CN111241331A

CN111241331A - Image searching method, device, equipment and medium based on artificial intelligence

Info

Publication number: CN111241331A
Application number: CN202010038730.6A
Authority: CN
Inventors: 张晓颖; 王季勇
Original assignee: Ping An Technology Shenzhen Co Ltd
Current assignee: Ping An Technology Shenzhen Co Ltd
Priority date: 2020-01-14
Filing date: 2020-01-14
Publication date: 2020-06-05
Also published as: WO2021143019A1

Abstract

The invention discloses an image searching method, device, equipment and medium based on artificial intelligence. The image searching method based on artificial intelligence comprises the following steps: acquiring an image search request, wherein the image search request comprises a target user identifier, an original CT image corresponding to the target user identifier and an original critical organ drawing image; inputting the original CT image and the original critical organ drawing image into a dose analysis model to generate an original dose distribution map corresponding to the target user identification; carrying out registration processing on the original dose distribution map to obtain a standard dose distribution map; inputting the standard dose distribution map into an image searching model, and acquiring a target characteristic vector corresponding to a target user identifier; and inquiring a radiotherapy plan database based on the target characteristic vector to obtain a target dose distribution map matched with the target characteristic vector. The image searching method based on artificial intelligence can improve the acquisition efficiency and accuracy of the target dose distribution map.

Description

Image searching method, device, equipment and medium based on artificial intelligence

Technical Field

The invention relates to the technical field of medical image processing, in particular to an image searching method, device, equipment and medium based on artificial intelligence.

Background

In the radiotherapy process, a radiotherapy plan is designed based on a dose distribution map so as to carry out radiotherapy on the basis of the radiotherapy plan.

With the development of computer technology, researchers have proposed that a dose distribution map which can be applied to patients for radiotherapy is automatically generated by using a deep learning trained model, but the generated dose distribution map usually ignores the spatial position relationship between organs, and has low accuracy, so that the reference value of the dose distribution map in clinical application is not high.

Disclosure of Invention

The embodiment of the invention provides an image searching method, device, equipment and medium based on artificial intelligence, and aims to solve the problems of low acquisition efficiency or low accuracy of a current radiotherapy dose distribution map.

An artificial intelligence based image search method, comprising:

acquiring an image search request, wherein the image search request comprises a target user identifier, an original CT image corresponding to the target user identifier and an original critical organ delineation image;

inputting the original CT image and the original organ-at-risk delineation image into a dose analysis model, and generating an original dose distribution map corresponding to the target user identification;

carrying out registration processing on the original dose distribution map to obtain a standard dose distribution map;

inputting the standard dose distribution map into an image search model to obtain a target characteristic vector corresponding to the target user identifier;

and inquiring a radiotherapy plan database based on the target characteristic vector to obtain a target dose distribution map matched with the target characteristic vector.

An artificial intelligence-based image search apparatus, comprising:

the system comprises an image search request acquisition module, a comparison module and a comparison module, wherein the image search request acquisition module is used for acquiring an image search request, and the image search request comprises a target user identifier, an original CT image and an original critical organ delineation image which correspond to the target user identifier;

the original dose distribution map acquisition module is used for inputting the original CT image and the original organ-at-risk delineation image into a dose analysis model and generating an original dose distribution map corresponding to the target user identifier;

the standard dose distribution map acquisition module is used for carrying out registration processing on the original dose distribution map to acquire a standard dose distribution map;

the target characteristic vector acquisition module is used for inputting the standard dose distribution map into an image search model and acquiring a target characteristic vector corresponding to the target user identifier;

and the target dose distribution map acquisition module is used for querying a radiotherapy plan database based on the target characteristic vector and acquiring a target dose distribution map matched with the target characteristic vector.

A computer device comprising a memory, a processor and a computer program stored in said memory and executable on said processor, said processor implementing the steps of the artificial intelligence based image search method described above when executing said computer program.

A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the artificial intelligence based image search method as described above.

According to the image searching method, the device, the equipment and the medium based on the artificial intelligence, the original CT image and the original critical organ delineation image in the image searching request are obtained, the original CT image and the original critical organ delineation image are input into the dose analysis model, the original dose distribution map is generated, and technical support is provided for image searching. The original dose distribution map is subjected to registration processing, a standard dose distribution map is obtained, so that the difference among different individuals is eliminated, the purpose of information fusion is achieved, the standard dose distribution map is input into an image search model, a target feature vector corresponding to the target user identification is obtained, a radiotherapy plan database is inquired based on the target feature vector, a target dose distribution map matched with the target feature vector is obtained, a server can conveniently search a historical radiotherapy plan which is stored in a relevant mode in the target radiotherapy plan database according to the target dose distribution map, and the historical radiotherapy plan is sent to a client side, so that a clinician can conveniently make the target radiotherapy plan according to the historical radiotherapy plan, and the making efficiency and accuracy of the target radiotherapy plan are improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments of the present invention will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without inventive labor.

FIG. 1 is a schematic diagram of an application environment of an artificial intelligence-based image search method according to an embodiment of the present invention;

FIG. 2 is a flowchart of an artificial intelligence based image search method according to an embodiment of the present invention;

FIG. 3 is another flow chart of an artificial intelligence based image search method in an embodiment of the invention;

FIG. 4 is another flow chart of an artificial intelligence based image search method in an embodiment of the invention;

FIG. 5 is another flow chart of an artificial intelligence based image search method in an embodiment of the invention;

FIG. 6 is another flow chart of an artificial intelligence based image search method in an embodiment of the invention;

FIG. 7 is another flow chart of an artificial intelligence based image search method in an embodiment of the invention;

FIG. 8 is another flow chart of an artificial intelligence based image search method in an embodiment of the invention;

FIG. 9 is a diagram of an artificial intelligence based image search apparatus according to an embodiment of the present invention;

FIG. 10 is a schematic diagram of a computer device according to an embodiment of the invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

The image searching method based on artificial intelligence provided by the embodiment of the invention can be applied to the application environment shown in figure 1. Specifically, the image searching method based on artificial intelligence is applied to an image searching system, the image searching system comprises a client and a server as shown in fig. 1, the client and the server are communicated through a network and are used for generating and processing an original CT image and an original critical organ delineation image corresponding to a target user identifier, a standard dose distribution map is generated, a historical registration distribution map similar to the standard dose distribution map is quickly searched through an image searching model to serve as a target dose distribution map, so that the acquisition efficiency and accuracy of the target dose distribution map are improved, and a target radiotherapy plan is made for reference by a clinician. The target radiotherapy plan is a radiotherapy plan for a target user. The client is also called a user side, and refers to a program corresponding to the server and providing local services for the client. The client may be installed on, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server may be implemented as a stand-alone server or as a server cluster consisting of a plurality of servers.

In one embodiment, as shown in fig. 2, an artificial intelligence based image searching method is provided, which is described by taking the server in fig. 1 as an example, and includes the following steps:

s201: and acquiring an image search request, wherein the image search request comprises a target user identifier, an original CT picture corresponding to the target user identifier and an original critical organ delineation picture.

The target user refers to a user who performs detection to determine the distribution of tumors so that a clinician can make a target radiotherapy plan. The target radiotherapy plan is a radiotherapy plan made for a target user, and one user corresponds to one radiotherapy plan. The target user identifier is an identifier for uniquely identifying the target user, and for example, the target user identifier may be a name of the target user, an identification number of the target user, and the like.

The original CT image is an image obtained by CT scanning by the target user. In CT (Computed Tomography), a layer surface where a part to be examined is located is scanned with an X-ray beam, the X-ray transmitted through the layer surface is received by a detector, converted into visible light, converted into an electrical signal by photoelectric conversion, and converted into a digital signal by an analog/digital converter (analog/digital converter), and the digital signal is processed by a computer to obtain a cross-sectional or stereoscopic image of the part to be examined, which is an original CT image, so that a fine lesion of the part to be examined can be found by using the original CT image. It is understood that the examined region includes a diseased region and a non-diseased region, for example, the examined region is a lung, the diseased region is a right lung, and the non-diseased region is a left lung and organs around the left lung, etc.

An organ at risk refers to a non-diseased, important tissue or organ within the radiation range of the radiation therapy radiation. The original organ-at-risk delineation map is a map obtained by delineating non-diseased important tissues or organs within the radiation range of the radiotherapy radiation in the original CT map, and corresponds to the target user.

Specifically, a target user goes to a hospital to perform CT scanning so as to obtain an original CT image of the target user, an organ at risk is outlined according to the original CT image so as to obtain an original organ at risk delineation image, and image data such as the original CT image and the original organ at risk delineation image of the target user and a target user identifier are stored in an image database in an associated mode. It will be appreciated that each user's image data, including but not limited to CT images and organ-at-risk delineation images, is in one-to-one correspondence with the user identification for management. The image database is a library for storing image data of all users.

As an example, a clinician generates an image search request with a target user identifier, an original CT image corresponding to the target user identifier, and an original endanger-hooked image by clicking an image search button on a client, and sends the image search request to a server, so that the server obtains the image search request.

S202: and inputting the original CT image and the original critical organ drawing image into a dose analysis model to generate an original dose distribution map corresponding to the target user identification.

Wherein the dose analysis model is a model for generating a predicted dose distribution map. The dose analysis model is a model generated based on deep neural network training, the dose analysis model is a model formed by training a training sample by adopting a deep neural network, and the training sample comprises a training CT image, a training critical organ sketch image and a corresponding training dose distribution map, which correspond to the same user identification.

The original dose distribution map is the radiation dose predicted by the dose analysis model, namely the radiation dose required by the diseased part of the target user when the target user is subjected to radiotherapy. Because the original dose distribution map is generated by the dose analysis model, the spatial positions existing between different organs in the original dose distribution map may be ignored, the accuracy is not high, and the clinical standard cannot be met, so that the clinician cannot directly generate the target radiotherapy plan from the original dose distribution map.

Specifically, an original CT image and an original critical organ drawing image are input into a dose analysis model, so that an original dose distribution map corresponding to a target user identification is quickly generated, and technical support is provided for image search.

S203: and carrying out registration processing on the original dose distribution map to obtain a standard dose distribution map.

The registration process is used for comparing or fusing images acquired by different users under different conditions so as to accurately search images subsequently. It can be understood that, because different users have different body types and different organ sizes or spatial positions of different users, the dose distribution map or CT map of different users is mapped to another image by finding a spatial transformation, and the points of the same image of different users corresponding to the same spatial position are in one-to-one correspondence to eliminate the difference between different individuals.

Specifically, an image registration algorithm is adopted to register an original CT image and a standard CT image to obtain standard registration parameters, the original dose distribution map is converted based on the standard registration parameters to obtain the standard dose distribution map, the organ of the original CT image corresponding to the target user identifier and the organ of the standard CT image are ensured to be in corresponding positions, the influence of factors such as the sizes and the spatial positions of the organs of different individuals on image searching can be eliminated, the subsequent searching of similar images is ensured, and the accuracy of image searching is improved. The standard CT image refers to a universal CT image template.

S204: and inputting the standard dose distribution map into an image searching model, and acquiring a target characteristic vector corresponding to the target user identifier.

The image search model is a pre-trained model for identifying a dose distribution map to output a feature vector. The image search model is specifically a model generated by applying convolutional neural network training based on a ternary loss function, the distance between the feature vectors generated by the similar image through the image search model is small, specifically, the distance between the corresponding feature vectors generated by the similar registration distribution map through the image search model is small, the distance between the feature vectors generated by the dissimilar image through the image search model is large, specifically, the distance between the corresponding feature vectors generated by the dissimilar registration distribution map through the image search model is large, i.e. the image search model ensures that the distance of the feature vectors generated by the image search model for similar images is smaller than the distance of the feature vectors generated by the image search model for dissimilar images, so that a historical registration profile similar to the standard dose profile can be subsequently accurately acquired to determine the target dose profile. Specifically, the standard dose distribution map is input into the image search model, so that a target feature vector corresponding to the target user identifier is obtained, and a basis is provided for subsequently searching the standard dose distribution map.

S205: and inquiring a radiotherapy plan database based on the target characteristic vector to obtain a target dose distribution map matched with the target characteristic vector.

The radiotherapy plan database is used for storing user data corresponding to historical users after radiotherapy, and the user data comprises historical user identification, a historical CT (computed tomography) graph, a historical dose distribution graph, a historical radiotherapy plan and historical feature vectors corresponding to the historical dose distribution graph which are stored in a related mode.

The historical user refers to a user who has undergone radiotherapy. The history user identifier is an identifier for uniquely identifying the history user. The historical CT map is an image acquired by a historical user through a CT scan. The historical dose profile is a dose profile that the historical user developed during radiotherapy. A historic radiotherapy plan is a radiotherapy plan that is acquired by a historic user during a radiotherapy procedure. The historical feature vector is a feature vector obtained by inputting the historical dose distribution map into the image search model. The target dose profile is a historical registration profile similar to the standard dose profile.

The historical characteristic vectors are obtained through the image searching model, the distance between the characteristic vectors of similar historical dose distribution maps can be ensured to be small, the distance between the characteristic vectors of dissimilar historical dose distribution maps is ensured to be large, so that the historical registration distribution map similar to the standard dose distribution map can be accurately obtained subsequently, and the target dose distribution map is determined based on the similar historical dose distribution maps.

Specifically, the similarity of the target feature vector and the historical feature vector corresponding to any historical user identifier in the radiotherapy plan database is calculated by adopting a similarity calculation formula, the top M (M is a positive integer) historical feature vectors with the maximum similarity are obtained, and the historical dose distribution map corresponding to the historical feature vectors is used as the target dose distribution map, so that the server searches the historical radiotherapy plan which is stored in a relevant manner in the target radiotherapy plan database according to the target dose distribution map and sends the historical radiotherapy plan to the client, so that a clinician formulates the target radiotherapy plan according to the historical radiotherapy plan, and the formulation efficiency and accuracy of the target radiotherapy plan are improved. It can be understood that, since the historical radiotherapy plan is a plan that the historical user has performed radiotherapy, the historical radiotherapy plan has a strong reference value, so as to shorten the time required by a clinician to formulate a target radiotherapy plan, shorten the radiotherapy period of the target user, and provide the target user with timely radiotherapy.

The image searching method based on artificial intelligence provided by the embodiment obtains the original CT image and the original critical organ delineation image in the image searching request, inputs the original CT image and the original critical organ delineation image into the dose analysis model, generates the original dose distribution map, and provides technical support for image searching. The method comprises the steps of registering an original dose distribution map, obtaining a standard dose distribution map to eliminate differences among different individuals so as to achieve the purpose of information fusion, inputting the standard dose distribution map into an image search model, obtaining a target feature vector corresponding to a target user identifier, inquiring a radiotherapy plan database based on the target feature vector, obtaining a target dose distribution map matched with the target feature vector, enabling a server to search a historical radiotherapy plan which is stored in a relevant mode in the target radiotherapy plan database according to the target dose distribution map, and sending the historical radiotherapy plan to a client side so that a clinician can make a target radiotherapy plan according to the historical radiotherapy plan, and improving the making efficiency and accuracy of the target radiotherapy plan.

In an embodiment, as shown in fig. 3, before step S204, that is, before inputting the standard dose distribution map into the image search model and acquiring the target feature vector corresponding to the target user identifier, the artificial intelligence based image search method further includes:

s301: historical user image data of the first historical user identification is acquired, the historical user image data including a historical CT map, a historical organ-at-risk delineation map, and a historical dose distribution map.

Wherein the first historical user identification is an identification of one historical user in the image database. The historical user image data is image data stored in an image database in association with the first historical user identification. The historical user image data includes, but is not limited to, historical CT maps, historical organ-at-risk maps, and historical dose distribution maps. The historical organ-at-risk sketch refers to a sketch obtained by sketching non-diseased important tissues or organs within the radiation range of radiotherapy radiation in the historical CT (computed tomography) picture of the same historical user.

S302: and inputting the historical CT image and the historical critical organ drawing image corresponding to the first historical user identifier into a dose analysis model, and acquiring an analysis dose distribution map corresponding to the first historical user identifier.

And predicting the historical CT image and the historical critical organ delineation image corresponding to the first historical user identification by using a dose distribution model to obtain a predicted dose distribution map, and taking the analyzed dose distribution map as training data for training an image search model.

S303: and acquiring a historical registration distribution diagram and an analysis registration distribution diagram based on the historical dose distribution diagram and the analysis dose distribution diagram corresponding to the first historical user identification.

The historical registration distribution map is obtained after image registration processing is carried out on the historical dose distribution map. The analysis registration distribution map refers to an image obtained after image registration processing is performed on the analysis dose distribution map.

Specifically, an image registration algorithm may be adopted to register the historical CT image of the first historical user identifier with the standard CT image to obtain a registration parameter corresponding to the first historical user identifier, perform image registration processing on the historical dose distribution map based on the registration parameter to obtain a historical registration distribution map, and perform image registration processing on the analysis dose distribution map based on the registration parameter to obtain an analysis registration distribution map. The historical registration distribution map and the analysis registration distribution map are obtained through registration processing, and the trained model can be ensured to be more accurate.

S304: and querying an image database, and determining a contrast dose distribution graph corresponding to the first historical user identifier based on the historical dose distribution graphs of other historical user identifiers.

Specifically, a historical dose distribution map corresponding to other historical user identifications except the first historical user identification is obtained from an image database, registration processing is carried out on the historical dose distribution maps corresponding to the other historical user identifications to generate a contrast dose distribution map, so that differences among images of different users are eliminated, and the accuracy of training an image search model is ensured. It is understood that, in order to obtain more training samples, the historical dose distribution maps corresponding to a plurality of other historical user identifiers except the first historical user identifier may be subjected to a registration process to obtain a contrast dose distribution map so as to obtain a sufficient number of training samples.

S305: and taking the historical registration distribution map, the analysis registration distribution map and the contrast dose distribution map corresponding to the first historical user identification as training samples.

Specifically, the historical registration distribution map, the analysis registration distribution map, and the contrast dose distribution map corresponding to the first historical user identifier are used as training samples, it can be understood that the similarity is high because the historical registration distribution map and the analysis registration distribution map are image data of the same historical user, and the contrast dose distribution map is a dose distribution map obtained after registration processing is performed on the historical dose distribution maps except for the first historical user identifier and is a dose distribution map which is dissimilar to the historical registration distribution map and the analysis registration distribution map corresponding to the first historical user identifier, so as to ensure that the generated image search model can make the distance of the feature vectors corresponding to the dissimilar images large, specifically, make the distance of the feature vectors corresponding to the dissimilar registration distribution maps large, and thus can ensure the accuracy of subsequent image search.

S306: and inputting the training sample into a convolutional neural network based on a ternary loss function to carry out model training, and obtaining an image search model.

Specifically, a historical registration distribution diagram, an analysis registration distribution diagram and a contrast dose distribution diagram corresponding to the first historical user identifier are input into a weight-sharing convolutional neural network based on a ternary loss function for training, and when the loss is smaller than a function convergence value, the image search model is trained completely. The function convergence value is presetThe value used to evaluate whether the loss function meets the convergence requirement may be zero. Wherein the ternary loss function is

M (M is a positive integer) represents the number of training samples, i (i is a positive integer, i ≦ M) represents the ith set of training samples, x^aVectors, x, representing the correspondence of the historical registration profiles^pVectors, x, corresponding to the representation analysis registration profilesⁿA vector corresponding to the contrast dose distribution map is represented,

representing a euclidean distance metric between the historical registration profile and the analysis registration profile,

representing a Euclidean distance measure between the historical registration profile and the contrast dose profile, α meaning x^aAnd xⁿThe sum of the distance between x^aAnd x^pThe distance between them is the smallest distance, + represents [, ]]When the value in (b) is greater than the function convergence value, take [ 2 ]]The inner value is loss; []And when the value of the image search model is smaller than the function convergence value, the training of the image search model is finished.

In the image search method based on artificial intelligence provided by this embodiment, the historical CT image and the historical critical organ delineation image corresponding to the first historical user identifier are input to the dose analysis model, so as to quickly obtain the analysis dose distribution map corresponding to the first historical user identifier. And acquiring a historical registration distribution diagram and an analysis registration distribution diagram based on the historical dose distribution diagram and the analysis dose distribution diagram corresponding to the first historical user identification so as to eliminate the influence of the difference between the images of different users on model training. And querying an image database, determining a contrast dose distribution graph corresponding to the first historical user identifier based on historical dose distribution graphs of other historical user identifiers, and taking the historical registration distribution graph, the analysis registration distribution graph and the contrast dose distribution graph corresponding to the first historical user identifier as training samples, so that the accuracy of subsequent image searching can be guaranteed. And inputting the training sample into a convolutional neural network based on a ternary loss function to perform model training, and quickly acquiring an image search model to ensure that the generated image search model can enable the distance of the feature vectors corresponding to dissimilar images to be large, and specifically enable the distance of the feature vectors corresponding to dissimilar registration distribution maps to be large.

In an embodiment, as shown in fig. 4, the step S303 of obtaining the historical registration distribution map and the analysis registration distribution map based on the historical dose distribution map and the analysis dose distribution map corresponding to the first historical user identifier includes:

s401: and registering the historical CT image corresponding to the first historical user identifier with the standard CT image by adopting an image registration algorithm to obtain historical registration parameters.

Specifically, the historical CT map is preprocessed to provide a basis for registration, and the preprocessing process includes: carrying out noise elimination processing on the historical CT image so as to eliminate interference factors; when the pixel sizes of the historical CT image and the standard CT image are different, the size of the historical CT image is adjusted to enable the pixel sizes of the historical CT image and the standard CT image to be matched, so that the characteristics of the historical CT image and the standard CT image correspond to each other, and therefore the accuracy of the acquired historical registration parameters is guaranteed.

Selecting a first feature from a specific position of the preprocessed historical CT image, and selecting a second feature from a position corresponding to the standard CT image and the historical CT image to obtain the first feature and the second feature corresponding to the same specific position, wherein in order to better determine the corresponding relationship between the historical CT image and the standard CT image, a plurality of corresponding position features can be selected from the historical CT image and the standard CT image; establishing a three-dimensional coordinate system, determining the three-dimensional coordinate of the first characteristic of the historical CT image and the three-dimensional coordinate of the second characteristic of the standard CT image, determining a registration function of the preprocessed historical CT image and the standard CT image according to the two three-dimensional coordinates, and acquiring historical registration parameters based on the registration function. Specifically, the historical registration parameters are parameters of a registration function, and the historical CT image is resampled to verify the accuracy of the historical registration parameters.

S402: and carrying out image registration on the historical dose distribution map and the analysis dose distribution map based on the historical registration parameters to obtain a historical registration distribution map and an analysis registration distribution map.

Specifically, the historical dose distribution map and the analyzed dose distribution map corresponding to the first historical user identifier are registered according to the historical registration parameters, that is, the historical dose distribution map and the analyzed dose distribution map corresponding to the first historical user identifier are spatially transformed according to the historical registration parameters. The spatial transformation can be conversion of rotation, reduction, amplification and the like to obtain a historical registration distribution map and an analysis registration distribution map, and the training samples are subjected to registration processing, so that the difference between images of different users can be eliminated, and the accuracy of generating an image search model is ensured.

In the image search method based on artificial intelligence provided by this embodiment, an image registration algorithm is used to register the historical CT image corresponding to the first historical user identifier with the standard CT image, so as to obtain historical registration parameters. And carrying out image registration on the historical dose distribution map and the analysis dose distribution map based on the historical registration parameters to obtain the historical registration distribution map and the analysis registration distribution map, so that the difference between images of different users can be eliminated, and the accuracy of generating an image search model is ensured.

In an embodiment, as shown in fig. 5, the step S304 of querying the image database, and determining a contrast dose distribution map corresponding to the first historical user identifier based on the historical dose distribution maps of the other historical user identifiers includes:

s501: a target region is determined based on the first historical user identification corresponding historical registration profile.

Wherein the target region is a tumor region. The historical registration distribution map comprises a target area part and an organ-at-risk part, and the target area part is outlined in the historical registration distribution map corresponding to the first historical user identification through a computer or manually so as to search the historical dose distribution map corresponding to other historical user identifications of the same target area part in the subsequent process. For example, if the target region is a lung tumor, the historical dose distribution map of the lung tumor is screened from the image database, so that the number of image searching is reduced, and the efficiency of subsequently acquiring the contrast dose distribution map is improved.

S502: and inquiring an image database based on the target region part to obtain a contrast dose distribution graph with the similarity to the historical registration distribution graph smaller than a first preset threshold value.

The first preset threshold is a threshold used for judging whether the dose distribution maps corresponding to different historical user identifications meet the similarity standard or not.

Specifically, the similarity between the historical registration distribution diagram corresponding to the first historical user identifier and the historical registration distribution diagrams corresponding to the other historical user identifiers is calculated through an image matching algorithm, and the historical registration distribution diagrams corresponding to the other historical user identifiers with the similarity smaller than a first preset threshold are determined as the contrast dose distribution diagram of the first historical user identifier, so that a sample of the training image search model is obtained. As can be understood, since the historical registration distribution maps corresponding to each user identifier may be different, X (X is a positive integer) historical registration distribution maps corresponding to other historical user identifiers with the smallest similarity may be obtained as training samples, so as to ensure that the number of samples used for training the image search model is sufficient. In this embodiment, the image matching algorithm includes, but is not limited to, a grayscale-based matching algorithm and a feature-based matching algorithm.

In the image searching method based on artificial intelligence provided by the embodiment, the target region is determined based on the historical registration distribution map corresponding to the first historical user identifier, so that the number of image searching is reduced, and the efficiency of obtaining the contrast dose distribution map is improved. And inquiring an image database based on the target region, and acquiring a contrast dose distribution graph with the similarity smaller than a first preset threshold with the historical registration distribution graph so as to acquire a sample of the training image search model and provide a technology for the training image search model.

In an embodiment, as shown in fig. 6, in step S304, querying the image database, and determining the historical dose distribution map of the other historical user identifiers as the contrast dose distribution map corresponding to the first historical user identifier includes:

s601: and acquiring a historical registration distribution map corresponding to the second historical user identification from the image database.

The second historical user identifier is the identifier of any other historical user except the first historical user identifier. Specifically, a historical registration distribution map corresponding to the second historical user identification is obtained from the image database, so that a contrast dose distribution map for training is obtained subsequently.

S602: and acquiring a first DVH (digital video graphics) graph corresponding to the first historical user identifier based on the historical registration distribution graph and the historical critical organ delineation graph of the first historical user identifier, and generating a corresponding second DVH graph based on the historical critical organ delineation graph of the first historical user identifier and the historical registration distribution graph corresponding to the second historical user identifier.

Wherein DVH is an abbreviation for Dose-Volume Histogram, referring to the Dose Volume Histogram. The volume of the diseased site is represented on the ordinate of the DVH plot, and the dose of radiation therapy is represented on the abscissa. The dose volume histogram specifically includes two curves, one of which represents the dose-volume relationship of the target region in the radiotherapy plan, and the other of which represents the dose-volume relationship of the organs at risk in the radiotherapy plan.

Specifically, a target area and an organ at risk are sketched out on a historical CT image corresponding to a first historical user identifier in advance, the area where the target area is located is converted into vector 1 to be represented, the area where the organ at risk is located is converted into vector 0 to be represented, and a first vector matrix is generated; similarly, the numerical value of the radiotherapy dose in the history registration distribution map corresponding to the first history user identifier is converted into a corresponding second vector matrix; and multiplying the first vector matrix by the second vector matrix to obtain a dose-volume curve of the target region corresponding to the first historical user identifier. It can be understood that, since the vector of the region where the target region is located is 1, the relationship between the dose and the volume of the target region is retained, and the dose and volume curve of the target region corresponding to the first historical user identifier can be obtained by multiplying the first vector matrix by the second vector matrix. In contrast, the area where the target region is located is converted into vector 0 to represent, the area where the organ at risk is located is converted into vector 1 to represent, a third vector matrix is generated, and the second vector matrix is multiplied by the third vector matrix to obtain the dose-volume relation in the organ at risk, namely the dose-volume curve of the organ at risk corresponding to the first historical user identifier, so that a first DVH graph is generated.

Similarly, the numerical value of the radiotherapy dose in the history registration distribution graph corresponding to the second history user identifier is converted into a fourth vector matrix, and the first vector matrix is multiplied by the fourth vector matrix to obtain the relationship between the dose and the volume of the target region corresponding to the second history user identifier; in contrast, the third vector matrix is multiplied by the fourth vector matrix to obtain a second historical user-identified organ-at-risk dose-volume relationship to generate a second DVH plot.

S603: and performing similarity calculation on the first DVH image and the second DVH image by adopting a similarity calculation method to obtain the target similarity.

Wherein the target similarity is a value representing a degree of similarity between the first DVH map and the second DVH map.

Specifically, the dose and volume curve of the target region in the first DVH image are spaced at equal intervals by N points, the dose and volume curve of the target region in the second DVH image are spaced at equal intervals by N points, and the distance difference between the dose and volume curve of the target region in the first DVH image and the N points in the second DVH image is calculated to form a first coordinate difference; similarly, calculating the distance difference between the dose of the organ at risk of the first DVH image and the N points in the volume curve to form a second coordinate difference, wherein the N points are equally spaced from the dose of the organ at risk of the first DVH image and the volume curve, and the N points are equally spaced from the dose of the organ at risk of the second DVH image and the volume curve; and calculating the average value of the first coordinate difference value and the second coordinate difference value to serve as target similarity, and subsequently accurately determining the contrast dose distribution map according to the target similarity. It is understood that the first DVH map is obtained according to the history registration distribution map of the first history user identifier and the history organ-at-risk delineation map, the second DVH map is obtained according to the history organ-at-risk delineation map of the first history user identifier and the history registration distribution map corresponding to the second history user identifier, and if the history registration distribution map corresponding to the first history user identifier is similar to the history registration distribution map corresponding to the second history user identifier, the first DVH map and the second DVH map should also be similar. Conversely, if the historical registration profiles corresponding to the first historical user identifier and the historical registration profiles corresponding to the second historical user identifier are not similar, the first DVH graph and the second DVH graph should also be dissimilar.

S604: and if the target similarity is smaller than a second preset threshold, taking the historical registration distribution map of the second historical user identifier as a contrast dose distribution map corresponding to the historical registration distribution map of the first historical user identifier.

The second preset threshold is used for judging whether the first DVH image and the second DVH image reach the value of the similar standard or not.

Specifically, when the target similarity is smaller than a preset threshold, it is indicated that the first DVH graph and the second DVH graph are not similar, the historical dose distribution graph corresponding to the second historical user identifier is used as a contrast dose distribution graph corresponding to the historical registration distribution graph of the first historical user identifier, and the dissimilar historical dose distribution graph is used as a contrast dose distribution graph, so that the generated image search model can accurately identify the distance between the similar image and the dissimilar image, the accuracy of the generated image search model is improved, and the dose distribution graph is ensured to be input into the feature vector generated by the image search model subsequently.

In the image searching method based on artificial intelligence provided by this embodiment, a historical registration distribution map corresponding to a second historical user identifier is acquired from an image database, so as to subsequently acquire a contrast agent distribution map for training, a first DVH image corresponding to the first historical user identifier is acquired based on the historical registration distribution map and a historical critical organ delineation map of the first historical user identifier, and a corresponding second DVH image is generated based on the historical critical organ delineation map of the first historical user identifier and the historical registration distribution map corresponding to any historical user identifier. And performing similarity calculation on the first DVH image and the second DVH image by adopting a similarity calculation method, and subsequently accurately determining a contrast dose distribution map according to the target similarity. And if the target similarity is smaller than a preset threshold, taking the historical dose distribution map of other user identifications as a contrast dose distribution map corresponding to the historical registration distribution map of the first historical user identification.

In one embodiment, as shown in fig. 7, step S205, querying the radiotherapy plan database based on the target feature vector to obtain a target dose distribution map matching the target feature vector, includes:

s701: and querying a radiotherapy plan database to obtain a historical feature vector corresponding to any historical user identifier.

Specifically, after the image search model is trained, the historical dose distribution maps corresponding to all the historical user identifiers are registered, and the registered historical registration distribution maps are input into the image search model to generate the historical feature vectors corresponding to each historical user identifier and stored in the radiotherapy plan database.

S702: and calculating a target similarity value of the target feature vector and the historical feature vector.

The target similarity value is a value indicating the degree of similarity between the target feature vector and the history feature vector.

Specifically, the server may calculate the target similarity value of the target feature vector and the history feature vector quickly through a similarity algorithm. In this embodiment, the similarity algorithm includes, but is not limited to, a cosine similarity algorithm, a euclidean distance algorithm, a manhattan algorithm, and the like.

S703: and if the target similarity value is larger than a third preset threshold value, determining the historical registration distribution map corresponding to the historical characteristic vector as a target dose distribution map.

The third preset threshold is a value used for judging whether the historical feature vector and the target feature vector reach a similarity standard or not.

Specifically, target similarity values of the target feature vectors and the historical feature vectors are calculated, the target similarity values are sorted from large to small, historical registration distribution maps corresponding to the top M (M is a positive integer) historical feature vectors of which the target similarity values are larger than a third preset threshold are selected according to sorting results, the selected historical dose distribution maps are used as target dose distribution maps, historical radiotherapy plans which are stored in a related mode are searched according to the historical dose distribution maps in a subsequent mode, and the historical radiotherapy plans of the target user are made to serve as references for clinicians.

The image search model provided by the embodiment queries a radiotherapy plan database and obtains a historical feature vector corresponding to any historical user identifier. And calculating a target similarity value of the target feature vector and the historical feature vector. And if the target similarity value is larger than a third preset threshold value, determining the historical registration distribution map corresponding to the historical characteristic vector as a target dose distribution map, so as to search a historical radiotherapy plan which is stored in a related manner according to the historical dose distribution map, and providing a reference for a clinician to formulate the historical radiotherapy plan of the target user.

In an embodiment, as shown in fig. 8, before step S205, before querying the radiotherapy plan database based on the feature vector corresponding to the target user identifier, the artificial intelligence based image search method further includes:

s801: n historical dose profiles are obtained from an image database.

Specifically, the image data of all history users who have undergone radiotherapy are stored in an image database, the image data of each history user is stored in association with the corresponding history user identifier and is stored in a server, the server can quickly acquire the history dose distribution maps of all history users through a query algorithm such as keyword matching, and for example, the server can search all history dose distribution maps through a keyword "dose distribution map".

S802: and carrying out registration processing on the historical dose distribution diagram by adopting an image registration algorithm to obtain a historical registration distribution diagram.

Specifically, the historical dose distribution map is registered by using an image registration algorithm, and the obtained historical registration distribution map is consistent with step S401, and is not described herein again to avoid repetition.

S803: and inputting the historical registration distribution map into an image search model to generate a corresponding historical feature vector.

Specifically, the history registration distribution map is input into the image search model, and the process of generating the corresponding history feature vector is consistent with the process of generating the target feature vector in step S204, which is not described herein again to avoid repetition.

S804: and storing the historical characteristic vector of each historical user identification and the corresponding historical registration distribution map in a radiotherapy plan database in an associated mode.

Specifically, the historical feature vectors, the historical registration distribution map, the historical dose distribution map and the historical radiotherapy plan of the same historical user identification are stored in a radiotherapy plan database in an associated mode, so that the historical feature vectors similar to the target feature vectors can be searched subsequently, the historical radiotherapy plan of the historical user with the diseased part similar to the target user can be obtained, and reference is provided for a clinician.

The image search model provided in this embodiment obtains N historical dose distribution maps from an image database, performs registration processing on the historical dose distribution maps by using an image registration algorithm to obtain a historical registration distribution map, inputs the historical registration distribution map into the image search model, and generates corresponding historical feature vectors, so as to calculate target similarity values of the historical feature vectors and the target feature vectors in the subsequent process. And storing the historical characteristic vector of each historical user identification and the corresponding historical registration distribution map in a radiotherapy plan database in an associated manner so as to search the historical characteristic vector similar to the target characteristic vector in the subsequent process, thereby obtaining the historical radiotherapy plan of the historical user of the diseased part similar to the target user and providing reference for a clinician.

It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

In one embodiment, an artificial intelligence based image search apparatus is provided, and the artificial intelligence based image search apparatus corresponds to the artificial intelligence based image search method in the above embodiments one to one. As shown in fig. 9, the artificial intelligence based image search apparatus includes an image search request acquisition module 901, an original dose distribution map acquisition module 902, a standard dose distribution map acquisition module 903, a target feature vector acquisition module 904, and a target dose distribution map acquisition module 905.

The functional modules are explained in detail as follows:

an image search request obtaining module 901, configured to obtain an image search request, where the image search request includes a target user identifier, an original CT image corresponding to the target user identifier, and an original critical organ delineation image.

And an original dose distribution map obtaining module 902, configured to input the original CT map and the original critical organ delineation map into the dose analysis model, and generate an original dose distribution map corresponding to the target user identifier.

And a standard dose distribution map acquisition module 903, configured to perform registration processing on the original dose distribution map to acquire a standard dose distribution map.

And a target feature vector obtaining module 904, configured to input the standard dose distribution map into the image search model, and obtain a target feature vector corresponding to the target user identifier.

And a target dose distribution map obtaining module 905, configured to query the radiotherapy plan database based on the target feature vector, and obtain a target dose distribution map matched with the target feature vector.

Further, before the target feature vector obtaining module 904, the artificial intelligence based image searching apparatus further includes: the device comprises a historical user image data acquisition module, an analysis dose distribution diagram acquisition module, an image registration processing module, a contrast dose distribution diagram determination module, a training sample determination module and an image search model acquisition module.

And the historical user image data acquisition module is used for acquiring historical user image data of the first historical user identifier, wherein the historical user image data comprises a historical CT (computed tomography) map, a historical organ-at-risk drawing map and a historical dose distribution map.

And the analysis dose distribution map acquisition module is used for inputting the historical CT map and the historical critical organ drawing map corresponding to the first historical user identifier into the dose analysis model to acquire the analysis dose distribution map corresponding to the first historical user identifier.

And the image registration processing module is used for acquiring a historical registration distribution diagram and an analysis registration distribution diagram based on the historical dose distribution diagram and the analysis dose distribution diagram corresponding to the first historical user identification.

And the contrast dose distribution map determining module is used for inquiring the image database and determining the contrast dose distribution map corresponding to the first historical user identifier based on the historical dose distribution maps of other historical user identifiers.

And the training sample determining module is used for taking the historical registration distribution map, the analysis registration distribution map and the contrast dose distribution map corresponding to the first historical user identification as training samples.

And the image search model acquisition module is used for inputting the training sample into a convolutional neural network based on a ternary loss function to carry out model training to acquire an image search model.

Further, an image registration processing module comprising:

and the historical registration parameter acquisition unit is used for registering the historical CT image corresponding to the first historical user identifier with the standard CT image by adopting an image registration algorithm to acquire historical registration parameters.

And the registration distribution diagram acquisition unit is used for carrying out image registration on the historical dose distribution diagram and the analysis dose distribution diagram based on the historical registration parameters to acquire the historical registration distribution diagram and the analysis registration distribution diagram.

Further, a contrast dose profile determination module comprising: a target region determination unit and a first judgment unit.

And the target region determining unit is used for determining the target region based on the historical registration distribution map corresponding to the first historical user identification.

And the first judgment unit is used for inquiring the image database based on the target region part and acquiring a contrast dose distribution graph of which the similarity with the historical registration distribution graph is smaller than a first preset threshold value.

Further, a contrast dose profile determination module comprising: the device comprises a history registration distribution diagram acquisition unit, a DVH diagram acquisition unit, a target similarity acquisition unit and a second judgment unit.

And the historical registration distribution image acquisition unit is used for acquiring a historical registration distribution image corresponding to the second historical user identification from the image database.

And the DVH image acquisition unit is used for acquiring a first DVH image corresponding to the first historical user identifier based on the historical registration distribution map of the first historical user identifier and the historical organ-at-risk delineation map, and generating a corresponding second DVH image based on the historical organ-at-risk delineation map of the first historical user identifier and the historical registration distribution map corresponding to the second historical user identifier.

And the target similarity obtaining unit is used for calculating the similarity of the first DVH image and the second DVH image by adopting a similarity calculation method to obtain the target similarity.

And the second judging unit is used for taking the historical registration distribution map of the second historical user identifier as a contrast dose distribution map corresponding to the historical registration distribution map of the first historical user identifier if the target similarity is smaller than a second preset threshold.

Further, the target dose distribution map obtaining module 905 includes: a radiotherapy plan database query unit, a feature vector calculation unit and a third judgment unit.

And the radiotherapy plan database query unit is used for querying the radiotherapy plan database and acquiring the historical characteristic vector corresponding to any historical user identifier.

And the characteristic vector calculating unit is used for calculating a target similarity value of the target characteristic vector and the historical characteristic vector.

And the third judging unit is used for determining the historical registration distribution map corresponding to the historical characteristic vector as the target dose distribution map if the target similarity value is greater than a third preset threshold value.

Further, before the target dose distribution map obtaining module 905, the artificial intelligence based image searching apparatus further includes: the radiotherapy treatment system comprises a historical dose distribution map acquisition unit, a registration processing unit, a historical feature vector generation unit and a radiotherapy plan database generation unit.

And the historical dose distribution map acquisition unit is used for acquiring the N historical dose distribution maps from the image database.

And the registration processing unit is used for carrying out registration processing on the historical dose distribution map by adopting an image registration algorithm to obtain a historical registration distribution map.

And the historical characteristic vector generating unit is used for inputting the historical registration distribution map into the image searching model and generating a corresponding historical characteristic vector.

And the radiotherapy plan database generation unit is used for storing the historical characteristic vector of each historical user identifier and the corresponding historical registration distribution map in a radiotherapy plan database in an associated mode.

For specific limitations of the artificial intelligence based image search apparatus, reference may be made to the above limitations of the artificial intelligence based image search method, which are not described herein again. The respective modules in the artificial intelligence based image search apparatus described above may be wholly or partially implemented by software, hardware, and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.

In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in fig. 10. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used to store historical registration profiles. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement an artificial intelligence based image search method.

In an embodiment, a computer device is provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the steps of the artificial intelligence based image search method in the foregoing embodiments are implemented, for example, steps S201 to S205 shown in fig. 2 or steps shown in fig. 3 to fig. 8, which are not repeated herein to avoid repetition. Alternatively, the processor implements the functions of each module/unit in the embodiment of the artificial intelligence based image search apparatus when executing the computer program, for example, the functions of the image search request obtaining module 901, the original dose distribution diagram obtaining module 902, the standard dose distribution diagram obtaining module 903, the target feature vector obtaining module 904, and the target dose distribution diagram obtaining module 905 shown in fig. 9, and are not described herein again to avoid repetition.

In an embodiment, a computer-readable storage medium is provided, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the artificial intelligence based image search method in the foregoing embodiments are implemented, for example, steps S201 to S205 shown in fig. 2 or steps shown in fig. 3 to fig. 8, and are not described herein again to avoid repetition. Alternatively, the processor implements the functions of each module/unit in the embodiment of the artificial intelligence based image search apparatus when executing the computer program, for example, the functions of the image search request obtaining module 901, the original dose distribution diagram obtaining module 902, the standard dose distribution diagram obtaining module 903, the target feature vector obtaining module 904, and the target dose distribution diagram obtaining module 905 shown in fig. 9, and are not described herein again to avoid repetition.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-mentioned division of the functional units and modules is illustrated, and in practical applications, the above-mentioned function distribution may be performed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-mentioned functions.

The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present invention, and are intended to be included within the scope of the present invention.

Claims

1. An image searching method based on artificial intelligence is characterized by comprising the following steps:

2. The artificial intelligence based image search method of claim 1, wherein before the inputting the standard dose distribution map into an image search model and obtaining the target feature vector corresponding to the target user identifier, the artificial intelligence based image search method further comprises:

acquiring historical user image data of a first historical user identifier, wherein the historical user image data comprises a historical CT (computed tomography) image, a historical organ-at-risk drawing image and a historical dose distribution map;

inputting a historical CT image and a historical critical organ drawing image corresponding to a first historical user identifier into a dose analysis model, and acquiring an analysis dose distribution map corresponding to the first historical user identifier;

acquiring a historical registration distribution diagram and an analysis registration distribution diagram based on a historical dose distribution diagram and an analysis dose distribution diagram corresponding to the first historical user identifier;

querying an image database, and determining a contrast dose distribution graph corresponding to the first historical user identifier based on historical dose distribution graphs of other historical user identifiers;

taking a historical registration distribution map, an analysis registration distribution map and a contrast dose distribution map corresponding to the first historical user identification as training samples;

and inputting the training sample into a convolutional neural network based on a ternary loss function to carry out model training, and obtaining an image search model.

3. The artificial intelligence based image searching method of claim 2, wherein the obtaining of the historical registration profile and the analysis registration profile based on the historical dose profile and the analysis dose profile corresponding to the first historical user identification comprises:

registering the historical CT image corresponding to the first historical user identifier with a standard CT image by adopting an image registration algorithm to obtain historical registration parameters;

and carrying out image registration on the historical dose distribution map and the analysis dose distribution map based on the historical registration parameters to obtain a historical registration distribution map and an analysis registration distribution map.

4. The artificial intelligence based image searching method of claim 2, wherein the querying the image database to determine the contrast dose distribution map corresponding to the first historical user identifier based on historical dose distribution maps of other historical user identifiers comprises:

determining a target region based on a historical registration distribution map corresponding to the first historical user identification;

and inquiring an image database based on the target region, and acquiring a contrast dose distribution graph with the similarity to the historical registration distribution graph being smaller than a first preset threshold value.

5. The artificial intelligence based image searching method of claim 2, wherein the querying the image database to determine the contrast dose distribution map corresponding to the first historical user identifier based on historical dose distribution maps of other historical user identifiers comprises:

acquiring a historical registration distribution diagram corresponding to the second historical user identification from an image database;

acquiring a first DVH (digital video graphics) graph corresponding to the first historical user identifier based on the historical registration distribution graph and the historical critical organ drawing graph of the first historical user identifier, and generating a second DVH graph corresponding to the second historical user identifier based on the historical critical organ drawing graph and the historical registration distribution graph corresponding to the second historical user identifier of the first historical user identifier;

performing similarity calculation on the first DVH image and the second DVH image by adopting a similarity calculation method to obtain target similarity;

and if the target similarity is smaller than a second preset threshold, taking the historical registration distribution map of the second historical user identifier as a contrast dose distribution map corresponding to the historical registration distribution map of the first historical user identifier.

6. The artificial intelligence based image searching method of claim 1, wherein the querying a radiotherapy plan database based on the target feature vector to obtain a target dose distribution map matching the target feature vector comprises:

inquiring a radiotherapy plan database, and acquiring a historical feature vector corresponding to any historical user identifier;

calculating a target similarity value of the target feature vector and the historical feature vector;

and if the target similarity value is larger than a third preset threshold value, determining the historical registration distribution map corresponding to the historical characteristic vector as a target dose distribution map.

7. The artificial intelligence based image search method of claim 1, wherein prior to the querying a radiotherapy plan database based on the target feature vector, the artificial intelligence based image search method further comprises:

acquiring N historical dose distribution maps from an image database;

adopting the image registration algorithm to perform registration processing on the historical dose distribution map to obtain a historical registration distribution map;

inputting the historical registration distribution map into an image search model to generate a corresponding historical feature vector;

storing the historical feature vector for each historical user identification in association with the corresponding historical registration profile in a radiotherapy plan database.

8. An image search device based on artificial intelligence, comprising:

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the steps of the artificial intelligence based image search method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the artificial intelligence based image search method according to any one of claims 1 to 7.