Clinical remote consultation system and method based on artificial intelligence

Abstract

The invention relates to the technical field of clinical remote consultation, in particular to a clinical remote consultation system and method based on artificial intelligence. The method comprises the steps of obtaining illness state description information of a patient and function information of each target consultation department, determining a to-be-consulted department corresponding to the patient based on the illness state description information of the patient and the function information of each target consultation department, establishing clinical remote consultation between the patient and the to-be-consulted department, obtaining a clinical image of the patient, and sending the clinical image to the patient. Inputting the clinical image into the clinical image segmentation model, and outputting the clinical image containing the target diagnosis part; receiving a clinical image containing the target diagnosis part, and registering the clinical image containing the target diagnosis part with the clinical image of the historical target diagnosis part of the patient to obtain a deformed clinical image; according to the method, the medical image data can be automatically analyzed, key features can be extracted, and doctors can be assisted in rapid and accurate diagnosis.

Description

Technical Field

[0001] The present invention relates to the technical field of clinical remote consultation, and in particular to a clinical remote consultation system and method based on artificial intelligence. Background Art

[0002] Traditional clinical consultation methods are limited by geographical distance, time arrangements, and uneven distribution of medical resources, which means that patients often need to travel long distances to large medical institutions to seek expert consultations. This not only increases the cost and burden of medical treatment for patients, but also limits the effective use of medical resources. With the rapid development of information technology, remote consultation has gradually become an important way to solve the above problems.

[0003] In the existing technology, clinical remote consultation has the following problems:

[0004] Low consultation efficiency: Traditional remote consultation systems often rely on manual operations, such as consultation discussions through video conferencing. This method is not only time-consuming and labor-intensive, but also easily affected by factors such as network delays and video quality, resulting in low consultation efficiency.

[0005] Low diagnostic accuracy: In remote consultations, doctors mainly rely on medical imaging data and other examination data provided by patients for diagnosis. However, due to the complexity and diversity of medical imaging data, it is difficult for doctors to comprehensively and accurately analyze these data in a short period of time, which affects the accuracy of diagnosis.

[0006] Therefore, there is an urgent need for artificial intelligence-based clinical remote consultation systems and methods to solve the above problems. Summary of the invention

[0007] The purpose of the present invention is to provide a clinical remote consultation system and method based on artificial intelligence: to solve the technical problem that the traditional remote consultation system in the existing solution relies on manual operation, such as consultation discussion through video conferencing, which leads to low consultation efficiency.

[0008] The purpose of the present invention can be achieved through the following technical solutions:

[0009] On the one hand, a clinical teleconsultation system based on artificial intelligence, the system includes a clinical teleconsultation activation unit, a patient clinical image segmentation unit, a patient clinical image registration unit and a consultation analysis unit;

[0010] The clinical remote consultation activation unit is used to obtain the patient's condition description information and the functional information of each target consultation department, determine the department to be consulted corresponding to the patient based on the patient's condition description information and the functional information of each target consultation department, and establish a clinical remote consultation between the patient and the department to be consulted;

[0011] The patient clinical image segmentation unit is used to obtain the patient's clinical image, input the clinical image into the clinical image segmentation model, output the clinical image containing the target diagnostic part, and transmit the clinical image containing the target diagnostic part to the patient clinical image registration unit;

[0012] The patient clinical image registration unit is used to receive the clinical image containing the target diagnostic part, and register the clinical image containing the target diagnostic part with the clinical image of the patient's historical target diagnostic part to obtain the deformed clinical image, and transmit the deformed clinical image to the consultation analysis unit;

[0013] The consultation analysis unit is used to receive the deformed clinical image and analyze the deformed clinical image to obtain a consultation result.

[0014] Furthermore, determining the department to be consulted corresponding to the patient based on the patient's condition description information and the functional information of each target consultation department specifically includes the following process:

[0015] Set the bipartite network G = (U, V), where U is the set of disease description information and V is the set of functional information of each target consultation department. Calculate the node association Q of the bipartite network G:

[0016]

[0017] Where N is the number of information in the condition description information that is associated with the functional information of each target consultation department, and A i，j Represents the adjacency matrix of the set of U and V, i, j represent the node numbers, and 2m represents the adjacency matrix A i，j The sum of all the edges in d i represents the degree of i, d j represents the degree of j, c i represents the community in which node i exists, c j represents the community in which node j exists, δ*(c i , c j ) is the conditional restriction coefficient. If N is greater than the preset value, then δ*(c i , c j ) is 1, if N is less than or equal to the preset value, then δ*(c i , c j ) is 0;

[0018] The target consultation department corresponding to the bipartite network G with the largest correlation degree Q is taken as the department to be consulted corresponding to the patient.

[0019] Furthermore, calculating the amount of information associated with the functional information of each target consultation department in the condition description information includes the following process:

[0020] Based on NLP natural language processing technology, all field names and field types in the disease description information and the functional information of each target consultation department are identified;

[0021] Determine the association between fields based on the field names and field types according to the association rules;

[0022] The condition description information related to the functional information of each target consultation department is recorded as related information.

[0023] Furthermore, the clinical image is input into the clinical image segmentation model, and the output of the clinical image containing the target diagnosis part specifically includes the following process:

[0024] Extracting each characteristic image block of the clinical image based on the clinical image, and acquiring a pixel set corresponding to the characteristic image block based on the characteristic image block;

[0025] Calculate the inner product between the feature vector of the feature image block pixel set and the feature vector of the feature image block pixel set containing the target diagnosis part, and then convert it into a similarity S through the softmax function:

[0026]

[0027] Among them, q i，j Represents the feature vector corresponding to the pixel (i, j) of the feature image block pixel set, Represents the transposition of the feature vector corresponding to the pixel (i, j) in the feature image block pixel set containing the target diagnosis part;

[0028] Calculate the proximity rate P between feature image block pixel sets:

[0029]

[0030] Among them, c i，j Represents the position coordinates of the feature image block pixels, d i，j represents the position coordinates of the pixels of the characteristic image block containing the target diagnostic part, ||c i，j -d i，j || means c i，j With d i，j The distance, e is 2.72, and σ is a learnable parameter;

[0031] The sum of the similarity S and the proximity P is calculated, and based on the sum, it is determined whether the feature image block and the feature image block containing the target diagnostic part belong to the same region. If the sum exceeds the preset sum threshold, the feature image block and the feature image block containing the target diagnostic part belong to the same region. If the sum does not exceed the preset sum threshold, the feature image block and the feature image block containing the target diagnostic part do not belong to the same region, the feature image block is discarded, and a clinical image containing the target diagnostic part is output.

[0032] Furthermore, the clinical image containing the target diagnosis part and the clinical image of the patient's historical target diagnosis part are registered to obtain the deformed clinical image, which specifically includes the following process:

[0033] A function f(M, F; θ, r) = φ is established based on the Transformer registration network, where F is the moving image, M is the fixed image, φ is a deformation field, θ is a learnable parameter of f, and r is a regularization parameter;

[0034] For the image pair M and F to be registered, where M represents the clinical image containing the target diagnostic part, and F represents the clinical image of the patient's historical target diagnostic part, the two images at the original resolution are defined as M1 and F1;

[0035] The image pairs M and F to be registered are downsampled to their original resolution based on the downsampling encoder. Where n is a positive integer, and the obtained image pairs are defined as (M2; F2), (M3; F3), ..., M n ; F n );

[0036] From the lowest resolution image pair (M n ; F n ) starts with training the nth layer of Transformer-based registration network to obtain the nth layer of deformation field φ. Next, the nth layer of deformation field is upsampled and applied to the n-1th layer of moving images to obtain the deformed Will and F n-1 As the input image, train the n-1th layer of Transformer-based registration network to obtain the n-1th layer deformation field Δφ n-1 , superimposed with the up-sampled Δφ n-2 , we get φ n-1 , repeat the training until the deformation field φ1 at the original resolution is obtained, and the deformed clinical image is obtained based on the function f(M, F; θ, r) = φ

[0037] Furthermore, the deformed clinical image is analyzed to obtain the consultation result, which specifically includes the following processes:

[0038] Three-dimensional reconstruction and volume measurement are performed based on the deformed clinical image and the clinical image of the patient before treatment. By comparing the results of the two measurements, the change in the lesion volume is obtained.

[0039] Three-dimensional reconstruction and shape comparison are performed based on the deformed clinical image and the clinical image of the patient before treatment. By comparing the results of the two measurements, the change in the shape of the lesion is obtained.

[0040] The CT values ​​of the deformed clinical images were compared with those of the patients before treatment, and the changes in the lesion density were obtained by comparing the results of the two measurements.

[0041] Furthermore, a platform for three-dimensional reconstruction and volume measurement of the deformed clinical images and the clinical images of the patient before treatment includes 3DSlicer.

[0042] Furthermore, platforms for three-dimensional reconstruction and shape comparison based on deformed clinical images and clinical images of patients before treatment include ITK-SNAP.

[0043] On the other hand, the clinical remote consultation method based on artificial intelligence includes:

[0044] Obtain the patient's condition description information and the functional information of each target consultation department, determine the patient's corresponding department to be consulted based on the patient's condition description information and the functional information of each target consultation department, and establish a clinical remote consultation between the patient and the department to be consulted;

[0045] Acquire a clinical image of the patient, input the clinical image into a clinical image segmentation model, and output a clinical image containing a target diagnostic site;

[0046] Receiving a clinical image containing a target diagnostic site, and registering the clinical image containing the target diagnostic site with a historical clinical image of the target diagnostic site of the patient to obtain a deformed clinical image;

[0047] The deformed clinical images are analyzed to obtain the consultation results.

[0048] Compared with the existing solutions, the present invention achieves the following beneficial effects:

[0049] The present invention obtains the patient's condition description information and the functional information of each target consultation department, determines the department to be consulted corresponding to the patient based on the patient's condition description information and the functional information of each target consultation department, and establishes a clinical remote consultation between the patient and the department to be consulted, which can quickly realize the remote consultation link between the patient and the department to be consulted and improve the diagnosis efficiency.

[0050] The present invention obtains a clinical image of a patient, inputs the clinical image into a clinical image segmentation model, and outputs a clinical image containing a target diagnostic site; receives a clinical image containing a target diagnostic site, and aligns the clinical image containing the target diagnostic site with a clinical image of the patient's historical target diagnostic site to obtain a deformed clinical image; and analyzes the deformed clinical image to obtain a consultation result. The present invention can automatically analyze medical imaging data, extract key features, and assist doctors in making rapid and accurate diagnoses. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0052] Figure 1 is a system block diagram of a clinical remote consultation system based on artificial intelligence according to an embodiment of the present invention;

[0053] Figure 2 This is a workflow diagram of a clinical remote consultation system based on artificial intelligence according to an embodiment of the present invention;

[0054] Figure 3 The present invention is a workflow diagram of an artificial intelligence-based clinical remote consultation method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0055] 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 only 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.

[0056] In addition, the described features, structures or characteristics may be combined in one or more example embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the example embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced while omitting one or more of the specific details, or other methods, components, steps, etc. may be adopted. In other cases, well-known structures, methods, implementations or operations are not shown or described in detail to avoid obscuring various aspects of the present disclosure.

[0057] This embodiment provides a clinical remote consultation system based on artificial intelligence. Figure 1 is a system block diagram of a clinical remote consultation system based on artificial intelligence according to an embodiment of the present invention, such as Figure 1 As shown, the system includes a clinical remote consultation activation unit, a patient clinical image segmentation unit, a patient clinical image registration unit and a consultation analysis unit;

[0058] The clinical remote consultation activation unit is used to obtain the patient's condition description information and the functional information of each target consultation department, determine the department to be consulted corresponding to the patient based on the patient's condition description information and the functional information of each target consultation department, and establish a clinical remote consultation between the patient and the department to be consulted;

[0059] The patient clinical image segmentation unit is used to obtain the patient's clinical image, input the clinical image into the clinical image segmentation model, output the clinical image containing the target diagnostic part, and transmit the clinical image containing the target diagnostic part to the patient clinical image registration unit;

[0060] The patient clinical image registration unit is used to receive the clinical image containing the target diagnostic part, and register the clinical image containing the target diagnostic part with the clinical image of the patient's historical target diagnostic part to obtain the deformed clinical image, and transmit the deformed clinical image to the consultation analysis unit;

[0061] The consultation analysis unit is used to receive the deformed clinical image and analyze the deformed clinical image to obtain a consultation result.

[0062] In summary, the present invention obtains the patient's condition description information and the functional information of each target consultation department, determines the patient's corresponding department to be consulted based on the patient's condition description information and the functional information of each target consultation department, and establishes a clinical remote consultation between the patient and the department to be consulted, which can quickly realize the remote consultation link between the patient and the department to be consulted and improve the diagnosis efficiency.

[0063] Furthermore, the present invention obtains a clinical image of a patient, inputs the clinical image into a clinical image segmentation model, and outputs a clinical image containing a target diagnostic site; receives a clinical image containing a target diagnostic site, and aligns the clinical image containing the target diagnostic site with a clinical image of the patient's historical target diagnostic site to obtain a deformed clinical image; and analyzes the deformed clinical image to obtain a consultation result. The present invention can automatically analyze medical imaging data, extract key features, and assist doctors in making quick and accurate diagnoses.

[0064] In some embodiments, determining the department to be consulted corresponding to the patient based on the patient's condition description information and the functional information of each target consultation department specifically includes the following process:

[0065] Set the bipartite network G = (U, V), where U is the set of disease description information and V is the set of functional information of each target consultation department. Calculate the node association Q of the bipartite network G:

[0066]

[0067] Where N is the number of information in the condition description information that is associated with the functional information of each target consultation department, and A i，j Represents the adjacency matrix of the set of U and V, i, j represent the node numbers, and 2m represents the adjacency matrix A i，j The sum of all the edges in d i represents the degree of i, d j represents the degree of j, c i represents the community in which node i exists, c j represents the community in which node j exists, δ*(c i , c j ) is the conditional restriction coefficient. If N is greater than the preset value, then δ*(c i , c j ) is 1, if N is less than or equal to the preset value, then δ*(c i , c j ) is 0;

[0068] The target consultation department corresponding to the bipartite network G with the largest correlation degree Q is taken as the department to be consulted corresponding to the patient.

[0069] In some embodiments, Figure 2 is a workflow diagram of a clinical remote consultation system based on artificial intelligence according to an embodiment of the present invention, such as Figure 2 As shown, calculating the number of information associated with the functional information of each target consultation department in the condition description information includes the following steps:

[0070] Step S201: Identify all field names and field types in the condition description information and the functional information of each target consultation department based on NLP natural language processing technology;

[0071] Specifically, based on the NLP (Natural Language Processing) natural language processing technology, all field names and field types in the condition description information and the functional information of each target consultation department are identified: field name, field type. Data fields are columns or attributes used to store data in the database. According to the type of data stored and the purpose, data fields can be divided into different types, including primary keys, foreign keys, text fields, and numeric fields. Primary key: The primary key is a field used to uniquely identify each row of data in the table. It can quickly locate a specific record and ensure the uniqueness of the data. The primary key must be unique and cannot be empty. Foreign key: A foreign key is a field in a table whose value comes from the primary key of another table. Foreign keys are used to establish a relationship between two tables and ensure referential integrity. Text field: Text fields are used to store character data, such as strings, text comments, etc. They are usually used to store non-numeric data of variable length. Numeric field: Numeric field is used to store numeric data, such as integers, floating point numbers, etc.

[0072] Step S202: Determine the association between fields based on the field names and field types according to the association rules;

[0073] Specifically, support and confidence are defined as:

[0074] Support: The frequency with which an item set appears in all transactions. In field correlation analysis, it can be understood as the frequency with which two or more field values ​​appear at the same time.

[0075] Confidence: The conditional probability that Y is also included in the transaction containing X, expressed as Support(X,Y) / Support(X). In field correlation analysis, it can be understood as the probability that field Y takes a certain value when field X takes a certain value.

[0076] Apply the association rule mining algorithm:

[0077] Association rule mining algorithms such as Apriori and FP-Growth are used to identify frequent item sets and generate association rules. The association between fields is determined based on field names and field types based on association rule mining.

[0078] Step S203: Record the condition description information associated with the functional information of each target consultation department as associated information.

[0079] In some embodiments, inputting a clinical image into a clinical image segmentation model and outputting a clinical image containing a target diagnostic site specifically includes the following process:

[0080] Extracting each characteristic image block of the clinical image based on the clinical image, and acquiring a pixel set corresponding to the characteristic image block based on the characteristic image block;

[0081] Calculate the inner product between the feature vector of the feature image block pixel set and the feature vector of the feature image block pixel set containing the target diagnosis part, and then convert it into a similarity S through the softmax function:

[0082]

[0083] Among them, q i，j Represents the feature vector corresponding to the pixel (i, j) of the feature image block pixel set, Represents the transposition of the feature vector corresponding to the pixel (i, j) in the feature image block pixel set containing the target diagnosis part;

[0084] Calculate the proximity rate P between feature image block pixel sets:

[0085]

[0086] Among them, c i，j Represents the position coordinates of the feature image block pixels, d i，j represents the position coordinates of the pixels of the characteristic image block containing the target diagnostic part, ||c i，j -d i，j || means c i，j With d i，j The distance, e is 2.72, and σ is a learnable parameter;

[0087] The sum of the similarity S and the proximity P is calculated, and based on the sum, it is determined whether the feature image block and the feature image block containing the target diagnostic part belong to the same region. If the sum exceeds the preset sum threshold, the feature image block and the feature image block containing the target diagnostic part belong to the same region. If the sum does not exceed the preset sum threshold, the feature image block and the feature image block containing the target diagnostic part do not belong to the same region, the feature image block is discarded, and a clinical image containing the target diagnostic part is output.

[0088] In some embodiments, registering the clinical image containing the target diagnosis site with the clinical image of the patient's historical target diagnosis site to obtain the deformed clinical image specifically includes the following process:

[0089] A function f(M, F; θ, r) = φ is established based on the Transformer registration network, where F is the moving image, M is the fixed image, φ is a deformation field, θ is a learnable parameter of f, and r is a regularization parameter;

[0090] For the image pair M and F to be registered, where M represents the clinical image containing the target diagnostic part, and F represents the clinical image of the patient's historical target diagnostic part, the two images at the original resolution are defined as M1 and F1;

[0091] The image pairs M and F to be registered are downsampled to their original resolution based on the downsampling encoder. Where n is a positive integer, and the obtained image pairs are defined as (M2; F2), (M3; F3), ..., M n ; F n );

[0092] From the lowest resolution image pair (M n ; F n ) starts with training the nth layer of Transformer-based registration network

[0093] The deformation field φ of the nth layer is obtained by upsampling the deformation field of the nth layer and applying it to the moving image of the n-1th layer to obtain the deformed Will and F n-1 As the input image, train the n-1th layer of Transformer-based registration network to obtain the n-1th layer deformation field Δφ n-1 , superimposed with the up-sampled Δφ n-2 , we get φ n-1 , repeat the training until the deformation field φ1 at the original resolution is obtained, and the deformed clinical image is obtained based on the function f(M, F; θ, r) = φ

[0094] In some embodiments, analyzing the deformed clinical image to obtain the consultation result specifically includes the following process:

[0095] Three-dimensional reconstruction and volume measurement are performed based on the deformed clinical image and the clinical image of the patient before treatment. By comparing the results of the two measurements, the change in the lesion volume is obtained.

[0096] Three-dimensional reconstruction and shape comparison are performed based on the deformed clinical image and the clinical image of the patient before treatment. By comparing the results of the two measurements, the change in the shape of the lesion is obtained.

[0097] The CT values ​​of the deformed clinical images were compared with those of the patients before treatment, and the changes in the lesion density were obtained by comparing the results of the two measurements.

[0098] It is worth mentioning that the platforms for three-dimensional reconstruction and volume measurement of deformed clinical images and clinical images of patients before treatment include 3DSlicer, and the platforms for three-dimensional reconstruction and shape comparison based on deformed clinical images and clinical images of patients before treatment include ITK-SNAP.

[0099] In some embodiments, Figure 3is a workflow diagram of a clinical remote consultation method based on artificial intelligence according to an embodiment of the present invention, such as Figure 3 As shown, the method comprises the following steps:

[0100] Step S301: Acquire the patient's condition description information and the functional information of each target consultation department, determine the department to be consulted corresponding to the patient based on the patient's condition description information and the functional information of each target consultation department, and establish a clinical remote consultation between the patient and the department to be consulted;

[0101] Step S302: Acquire a clinical image of the patient, input the clinical image into a clinical image segmentation model, and output a clinical image containing a target diagnostic site;

[0102] Step S303: receiving a clinical image containing the target diagnosis part, and registering the clinical image containing the target diagnosis part with a clinical image of the patient's historical target diagnosis part to obtain a deformed clinical image;

[0103] Step S304: Analyze the deformed clinical image to obtain a consultation result.

[0104] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0105] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0106] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0107] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only some logical function divisions. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0108] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0109] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. The clinical remote consultation system based on artificial intelligence is characterized by: The system includes a clinical remote consultation activation unit, a patient clinical image segmentation unit, a patient clinical image registration unit and a consultation analysis unit; The clinical remote consultation activation unit is used to obtain the patient's condition description information and the functional information of each target consultation department, determine the department to be consulted corresponding to the patient based on the patient's condition description information and the functional information of each target consultation department, and establish a clinical remote consultation between the patient and the department to be consulted; The patient clinical image segmentation unit is used to obtain the patient's clinical image, input the clinical image into the clinical image segmentation model, output the clinical image containing the target diagnostic part, and transmit the clinical image containing the target diagnostic part to the patient clinical image registration unit; The patient clinical image registration unit is used to receive the clinical image containing the target diagnostic part, and register the clinical image containing the target diagnostic part with the clinical image of the patient's historical target diagnostic part to obtain the deformed clinical image, and transmit the deformed clinical image to the consultation analysis unit; The consultation analysis unit is used to receive the deformed clinical image and analyze the deformed clinical image to obtain a consultation result.

2. The artificial intelligence-based clinical remote consultation system according to claim 1, characterized in that: Based on the patient's condition description information and the functional information of each target consultation department, the specific department to be consulted for the patient is determined. The process includes: Set the bipartite network G = (U, V), where U is the set of disease description information and V is the set of functional information of each target consultation department. Calculate the node association Q of the bipartite network G: Where N is the number of information in the condition description information that is associated with the functional information of each target consultation department, and A i，j Represents the adjacency matrix of the set of U and V, i, j represent the node numbers, and 2m represents the adjacency matrix A i，j The sum of all the edges in d i represents the degree of i, d j represents the degree of j, c i represents the community in which node i exists, c j represents the community in which node j exists, δ*(c i , c j ) is the conditional restriction coefficient. If N is greater than the preset value, then δ*(c i , c j ) is 1, if N is less than or equal to the preset value, then δ*(c i , c j ) is 0; The target consultation department corresponding to the bipartite network G with the largest correlation degree Q is taken as the department to be consulted corresponding to the patient.

3. The artificial intelligence-based clinical remote consultation system according to claim 2 is characterized in that: Calculate the number of information in the condition description information that is associated with the functional information of each target consultation department The process includes: Based on NLP natural language processing technology, all field names and field types in the disease description information and the functional information of each target consultation department are identified; Determine the association between fields based on field names and field types according to association rules; The condition description information related to the functional information of each target consultation department is recorded as related information.

4. The artificial intelligence-based clinical remote consultation system according to claim 1, characterized in that: Inputting clinical images into the clinical image segmentation model and outputting clinical images containing the target diagnostic site specifically includes the following processes: Extracting each characteristic image block of the clinical image based on the clinical image, and acquiring a pixel set corresponding to the characteristic image block based on the characteristic image block; Calculate the inner product between the feature vector of the feature image block pixel set and the feature vector of the feature image block pixel set containing the target diagnosis part, and then convert it into a similarity S through the softmax function: Among them, q i，j Represents the feature vector corresponding to the pixel (i, j) of the feature image block pixel set, Represents the transposition of the feature vector corresponding to the pixel (i, j) in the feature image block pixel set containing the target diagnosis part; Calculate the proximity rate P between feature image block pixel sets: Among them, c i，j Represents the position coordinates of the feature image block pixels, d i，j represents the position coordinates of the pixels of the characteristic image block containing the target diagnostic part, ||c i，j -d i，j || means c i，j With d i，j The distance, e is 2.72, and σ is a learnable parameter; The sum of the similarity S and the proximity P is calculated, and based on the sum, it is determined whether the feature image block and the feature image block containing the target diagnostic part belong to the same region. If the sum exceeds the preset sum threshold, the feature image block and the feature image block containing the target diagnostic part belong to the same region. If the sum does not exceed the preset sum threshold, the feature image block and the feature image block containing the target diagnostic part do not belong to the same region, the feature image block is discarded, and a clinical image containing the target diagnostic part is output.

5. The artificial intelligence-based clinical remote consultation system according to claim 1 is characterized in that: The clinical image containing the target diagnostic part and the clinical image of the patient's historical target diagnostic part are registered to obtain the deformed clinical image, which specifically includes the following process: A function f(M, F; θ, r) = φ is established based on the Transformer registration network, where F is the moving image, M is the fixed image, φ is a deformation field, θ is a learnable parameter of f, and r is a regularization parameter; For the image pair M and F to be registered, where M represents the clinical image containing the target diagnostic part, and F represents the clinical image of the patient's historical target diagnostic part, the two images at the original resolution are defined as M1 and F1; The image pairs M and F to be registered are downsampled to their original resolution based on the downsampling encoder. Where n is a positive integer, and the obtained image pairs are defined as (M2; F2), (M3; F3), ..., (M n ; F n ); From the lowest resolution image pair (M n ; F n ) starts with training the nth layer of Transformer-based registration network to obtain the nth layer of deformation field φ. Next, the nth layer of deformation field is upsampled and applied to the n-1th layer of moving images to obtain the deformed Will and F n-1 As the input image, train the n-1th layer of Transformer-based registration network to obtain the n-1th layer deformation field Δφ n-1 , superimposed with the up-sampled Δφ n-2 , we get φ n-1 , repeat the training until the deformation field φ1 at the original resolution is obtained, and the deformed clinical image is obtained based on the function f(M, F; θ, r) = φ 6. The artificial intelligence-based clinical remote consultation system according to claim 1 is characterized in that: Analyzing the deformed clinical images and obtaining the consultation results specifically includes the following processes: Three-dimensional reconstruction and volume measurement are performed based on the deformed clinical image and the clinical image of the patient before treatment. By comparing the results of the two measurements, the change in the lesion volume is obtained. Three-dimensional reconstruction and shape comparison are performed based on the deformed clinical image and the clinical image of the patient before treatment. By comparing the results of the two measurements, the change in the shape of the lesion is obtained. The CT values ​​of the deformed clinical images were compared with those of the patients before treatment, and the changes in the lesion density were obtained by comparing the results of the two measurements.

7. The artificial intelligence-based clinical remote consultation system according to claim 6, characterized in that: Platforms for three-dimensional reconstruction and volume measurement of deformed clinical images and clinical images of patients before treatment include 3DSlicer.

8. The artificial intelligence-based clinical remote consultation system according to claim 6, characterized in that: Platforms for 3D reconstruction and shape comparison based on deformed clinical images and clinical images of patients before treatment include ITK-SNAP.

9. A clinical remote consultation method based on artificial intelligence, characterized in that: The artificial intelligence-based clinical remote consultation system according to any one of claims 1 to 8, the method comprising: Obtain the patient's condition description information and the functional information of each target consultation department, determine the patient's corresponding department to be consulted based on the patient's condition description information and the functional information of each target consultation department, and establish a clinical remote consultation between the patient and the department to be consulted; Acquire a clinical image of the patient, input the clinical image into a clinical image segmentation model, and output a clinical image containing a target diagnostic site; Receiving a clinical image containing a target diagnostic site, and registering the clinical image containing the target diagnostic site with a historical clinical image of the target diagnostic site of the patient to obtain a deformed clinical image; The deformed clinical images are analyzed to obtain the consultation results.

Images