Patient diagnosis and treatment method based on vector database and medium

Abstract

The invention relates to a patient diagnosis and treatment method based on a vector database and a medium, and relates to the technical field of intelligent medical treatment, and the method comprises the steps: collecting patient medical data, vectorizing the medical data comprising digital information, text information and image information, carrying out the feature fusion of the vectorized data, and obtaining a first fusion vector; calculating the similarity between the first fusion vector and a second fusion vector corresponding to each historical patient to obtain a preset number of second fusion vectors with the highest similarity; and obtaining an optimal treatment scheme based on a preset number of second fusion vectors with the highest similarity. Compared with the prior art, dependence on the diagnosis and treatment level of doctors can be reduced with the help of previous excellent diagnosis and treatment cases, a better treatment scheme can be quickly obtained, and the diagnosis and treatment level of existing doctors can be quickly improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent medical technology, and in particular to a patient diagnosis and treatment method and medium based on a vector database. Background Art

[0002] With the continuous development of artificial intelligence technology, big data, vector databases and large model technologies have been widely used in various fields. Especially in the medical field, these technologies can be used to analyze large amounts of medical data and help doctors improve their diagnosis and treatment levels. However, in some underdeveloped areas, the medical structure faces two major problems: one is the loss of young and excellent medical talents, and the other is the retirement of elderly medical talents. This has led to a sharp decline in the discipline capabilities of some hospitals, which is not conducive to the construction of excellent medical teams and affects the improvement of medical standards.

[0003] The existing solution is to improve the medical level in these areas mainly by introducing talents and strengthening training. However, this method requires a long time of introduction and training, more capital investment, and it is difficult to achieve certain results in the short term.

[0004] In addition, some hospitals have tried to use artificial intelligence technology to assist in diagnosis and treatment, but these technologies often lack local adaptability and expertise support, making it difficult to achieve the desired results. Summary of the invention

[0005] The technical problem to be solved by the present application is to provide a patient diagnosis and treatment method and medium based on a vector database, which has the characteristics of being able to quickly obtain a better treatment plan and quickly improve the diagnosis and treatment level of existing doctors.

[0006] In a first aspect, an embodiment provides a patient diagnosis and treatment method based on a vector database, comprising: Collecting patient medical data, including digital information, text information and image information; Vectorizing the digital information, text information and image information respectively to obtain corresponding digital vectors, text vectors and image vectors; Performing feature fusion on the digital vector, the text vector and the image vector to obtain a first fused vector; Calculating the similarity between the first fused vector and each second fused vector to obtain a preset number of second fused vectors with the highest similarity; Obtaining an optimal treatment plan based on a preset number of second fusion vectors with the highest similarity; The method for obtaining the second fusion vector includes: obtaining medical data of historical patients, obtaining corresponding digital vectors, text vectors and image vectors for the medical data of any historical patient, and performing feature fusion on the digital vectors, text vectors and image vectors to obtain the second fusion vector corresponding to the any historical patient.

[0007] In one embodiment, the digital information, text information and image information are respectively vectorized to obtain corresponding digital vectors, text vectors and image vectors, including: For digital information, the numerical data is standardized, normalized or feature-encoded to obtain a digital vector; For text information, medical terminology keywords are set, and word embedding is used to learn the distributed representation of keywords in medical diagnosis texts, mapping each keyword to a vector space. These vectors can reflect the semantic distance relationship between different keywords.

[0008] In one embodiment, the digital information, text information and image information are respectively vectorized to obtain corresponding digital vectors, text vectors and image vectors, including: For image information, the medical image is input into the convolutional neural network, processed by multiple convolutional layers and pooling layers, and the convolutional layer is used to automatically learn the features in the image. After full connection processing, the output of the fully connected layer is used as the vector representation of the image information.

[0009] In one embodiment, after the processing of multiple convolutional layers and pooling layers and before the full connection processing, the following is further included: Perform PCA processing to find the direction with the largest variance in the data and project the data into a low-dimensional space.

[0010] In one embodiment, the step of fusing the digital vector, the text vector and the image vector includes: Inputting the digital vector and the text vector into a first bimodal neural network, performing feature fusion and interactive learning of the digital vector and the text vector in the first bimodal neural network, and obtaining a first comprehensive vector representation; The image vector and the text vector are input into a second bimodal neural network, and in the second bimodal neural network, feature fusion and interactive learning of the image vector and the text vector are performed to obtain a second comprehensive vector representation; Inputting the first integrated vector representation and the second integrated vector representation into a third bimodal neural network, and performing feature fusion and interactive learning of the first integrated vector representation and the second integrated vector representation in the third bimodal neural network to obtain a third integrated vector representation; The third comprehensive vector representation is taken as a feature fusion result of the digital vector, the text vector and the image vector.

[0011] In one embodiment, the optimal treatment plan includes at least one or more of the treatment plan with the best treatment effect, the treatment plan with the lowest treatment cost, the treatment plan with the best treatment effect within a preset treatment cost range, and the treatment plan with the lowest treatment cost within a preset treatment effect range.

[0012] In one embodiment, the calculating the similarity between the first fused vector and each second fused vector includes: Calculate the modulus lengths of the first fused vector and the second fused vector respectively; Calculate the product of the modulus of the first fused vector and the modulus of the second fused vector as a first product; calculating a product of the first fused vector and the second fused vector as a second product; Calculate the ratio of the second product to the first product. The closer the ratio is to 1, the higher the similarity. In one embodiment, obtaining the best treatment plan based on the preset number of second fusion vectors with the highest similarity includes: For each second fusion vector, obtain its historical treatment plan, and vectorize the historical treatment plan; Each second fusion vector and its vectorized treatment plan are used as the input of the Hua Tuo model, the objective function is set, the Hua Tuo model is trained, and the optimized Hua Tuo model is obtained; The first fused vector is used as the input of the optimized Hua Tuo model to obtain the best treatment plan.

[0013] In a second aspect, an embodiment provides a computer-readable storage medium, in which a program is stored, and the program can be loaded by a processor to execute the patient diagnosis and treatment method based on the vector database described in any one of the above embodiments.

[0014] The beneficial effects of the present invention are: Since the first fusion vector and the second fusion vector are obtained by feature fusion of digital vectors, text vectors and image vectors, the deep and complex relationship between numbers, texts and images can be integrated. Since the similarity between the first fusion vector and each second fusion vector is calculated, the second fusion vectors with the highest similarity are obtained, so that the best treatment plan can be obtained based on the second fusion vectors with the highest similarity. The sample size of the training data set, the number of input variables, the characteristics of the training model algorithm, the level of the training model personnel, and the degree of separation between the characteristics of new patients and the training model samples are reduced when constructing the mathematical model. In this way, with the help of excellent diagnosis and treatment cases in the past, the dependence on the diagnosis and treatment level of doctors can be reduced, not only can a better treatment plan be quickly obtained, but also the diagnosis and treatment level of existing doctors can be quickly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a schematic diagram of a patient diagnosis and treatment method based on a vector database according to an embodiment of the present application; Figure 2 This application Figure 1 A schematic flow chart of an embodiment of step S30; Figure 3 is a schematic flow chart of a method for calculating the similarity between any second fused vector and the first fused vector according to an embodiment of the present application; Figure 4 This application Figure 1 A flowchart of an embodiment of step S50 in FIG. DETAILED DESCRIPTION

[0016] The present invention is further described in detail below by specific embodiments in conjunction with the accompanying drawings. Wherein similar elements in different embodiments adopt associated similar element numbers. In the following embodiments, many detailed descriptions are for making the present application better understood. However, those skilled in the art can easily recognize that some features can be omitted in different situations, or can be replaced by other elements, materials, methods. In some cases, some operations related to the present application are not shown or described in the specification, this is to avoid the core part of the present application being overwhelmed by too much description, and for those skilled in the art, it is not necessary to describe these related operations in detail, and they can fully understand the related operations according to the description in the specification and the general technical knowledge in the art.

[0017] In addition, the features, operations or characteristics described in the specification can be combined in any appropriate manner to form various implementations. At the same time, the steps or actions in the method description can also be interchanged or adjusted in a manner that is obvious to those skilled in the art. Therefore, the various sequences in the specification and the drawings are only for the purpose of clearly describing a certain embodiment and are not meant to be a required sequence, unless otherwise specified that a certain sequence must be followed.

[0018] The serial numbers assigned to the components in this article, such as "first", "second", etc., are only used to distinguish the objects described and do not have any order or technical meaning.

[0019] In order to explain this scheme more clearly, the current patient diagnosis and treatment technology is first explained below.

[0020] The current AI-assisted diagnosis and treatment models are mostly based on traditional AI technology to build auxiliary diagnosis models for deterministic medical problems, find the mapping relationship between data sets and prediction labels, and are not widely adaptable. In addition, the applicant found in the study that, on the one hand, although the training data samples used by the current AI-assisted diagnosis and treatment software include digital, text and image data, they are all based on digital, text and image data for separate training models. When used, they only integrate the model results of each modality. This model makes it difficult to discover the deep and complex relationship between numbers, text and images. On the other hand, the accuracy of the current AI-assisted diagnosis and treatment software is heavily dependent on the sample size of the training data set for building the model, the number of input variables, the characteristics of the training model algorithm, the level of the training model personnel, and the degree of separation between the characteristics of new patients and the training model samples. These factors have led to the current auxiliary diagnosis and treatment methods being not adaptable and difficult to achieve the expected technical effects.

[0021] In view of this, a patient diagnosis and treatment method and medium based on a vector database are provided in the embodiment of the present application. In the method, the medical data of historical patients can be vectorized based on any successful diagnosis and treatment case to obtain a second fusion vector, thereby obtaining a vector database containing multiple second fusion vectors, collecting the medical data of the current patient, and vectorizing the medical data of the current patient to obtain a first fusion vector. Since the first fusion vector and the second fusion vector are both obtained by feature fusion of digital vectors, text vectors and image vectors, the deep and complex relationship between numbers, texts and images can be fused. Since the similarity between the first fusion vector and each second fusion vector is calculated, the second fusion vectors with the highest similarity are obtained, so that the best treatment plan can be obtained based on the second fusion vectors with the highest similarity. The sample size of the training data set, the number of input variables, the characteristics of the training model algorithm, the level of the training model personnel, and the degree of separation between the characteristics of new patients and the training model samples are reduced when constructing a mathematical model. In this way, we can use the help of previous excellent diagnosis and treatment cases to reduce dependence on the doctor's diagnosis and treatment level. We can not only quickly get a better treatment plan, but also quickly improve the diagnosis and treatment level of existing doctors.

[0022] An embodiment of the present application provides a patient diagnosis and treatment method based on a vector database, please refer to Figure 1 ,include: Step S10, collecting patient medical data, wherein the medical data includes digital information, text information and image information.

[0023] Medical data may include electronic medical record data, test data, doctor's order data, diagnosis and treatment data, drug data, surgery data, hospitalization data, outpatient data, and follow-up data, etc. These data are classified into digital information, text information, and image information.

[0024] Step S20, respectively vectorize the digital information, text information and image information to obtain corresponding digital vectors, text vectors and image vectors.

[0025] In one embodiment, for digital information, the numerical data is standardized, normalized or feature-encoded to obtain a digital vector.

[0026] Medical data includes a large amount of digital information, including numerical data from medical tests, the patient's age, and the stage of the disease.

[0027] Numerical data in medical tests, such as vital signs (blood pressure, heart rate, and body temperature, etc.) and laboratory test results (blood sugar, blood lipids, and blood routine tests, etc.), can be standardized. In one embodiment, a Z-score standardization method can be used for standardization, which can be specifically expressed as: ; in, x Represents the original data, represents the standard deviation of the original data, represents the mean of the original data, z Represents the standardized result.

[0028] Through standardization, the data is transformed into a distribution with a mean of 0 and a standard deviation of 1, so that data of different magnitudes can be compared on the same scale.

[0029] In one embodiment, the numerical data may also be normalized to map the data to the interval [0, 1].

[0030] Simple numerical values ​​can be directly used as a dimension of a vector. For example, the patient's age data can be used as an element of the vector. For categorical numerical data, such as disease stages (stage I, stage II, stage III, etc.), one-hot encoding can be used. Assuming there are three disease stages, stage I can be encoded as [1,0,0], stage II can be encoded as [0,1,0], and stage III can be encoded as [0,0,1].

[0031] However, for text information, such as a doctor's medical diagnosis information, word embedding can be used to convert text information into text vectors. For example, in the medical text "The patient underwent heart bypass surgery", for the central word "bypass", the context is "the patient underwent heart" and "surgery". Through a large amount of text training, the distribution representation of keywords can be learned. Therefore, medical terminology keywords can be set, and word embedding can be used to learn the distributed representation of keywords in medical diagnosis texts, map each keyword to a vector space, and thus obtain text vectors. These vectors can reflect the semantic distance relationship between different keywords. For example, the vectors of "heart disease" and "myocardial infarction" may be close in space because they are semantically related.

[0032] For medical images, such as X-rays, CT and MRI, the most common vector conversion method is to extract the grayscale features, texture features or shape features of the image. For example, the gray-level co-occurrence matrix (GLCM) can be used to extract texture features. By calculating the frequency of occurrence of pixel pairs with different grayscale values ​​in a specific direction and distance in the image, features such as contrast, correlation, energy and homogeneity are obtained. These features can form a vector.

[0033] In one embodiment of the present application, a new method for converting image information into an image vector is provided, including: inputting a medical image into a convolutional neural network, processing through multiple convolutional layers and pooling layers, using the convolutional layer to automatically learn features in the image, and then processing through full connection, using the output of the fully connected layer as a vector representation of the image information. The CNN convolutional layer can be used to automatically learn features in the image. For example, when identifying a tumor in a medical image, the convolutional layer can learn features such as the shape and edge of the tumor. After multiple convolutional layers and pooling layers, the output of the fully connected layer can finally be used as a vector representation of the image.

[0034] The above-mentioned results after feature extraction may still have a high dimension, which is not convenient for subsequent data analysis and processing. In view of this, in one embodiment of the present application, after the processing of multiple convolutional layers and pooling layers, and before the full connection processing, PCA (principal component analysis) processing is performed to find the direction with the largest variance in the data and project the data into a low-dimensional space. For example, reducing the high-dimensional medical image feature vector from hundreds of dimensions to dozens of dimensions not only reduces the storage space of the data, but also facilitates subsequent data analysis and processing.

[0035] Step S30, performing feature fusion on the digital vector, the text vector and the image vector to obtain a first fused vector.

[0036] For the feature fusion of digital vectors, text vectors and image vectors, a simple splicing method can be used. However, the applicant found in the research that this simple splicing method may be affected by the difference in the dimensions of vectors of different data types.

[0037] In view of this, please refer to Figure 2 In one embodiment of the present application, the feature fusion of digital vectors, text vectors and image vectors includes: Step S301, inputting the digital vector and the text vector into the first bimodal neural network, performing feature fusion and interactive learning of the digital vector and the text vector in the first bimodal neural network, and obtaining a first comprehensive vector representation.

[0038] Step S302: input the image vector and the text vector into a second bimodal neural network, perform feature fusion and interactive learning of the image vector and the text vector in the second bimodal neural network, and obtain a second comprehensive vector representation.

[0039] Step S303: input the first comprehensive vector representation and the second comprehensive vector representation into a third bimodal neural network, and perform feature fusion and interactive learning of the first comprehensive vector representation and the second comprehensive vector representation in the third bimodal neural network to obtain a third comprehensive vector representation.

[0040] Step S304: The third comprehensive vector representation is used as a feature fusion result of the digital vector, the text vector and the image vector.

[0041] Based on the above method, the association between different modal data can be better utilized, the quality of vector representation can be improved, and it can be applied to more complex medical data analysis tasks.

[0042] Step S40, calculating the similarity between the first fused vector and each second fused vector, and obtaining a preset number of second fused vectors with the highest similarity.

[0043] The method for obtaining the second fusion vector includes: obtaining medical data of historical patients, obtaining corresponding digital vectors, text vectors and image vectors for the medical data of any historical patient, and performing feature fusion on the digital vectors, text vectors and image vectors to obtain the second fusion vector corresponding to the any historical patient. The process of obtaining the second fusion vector can refer to the process of obtaining the first fusion vector.

[0044] In one embodiment, the preset number is five, and the similarity between the first fused vector and each second fused vector is calculated to obtain the top five second fused vectors with the highest similarity.

[0045] Those skilled in the art can understand that, for the second fusion vector, in order to obtain a better treatment plan, more excellent medical record data should be used to obtain each second fusion vector, so that an excellent treatment plan can be recommended. On the one hand, the treatment efficiency of patients can be improved, and on the other hand, the learning efficiency of doctors can be promoted, and the diagnosis and treatment level of doctors can be quickly improved under a large number of excellent treatment plans.

[0046] For the similarity calculation between two vectors, the current method can be implemented based on methods such as Euclidean distance, cosine similarity, Pearson correlation coefficient and Manhattan distance. In one embodiment of the present application, a new vector similarity calculation method is provided. For any second fused vector, the method for calculating its similarity with the first fused vector is referred to as Figure 3 ,include: Step S401, respectively calculating the module lengths of the first fused vector and the second fused vector.

[0047] In one embodiment, step S401 can be expressed as: ; ; in, A represents the first fusion vector, B represents the second fused vector, represents the modulus of the first fused vector, represents the modulus of the second fused vector, represents the first fused vector i element values, represents the first i element values, n Represents the second fused vector or the total number of elements in the second fused vector, 1≤ i ≤ n .

[0048] Those skilled in the art can understand that in order to compare the similarity between the first fused vector and the second fused vector, the first fused vector and the second fused vector have the same form and the same number of elements.

[0049] Step S402: Calculate the product of the modulus of the first fused vector and the modulus of the second fused vector as a first product.

[0050] Step S403: Calculate the product of the first fused vector and the second fused vector as the second product.

[0051] Step S404, calculating the ratio of the second product to the first product, the closer the ratio is to 1, the higher the similarity is.

[0052] In one embodiment, step S404 can be expressed as: ; in, Represents the similarity value.

[0053] For the obtained similarity value, when the value is 1, it is considered that the two vectors are exactly the same, when the value is -1, it is considered that the two vectors are completely opposite, and when the value is 0, it means that the two vectors are orthogonal. The closer to 1, the more similar the symptoms of the two patients are.

[0054] Step S50, obtaining an optimal treatment plan based on a preset number of second fusion vectors with the highest similarity.

[0055] In one embodiment, based on the first five second fusion vectors with the highest similarity obtained, the treatment plan, treatment effect and treatment cost corresponding to the five second fusion vectors can be obtained based on the database. Therefore, based on the needs, the treatment plan with the best treatment effect can be obtained as the best treatment plan; the treatment plan with the lowest treatment cost can also be obtained as the best treatment plan; it can also be the treatment plan with the best treatment effect within a preset treatment cost range, for example, within an affordable treatment cost, the treatment plan with the best treatment effect can be obtained as the best treatment plan; it can also be the treatment plan with the lowest treatment cost within a preset treatment effect range, for example, in terms of achieving the most basic treatment effect requirements, the treatment plan with the lowest treatment cost can be obtained as the best treatment plan.

[0056] Based on the scheme of the above embodiment, since the first fusion vector and the second fusion vector are obtained by feature fusion of digital vectors, text vectors and image vectors, the deep and complex relationship between numbers, texts and images can be fused. Since the similarity between the first fusion vector and each second fusion vector is calculated, the preset number of second fusion vectors with the highest similarity is obtained, so that the best treatment plan can be obtained based on the preset number of second fusion vectors with the highest similarity, which reduces the reliance on factors such as the sample size of the training data set, the number of input variables, the characteristics of the training model algorithm, the level of the training model personnel, and the degree of separation between the characteristics of new patients and the training model samples when constructing the mathematical model. In this way, with the help of previous excellent diagnosis and treatment cases, the reliance on the diagnosis and treatment level of doctors can be reduced, not only can a better treatment plan be quickly obtained, but also the diagnosis and treatment level of existing doctors can be quickly improved.

[0057] As an existing medical consultation model, the Hua Tuo model can take the patient's symptoms, medical history, examination reports and other information as input, and then the model will combine the input information with its own medical knowledge and logic to generate corresponding consultation results, including condition analysis, possible disease diagnosis and treatment suggestions. However, the current treatment suggestions obtained by the Hua Tuo model are limited to directional suggestions, and detailed treatment plans still need to be formulated based on the doctor's experience and the patient's condition, which still poses a great challenge to the doctor's professional level.

[0058] In view of this, in one embodiment of the present application, please refer to Figure 4 , a method for obtaining the best treatment plan based on a preset number of second fusion vectors with the highest similarity obtained in this application and combined with the Hua Tuo model is proposed, including: Step S501: for each second fusion vector, obtain its historical treatment plan, and vectorize the historical treatment plan.

[0059] Historical treatment plans may include information such as drug treatment (e.g. drug name, dosage, course of treatment, etc.), surgical treatment (e.g. type of surgery and time of surgery, etc.) and physical therapy (e.g. method and frequency of physical therapy, etc.). Historical treatment plans may also include treatment effect evaluation, such as symptom improvement (e.g. complete relief, partial relief, no change and deterioration, etc.), changes in the patient's quality of life and the occurrence of complications.

[0060] Step S502, taking each second fusion vector and the treatment plan after vectorization as the input of the Hua Tuo model, setting the objective function, training the Hua Tuo model, and obtaining the optimized Hua Tuo model.

[0061] Training is performed based on the collected similar medical records. During the training process, a suitable objective function should be set. The objective function can be set based on demand. For example, the goal can be to improve treatment efficiency or reduce the incidence of complications. During training, supervised learning methods can be used to use historical treatment plans and effects as labels to allow the Hua Tuo model to learn the relationship between different treatment plans and treatment effects.

[0062] Those skilled in the art can understand that, in order to be used with the Hua Tuo model, the form of the first fusion vector, the second fusion vector and the vectorized treatment plan can be the vector form required by the input vector of the Hua Tuo model. If not, they can be converted into the vector form required by the input vector of the Hua Tuo model and then used as the input of the Hua Tuo model.

[0063] Step S503: Use the first fusion vector as input of the optimized Hua Tuo model to obtain the best treatment plan.

[0064] The optimized large model can generate one or more potential more suitable treatment plans based on the input current patient conditions or historical treatment plans of similar medical records through the internal neural network architecture, taking into account various factors such as the similarity of symptoms, individual differences of patients, feasibility of treatment plans, etc. These plans can be more accurate in drug selection (taking into account drug tolerance, drug interactions, etc.), and more optimized in treatment time and intensity (avoiding overtreatment or undertreatment).

[0065] In some embodiments, the best treatment plan is obtained based on the optimized Hua Tuo model, and the medical team (including doctors, pharmacists, etc.) can also judge based on medical knowledge and clinical experience. For example, check whether the drug dosage is within the safe range and whether the surgical plan is in line with the current level of medical technology and the patient's physical condition. The cost-benefit ratio of the treatment plan can also be considered, including factors such as medical expenses, patient recovery time, and occupation of social resources. According to the evaluation results, the most appropriate treatment plan is selected. If there are some unreasonable aspects of the plan output by the model, it can be adjusted according to medical expertise. For example, adjust the type or dosage of the drug, modify the order of treatment, etc. In the process of implementing the treatment plan, it is necessary to continuously monitor the patient's response so that the plan can be further optimized in a timely manner.

[0066] Based on the optimized diagnosis and treatment plan based on the Hua Tuo model in the above embodiment, on the one hand, it makes up for the defects of the existing Hua Tuo model, and on the other hand, it improves the adaptability of the patient's diagnosis and treatment plan, which can help obtain a better treatment plan and improve the diagnosis and treatment level of existing doctors more quickly.

[0067] In one embodiment of the present application, a computer-readable storage medium is provided, on which a program is stored. The stored program includes a method that can be loaded by a processor and process any of the above embodiments.

[0068] Those skilled in the art will appreciate that all or part of the functions of the various methods in the above-mentioned embodiments can be implemented by hardware or by computer programs. When all or part of the functions in the above-mentioned embodiments are implemented by computer programs, the program can be stored in a computer-readable storage medium, and the storage medium can include: read-only memory, random access memory, disk, optical disk, hard disk, etc., and the program is executed by a computer to implement the above-mentioned functions. For example, the program is stored in the memory of the device, and when the program in the memory is executed by the processor, all or part of the above-mentioned functions can be implemented. In addition, when all or part of the functions in the above-mentioned embodiments are implemented by computer programs, the program can also be stored in a storage medium such as a server, another computer, disk, optical disk, flash disk or mobile hard disk, and can be downloaded or copied and saved in the memory of the local device, or the system of the local device is updated, and when the program in the memory is executed by the processor, all or part of the functions in the above-mentioned embodiments can be implemented.

[0069] The above specific examples are used to illustrate the present invention, which is only used to help understand the present invention and is not intended to limit the present invention. For those skilled in the art, according to the concept of the present invention, some simple deductions, modifications or substitutions can be made.

Claims

1. A patient diagnosis and treatment method based on a vector database, characterized in that: include: Collecting patient medical data, including digital information, text information and image information; Vectorizing the digital information, text information and image information respectively to obtain corresponding digital vectors, text vectors and image vectors; Performing feature fusion on the digital vector, the text vector and the image vector to obtain a first fused vector; Calculating the similarity between the first fused vector and each second fused vector to obtain a preset number of second fused vectors with the highest similarity; Obtaining an optimal treatment plan based on a preset number of second fusion vectors with the highest similarity; The method for obtaining the second fusion vector includes: obtaining medical data of historical patients, obtaining corresponding digital vectors, text vectors and image vectors for the medical data of any historical patient, and performing feature fusion on the digital vectors, text vectors and image vectors to obtain the second fusion vector corresponding to the any historical patient.

2. The patient diagnosis and treatment method based on the vector database as claimed in claim 1, characterized in that: The digital information, text information and image information are respectively vectorized to obtain corresponding digital vectors, text vectors and image vectors, including: For digital information, the numerical data is standardized, normalized or feature-encoded to obtain a digital vector; For text information, medical terminology keywords are set, and word embedding is used to learn the distributed representation of keywords in medical diagnosis texts, mapping each keyword to a vector space. These vectors can reflect the semantic distance relationship between different keywords.

3. The patient diagnosis and treatment method based on the vector database as claimed in claim 1, characterized in that: The digital information, text information and image information are respectively vectorized to obtain corresponding digital vectors, text vectors and image vectors, including: For image information, the medical image is input into the convolutional neural network, processed by multiple convolutional layers and pooling layers, and the convolutional layer is used to automatically learn the features in the image. After full connection processing, the output of the fully connected layer is used as the vector representation of the image information.

4. The patient diagnosis and treatment method based on the vector database as claimed in claim 3, characterized in that: After multiple convolutional layers and pooling layers, and before full connection processing, it also includes: Perform PCA processing to find the direction with the largest variance in the data and project the data into a low-dimensional space.

5. The patient diagnosis and treatment method based on vector database according to claim 1, characterized in that: The step of fusing the digital vector, text vector and image vector with features includes: Inputting the digital vector and the text vector into a first bimodal neural network, performing feature fusion and interactive learning of the digital vector and the text vector in the first bimodal neural network, and obtaining a first comprehensive vector representation; The image vector and the text vector are input into a second bimodal neural network, and in the second bimodal neural network, feature fusion and interactive learning of the image vector and the text vector are performed to obtain a second comprehensive vector representation; Inputting the first integrated vector representation and the second integrated vector representation into a third bimodal neural network, and performing feature fusion and interactive learning of the first integrated vector representation and the second integrated vector representation in the third bimodal neural network to obtain a third integrated vector representation; The third comprehensive vector representation is taken as a feature fusion result of the digital vector, the text vector and the image vector.

6. The patient diagnosis and treatment method based on vector database according to claim 1, characterized in that: The calculating the similarity between the first fusion vector and each second fusion vector to obtain a preset number of second fusion vectors with the highest similarity includes: calculating the similarity between the first fusion vector and each second fusion vector to obtain the first five second fusion vectors with the highest similarity.

7. The patient diagnosis and treatment method based on vector database according to claim 1, characterized in that: The optimal treatment plan includes at least one or more of the treatment plan with the best treatment effect, the treatment plan with the lowest treatment cost, the treatment plan with the best treatment effect within a preset treatment cost range, and the treatment plan with the lowest treatment cost within a preset treatment effect range.

8. The patient diagnosis and treatment method based on vector database according to claim 1, characterized in that: The calculating the similarity between the first fused vector and each second fused vector includes: Calculate the modulus lengths of the first fused vector and the second fused vector respectively; Calculate the product of the modulus of the first fused vector and the modulus of the second fused vector as a first product; calculating a product of the first fused vector and the second fused vector as a second product; The ratio of the second product to the first product is calculated. The closer the ratio is to 1, the higher the similarity.

9. The patient diagnosis and treatment method based on vector database according to claim 1, characterized in that: The obtaining of the best treatment plan based on the preset number of second fusion vectors with the highest similarity comprises: For each second fusion vector, obtain its historical treatment plan, and vectorize the historical treatment plan; Each second fusion vector and its vectorized treatment plan are used as the input of the Hua Tuo model, the objective function is set, the Hua Tuo model is trained, and the optimized Hua Tuo model is obtained; The first fused vector is used as the input of the optimized Hua Tuo model to obtain the best treatment plan.

10. A computer-readable storage medium, characterized in that: The medium stores a program, which can be loaded by a processor and execute the patient diagnosis and treatment method based on a vector database as described in any one of claims 1 to 9.

Images