment results.

[0070] In view of this, the present invention collects spinal images at different health stages to obtain spinal image big data, and then trains a deep recognition model based on the training set obtained therefrom, so that the deep recognition model establishes a functional relationship that can be used to accurately describe the changes in spinal health. In actual application, the captured spinal image data of the current health stage is input into the trained deep recognition model to obtain the corresponding spinal health prediction atlas. Outputting these atlases can enable doctors and patients to have a deeper understanding of the actual spinal health status, which is conducive to the accurate determination of subsequent treatment plans.

[0071] Among them, the deep recognition model can be built through a variety of intelligent prediction algorithms existing in the prior art, such as Deep Belief Network (DBN), Recurrent Neural Network (RNN), Long / Short Term Memory (LSTM), Gated Recurrent Unit (GRU), Auto Encoder Neural Network (Auto Encoder (AE)), Hopfield Network (HN), Deep Belief Network (DBN), etc., and the present invention is not limited to this.

[0072] It should be noted that the solution of the present invention can be implemented in a variety of ways, mainly including a field end and a server. Among them, the field end can be a terminal with wireless connection function and network connection function, such as a user device such as a smart phone, a tablet computer, a desktop computer, a laptop computer, a super personal computer and a wearable device; and the server can be an independent physical server, or it can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), big data and artificial intelligence platforms.

[0073] Further, the spinal image big data includes first spinal images at different health stages;

[0074] Then the step of forming a training set based on the spinal image big data includes:

[0075] Performing first spine parameter extraction processing on each of the first spine images, determining a first attribute of each of the first spine images based on the extracted first spine parameters, and labeling each of the first spine images using the first attribute;

[0076] Performing fitting processing on each of the first spinal images having the same first attribute to obtain a second spinal image and corresponding second spinal parameters, and determining a second attribute of the second spinal image according to the second spinal parameters;

[0077] Calculating a first matching degree between the second attribute of each second spinal column image and a standard attribute stage table, determining a first stage sequence number of each second spinal column image according to the first matching degree, and associating a plurality of first spinal column images corresponding to each second spinal column image with the first stage sequence number;

[0078] Starting from the first stage sequence number of each second spinal column image, traversing backward, and associating a number of the first spinal column images corresponding to the second spinal column image involved in the traversal with the second stage sequence number;

[0079] A plurality of training data are obtained according to the first spinal column image and / or the second spinal column image, and each of the training data is organized into the training set.

[0080] In this step, the first attribute of each first spinal image is determined according to the spinal parameters. The first attribute is used to describe the health status of the spine, including but not limited to whether there is scoliosis, the number of scoliosis segments, the scoliosis direction, the scoliosis Cobb angle, etc. Figure 2 As shown, for each first spine image having the same first attribute (for example, Figure 21.1-1.3) shown in the figure, and then a virtual, equivalent second spinal image (for example, Figure 2 2.1) shown in, similar to the above, the second attribute is determined based on the second spinal parameter of the second spinal image; at the same time, a standard attribute stage table can be pre-set based on artificial experience, and the standard attribute stage table can describe the stage attribute of spinal health, and then the serial number of the health stage of each equivalent second spinal image can be determined by matching calculation; finally, the second spinal image corresponding to the different stage serial numbers is used as the starting point to traverse backwards, that is, the data after a certain health stage is used as a training data, so that a number of training data starting from different health stages can be obtained, and then a training set can be formed, wherein, when training, the first spinal image can be used, the second spinal image can be used, or both can be used at the same time; accordingly, the deep recognition model obtained by training also corresponds to the establishment of multiple aforementioned functional relationships. Wherein, in addition to the corresponding first spinal image and / or second spinal image, the training data can also include the first spinal parameters and first attributes corresponding to the first spinal image, and the second spinal parameters and second attributes corresponding to the second spinal image. With such a setting, the deep recognition model can avoid repeated parameter extraction.

[0081] Through the above analysis, it can be seen that when determining the training set, the present invention determines the real health stage corresponding to each first spinal image through a two-step process. Such a setting can solve the problem of accurate identification of the health stage when there are slight differences in the first spinal parameters of each first spinal image; and, since there are large differences in the current health conditions of the spines of the input patients, the present invention also uses the second spinal images of different health stages as the starting point to traverse backward to obtain the corresponding training data. Such a setting can derive the functional relationship corresponding to different health stages, making the predicted spinal health prediction atlas more accurate.

[0082] It should be noted that the standard attribute stage table is an attribute comparison table that takes the healthy spine as the starting point and the extremely severe spine as the end point. Figure 3 As shown, the first stage sequence number corresponds one-to-one to the standard attribute stage table, while the second stage sequence number is a renumbering of the traversal result. For example, when the first stage sequence number is 5, but it is located at the 4th position of the traversal result, the second stage sequence number is 4.

[0083] Further, the first spinal image is associated with a user attribute and is associated with a third stage sequence number;

[0084] Then, associating a plurality of the first spinal images corresponding to each of the second spinal images with the first stage sequence number includes:

[0085] Determine at least two of the first spinal images based on the third stage sequence number, and calculate the mutation degree according to the first spinal parameters of the at least two first spinal images;

[0086] If the mutation degree is greater than or equal to a first threshold, the corresponding first spinal column image is deleted, and the remaining first spinal column images are associated with the first stage sequence number.

[0087] In this step, the collected first spinal images can be associated with each user and marked with serial numbers, and the spinal health of a single user should develop gradually. Therefore, the present invention calculates the mutation degree between adjacent second spinal images and deletes those second spinal images with too high mutation degree, thereby making the training set more accurate.

[0088] The following are some examples:

[0089] There are 20 first spine images of user A collected. The mutation degree between adjacent first spine images is calculated one by one according to the first spine parameters (such as cobb angle). For example, if the mutation degree between the 9th and 10th first spine images is too large, it means that the 10th first spine image is abnormal, for example, the first spine image of user B is incorrectly associated with user A, or user A has an abnormal "mutation" in the first spine parameters due to reasons other than traffic. At this time, in order to prevent the abnormal "mutation" of the first spine image from affecting the establishment of the aforementioned functional relationship, it should be deleted.

[0090] It should be noted that the mutation degree can be determined based on the difference, change trend, similarity, etc. between the first spine parameters of adjacent first spine images. For the similarity, for example, the 9th image is C-type scoliosis, and the 10th image is S-type scoliosis. At this time, the similarity is low and the corresponding mutation degree is high. The similarity calculation involved here can be obtained based on cosine similarity, Pearson correlation coefficient, Tanimoto coefficient, Hamming distance, etc., and the present invention is not limited to this.

[0091] Furthermore, after the mutation degree is greater than or equal to the first threshold, the method further includes:

[0092] Calculate whether the mutation degree is less than the second threshold, if so, then:

[0093] Determine a plurality of first mechanical models based on the first spinal parameters of the first spinal image that is adjacent to the first spinal image, and calculate a correlation value between the first spinal parameters of the first spinal image and the corresponding first mechanical models;

[0094] If the correlation value is greater than or equal to the third threshold, the mutation degree is determined to be greater than or equal to the first threshold, and the corresponding first spinal image is deleted; otherwise, the mutation degree is determined to be less than the first threshold, and the corresponding first spinal image is not deleted.

[0095] In this improved step, although the change in the health status of the spine occurs gradually overall, there are also rapid deteriorations at certain stages, and when the sampling interval is long, the situation where the health status difference between adjacent first spinal images is large is more likely to occur. In view of this, the present invention further analyzes several first mechanical models (mainly for the parts where scoliosis has already occurred) based on the first spinal parameters in the adjacent image before the first spinal image with the "mutation", and then analyzes the correlation value between the first spinal parameters in the first spinal image with the "mutation" and the mechanical model (where the mechanical model is the cause, the first spinal parameters corresponding to the "mutation" are the result, and the correlation value is the degree of correspondence between the cause and the result). When the correlation is higher than the third threshold, it can be determined that the "mutation" is a local deterioration of the disease development and is normal. Otherwise, it is determined that the "mutation" may be caused by other factors and does not belong to normal pathological changes.

[0096] It should be noted that the second threshold is used to describe an intermediate situation where the mutation degree is greater than or equal to the first threshold but has not exceeded an obvious limit.

[0097] Furthermore, the third threshold can be determined as follows:

[0098] Determine a plurality of second mechanical models based on the first spinal parameters of the first spinal image that is second precedingly adjacent to the first spinal image, calculate a first similarity between the second mechanical models and the corresponding first mechanical models, and determine the third threshold based on the first similarity;

[0099] The third threshold is negatively correlated with the first similarity.

[0100] In this step, the present invention further calculates the first similarity of the mechanical models corresponding to the two images before the "mutated" first spinal image. When the first similarity is lower, that is, the difference between the two mechanical models is greater, it means that the "mutation" or the precursor of local deterioration is more obvious. At this time, the smaller the third threshold is set, the higher the sensitivity of judging the "mutation" of the "mutated" first spinal image is, and the calculation accuracy of the correlation value is further improved.

[0101] Furthermore, when the depth recognition model is trained using the training set, the following loss function is used:

[0102]

[0103] In the formula, L represents the loss function of the deep recognition model during training, M represents the current number of training iterations, and F i represents the first equivalent feature matrix output by the deep recognition model after receiving the i-th round of training data, f i represents the second equivalent feature matrix of the i-th round of training data received by the deep recognition model, D(F i , f i ) represents F i With f i The distance between (such as Euclidean distance, cosine distance, etc.); F j Represents the output corresponding to j≠y after receiving the i-th round of training data i The characteristic matrix of i The output feature matrix exceeds the set range of the i-th round of training. Accordingly, D(F j , f i ) represents F j With f i The distance between (such as Euclidean distance, cosine distance, etc.); m is the adjustment base. Among them, the equivalent matrix is ​​obtained by fitting the characteristic matrices of multiple input data and output data.

[0104] In this step, the loss function directly affects the training results of the deep recognition model. The present invention designs the above-mentioned loss function, wherein the training data are input into the deep recognition model for training according to the rounds. After each round of training, the first equivalent feature matrix output in this round within the set range and each feature matrix exceeding the set range are calculated respectively (wherein the deep recognition model has multiple outputs in each training round). The above formula can then be used to obtain the distance between the failed data and the successful data in the training results of this round. When the distance is close enough, that is, the failed data is closely surrounding the successful data center, it means that the training meets the requirements and the training can be terminated.

[0105] Furthermore, the step of inputting the spinal image data into the trained deep recognition model, and the deep recognition model outputting a spinal health prediction atlas, comprises:

[0106] Calculating a third spinal column parameter corresponding to the spinal column image data, inputting the third spinal column parameter into the trained deep recognition model, wherein the deep recognition model obtains an initial prediction atlas, wherein the initial prediction atlas includes a plurality of second spinal column images and corresponding fourth stage serial numbers;

[0107] calculating a second similarity between the third spine parameter and the first spine parameter corresponding to each of the first spine images associated with the second spine image;

[0108] If the second similarity is greater than or equal to a fourth threshold, the first spinal image associated with the second spinal image is used as a target prediction image, and the target prediction image corresponding to each second spinal image is associated with the fourth stage sequence number to obtain the spinal health prediction atlas;

[0109] The deep recognition model outputs the spinal health prediction atlas.

[0110] In this step, the depth recognition model of the present invention first screens out the corresponding second spinal images and their corresponding serial numbers based on the second spinal parameters corresponding to the spinal image data and the functional relationship obtained through training, and then screens out the first spinal image associated with the second spinal image based on the second similarity calculation, wherein the screened first spinal image serves as the target prediction image for the final output. The present invention sets the depth recognition model into two parts, which work in an orderly manner and can improve the computational efficiency of the depth recognition model. Among them, there may be multiple first spinal images that satisfy the second similarity greater than or equal to the fourth threshold, and the output method in this case will be introduced later.

[0111] Furthermore, the second similarity is calculated by the following formula:

[0112]

[0113] In the formula, s(data i ) represents the similarity between the second spine parameter corresponding to the spine image data and the first spine parameter of the i-th first spine image associated with the second spine image; d(data i ) represents the similarity value between the second spine parameter and the first spine parameter calculated by using the benchmark similarity algorithm; d min d max represents the minimum and maximum values ​​of the similarity values ​​between the second spine parameters and the first spine parameters calculated using different similarity algorithms; k represents the number of the first spine images associated with the second spine image, that is, the number of the first spine parameters.

[0114] Among them, the present invention comprehensively uses multiple traditional similarity algorithms to obtain a more accurate second similarity calculation result. For example, the baseline similarity algorithm can be a Euclidean distance algorithm, and other similarity algorithms can be cosine similarity, Pearson correlation coefficient, Tanimoto coefficient, Hamming distance, etc., so different similarity algorithms are used to obtain corresponding similarity values, and then a more accurate second similarity value is obtained based on the above-mentioned similarity fusion formula.

[0115] Furthermore, the outputting of the spinal health prediction atlas according to a preset strategy includes:

[0116] Calculating a second matching degree between a third attribute corresponding to the spinal image data of the current health stage and the standard attribute stage table, determining a fifth stage sequence number based on the second matching degree, and determining the number of images in the spinal health prediction atlas according to the fifth stage sequence number;

[0117] Determine a sequence number of a key stage according to the third attribute, and correct the number of images in the key stage according to the sequence number of the key stage;

[0118] The degree of correction is positively correlated with the criticality value of the critical stage.

[0119] In this step, different health stages of the spine have different focuses, so the present invention first determines the fifth stage number corresponding to the spinal image data of the current health stage in a manner similar to the above, that is, determines which health stage the currently input spinal image data belongs to. Based on this, the appropriate number of images in the spinal health prediction atlas can be determined, and the number of images in the key stage can be appropriately corrected (for example, increasing the number of images), and the higher the criticality value of the key stage, the more images of the key stage are set.

[0120] Among them, when the current monitoring stage is in the early stage (i.e., mild scoliosis), it is necessary to output a relatively sparse, holistic spinal health prediction atlas that can describe the changes in spinal health, so that at least the patient can know the necessity of scoliosis treatment. When the current monitoring stage is in the middle stage, especially when the later development stage is a stage where surgical treatment is difficult, a more detailed and larger number of spinal health prediction atlases should be output. At the same time, the middle stage will also correspond to the key stage of treatment (such as the aforementioned optimal surgical treatment period, or the last surgical treatment period), and the number of images in the spinal health prediction atlas should also be appropriately increased in this key stage. Of course, the above is only an example, and is not intended to limit the scope of protection of the present invention. Other situations can be set accordingly, and the present invention will not be repeated here. In addition, for the key stage, the correlation between different current monitoring stages and corresponding key stages can be established based on artificial experience, and it can also be predicted by building a deep recognition model.

[0121] It should be noted that the spine health prediction atlas can be output in the form of conventional static images, GIF animations, or videos, and can be output based on a timeline. Of course, the present invention does not exclude other output forms, which will not be described in detail here. Figure 4As shown, when there may be multiple first spinal images satisfying the second similarity greater than or equal to the fourth threshold, when outputting longitudinally based on the time axis, multiple first spinal images are output laterally at corresponding positions simultaneously or separately.

[0122] Embodiment 2

[0123] See also Figure 5 , Figure 5 Schematic diagram of a spinal health change prediction system based on big data disclosed in an embodiment of the present invention. Figure 5 As shown, a spinal health change prediction system based on big data according to an embodiment of the present invention comprises a first acquisition module (101), a second acquisition module (102), a processing module (103), a storage module (104), and an output module (105), wherein the processing module (101) is connected to the first acquisition module (101), the second acquisition module (102), the storage module (104), and the output module (105); wherein,

[0124] The first acquisition module (101) is used to acquire big data of spinal images at different health stages and transmit the big data to the processing module (103);

[0125] The second acquisition module (102) is used to acquire spinal image data of the current healthy stage and transmit it to the processing module (103);

[0126] The storage module (104) is used to store executable computer program code;

[0127] The output module (103) is used to output the spine health prediction atlas according to a preset strategy;

[0128] The processing module (103) is used to execute the method described in the first embodiment by calling the executable computer program code in the storage module (104).

[0129] The specific functions of a spinal health change prediction system based on big data in this embodiment refer to the above-mentioned embodiment 1. Since the system in this embodiment adopts all the technical solutions of the above-mentioned embodiments, it at least has all the beneficial effects brought by the technical solutions of the above-mentioned embodiments, which will not be described one by one here.

[0130] Embodiment 3

[0131] See also Figure 6 , Figure 6An electronic device disclosed in an embodiment of the present invention comprises: a memory storing executable program code; a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the method described in the first embodiment.

[0132] Embodiment 4

[0133] The embodiment of the present invention further discloses a computer storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method described in the first embodiment is executed.

[0134] The computer storage medium of the embodiment of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0135] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0136] The program code embodied on the computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0137] Computer program code for performing the operations of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0138] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may also include more other equivalent embodiments without departing from the concept of the present invention, all of which belong to the scope of protection of the present invention.

Claims

1. A method for predicting spinal health changes based on big data, characterized in that: The steps include: Acquire big data of spinal images at different health stages, build a training set based on the big data of spinal images, and use the training set to train a deep recognition model; Acquire spinal image data of the current health stage, input the spinal image data into the trained deep recognition model, and the deep recognition model outputs a spinal health prediction atlas; Outputting the spinal health prediction atlas according to a preset strategy; Wherein, the spinal image big data includes first spinal images at different health stages; and the forming of a training set based on the spinal image big data includes: Performing first spine parameter extraction processing on each of the first spine images, determining a first attribute of each of the first spine images based on the extracted first spine parameters, and labeling each of the first spine images using the first attribute; Performing fitting processing on each of the first spinal images having the same first attribute to obtain a second spinal image and corresponding second spinal parameters, and determining a second attribute of the second spinal image according to the second spinal parameters; Calculating a first matching degree between the second attribute of each second spinal column image and a standard attribute stage table, determining a first stage sequence number of each second spinal column image according to the first matching degree, and associating a plurality of first spinal column images corresponding to each second spinal column image with the first stage sequence number; Starting from the first stage sequence number of each second spinal column image, traversing backward, and associating a number of the first spinal column images corresponding to the second spinal column image involved in the traversal with the second stage sequence number; Obtaining a number of training data according to the first spinal column image and / or the second spinal column image, and forming each of the training data into the training set; The step of inputting the spinal image data into the trained deep recognition model, and the deep recognition model outputting a spinal health prediction atlas, comprises: Calculating a third spinal column parameter corresponding to the spinal column image data, inputting the third spinal column parameter into the trained deep recognition model, wherein the deep recognition model obtains an initial prediction atlas, wherein the initial prediction atlas includes a plurality of second spinal column images and corresponding fourth stage serial numbers; calculating a second similarity between the third spine parameter and the first spine parameter corresponding to each of the first spine images associated with the second spine image; If the second similarity is greater than or equal to a fourth threshold, the first spinal image associated with the second spinal image is used as a target prediction image, and the target prediction image corresponding to each second spinal image is associated with the fourth stage sequence number to obtain the spinal health prediction atlas; The deep recognition model outputs the spinal health prediction atlas; The step of outputting the spinal health prediction atlas according to a preset strategy includes: Calculating a second matching degree between a third attribute corresponding to the spinal image data of the current health stage and the standard attribute stage table, determining a fifth stage sequence number based on the second matching degree, and determining the number of images in the spinal health prediction atlas according to the fifth stage sequence number; Determine a sequence number of a key stage according to the third attribute, and modify the number of images of the key stage according to the sequence number of the key stage; The degree of correction is positively correlated with the criticality value of the critical stage.

2. The method for predicting spinal health changes based on big data according to claim 1 is characterized in that: The first spinal image is associated with a user attribute and is associated with a third stage sequence number; Then, associating a plurality of the first spinal images corresponding to each of the second spinal images with the first stage sequence number includes: Determine at least two of the first spinal images based on the third stage sequence number, and calculate the mutation degree according to the first spinal parameters of the at least two first spinal images; If the mutation degree is greater than or equal to a first threshold, the corresponding first spinal column image is deleted, and the remaining first spinal column images are associated with the first stage sequence number.

3. The method for predicting spinal health changes based on big data according to claim 2 is characterized in that: After the mutation degree is greater than or equal to the first threshold, the method further includes: Calculate whether the mutation degree is less than the second threshold, if so, then: Determine a plurality of first mechanical models based on the first spinal column parameters of the first spinal column image that is adjacent to the first spinal column image, and calculate a correlation value between the first spinal column parameters of the first spinal column image and the corresponding first mechanical models; If the correlation value is greater than or equal to the third threshold, the mutation degree is determined to be greater than or equal to the first threshold, and the corresponding first spinal image is deleted; otherwise, the mutation degree is determined to be less than the first threshold, and the corresponding first spinal image is not deleted.

4. The method for predicting spinal health changes based on big data according to claim 3 is characterized by: The third threshold is determined as follows: Determine a plurality of second mechanical models based on the first spinal parameters of the first spinal image that is second precedingly adjacent to the first spinal image, calculate a first similarity between the second mechanical models and the corresponding first mechanical models, and determine the third threshold based on the first similarity; The third threshold is negatively correlated with the first similarity.

5. A spinal health change prediction system based on big data, comprising a first acquisition module, a second acquisition module, a processing module, a storage module, and an output module, wherein the processing module is connected to the first acquisition module, the second acquisition module, the storage module, and the output module; wherein: The first acquisition module is used to acquire big data of spinal images at different health stages and transmit them to the processing module; The second acquisition module is used to acquire the spinal image data of the current health stage and transmit it to the processing module; The storage module is used to store executable computer program code; The output module is used to output the spine health prediction atlas according to a preset strategy; It is characterized in that: the processing module is used to execute the method according to any one of claims 1-4 by calling the executable computer program code in the storage module.

6. An electronic device comprising: A memory storing executable program code; a processor coupled to the memory; It is characterized in that: the processor calls the executable program code stored in the memory to execute the method according to any one of claims 1-4.

7. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 4 is executed.

Images