A medical diagnosis large model training method

Abstract

The application relates to the field of disease diagnosis, and discloses a medical diagnosis large model training method, which comprises the following steps: firstly, collecting the disease symptoms of historical patients, the diagnosis results of the historical patients, different age stages of the patients, physical conditions of the patients, and corresponding treatment schemes; establishing multiple disease symptom tree diagrams of different levels according to the different age stages and the physical conditions of the patients, and taking different sensory parts as the center points of the disease symptom tree diagrams; counting the disease symptoms of the historical patients and the treatment schemes of different disease symptoms in the corresponding treatment schemes, comparing and screening the treatment schemes of the same disease symptoms through a large language model to obtain optimal treatment schemes; and binding the optimal treatment schemes according to the disease symptom tree diagrams to complete the training of the disease symptom tree diagrams; and the application solves the problems in the prior art that it is slow to explore the causes, easy to misdiagnose, difficult for doctors to consider multiple cause treatment methods, and pathological knowledge is scarce in remote areas.

Description

Technical Field

[0001] This invention relates to the field of disease diagnosis, and more particularly to a method for training a large medical diagnostic model. Background Technology

[0002] With the increasing population, various diseases are becoming more frequent. People often suffer from various illnesses. However, due to the large population and limited medical resources, the medical level in many remote areas is not high. In some remote areas, doctors lack medical knowledge, which sometimes leads to the use of the wrong medication or misdiagnosis of the cause of the disease. As a result, patients do not receive timely treatment, which delays treatment time and may even cause secondary harm to patients.

[0003] Currently, doctors in various regions learn about medical treatment through medical books or videos. However, the data and videos are relatively general and do not take into account the rejection of treatment plans when patients have multiple diseases or the different medication situations that arise from similar symptoms. Therefore, there is an urgent need for a method that can consider multiple causes of patients, record multiple scientific treatment methods, and take into account the rejection of treatment methods. Summary of the Invention

[0004] The present invention aims to provide a method for training a large medical diagnostic model to solve the problems of slow disease etiology investigation, easy misdiagnosis, difficulty for doctors to consider the rejection of multiple etiologies and treatment methods, and lack of pathological knowledge in remote areas in the existing technology.

[0005] To achieve the above objectives, the present invention provides the following method:

[0006] The present invention provides a method for training a large medical diagnostic model:

[0007] S1: Collect information on the patient's symptoms, diagnosis results, age, physical condition, and corresponding treatment plan.

[0008] S2: Based on the different age stages and physical conditions of the patients, establish multiple different levels of disease symptom tree diagrams, with different sensory parts as the center points of the disease symptom tree diagrams;

[0009] S3: Statistically analyze the symptoms of the patients in the history and the treatment plans for different symptoms in the corresponding treatment plans. Use a large language model to compare and filter the treatment plans for the same symptoms to obtain the optimal treatment plan.

[0010] S4: Bind the optimal treatment plan according to the disease symptom tree diagram to complete the training of the disease symptom tree diagram.

[0011] Preferably, the collected information on the patient's symptoms, age, physical condition, and corresponding treatment plan includes: the patient's symptoms, duration of illness, age, physical condition, treatment cycle, daily diet, and differences in treatment plans for patients with the same illness.

[0012] Preferably, the step of establishing multiple hierarchical symptom tree diagrams based on the patient's different age stages and physical condition, with different sensory organs as the center points of the symptom tree diagrams, includes: dividing the patient into four age stages: infancy, childhood, adolescence, and old age; dividing the patient into three physical condition stages: good physical condition, average physical condition, and poor physical condition; establishing twelve-level symptom tree diagrams based on the four age stages and three physical condition stages, wherein each symptom tree diagram includes a core point, multiple center points, and multiple levels of branch points extending from the center points; and expanding the symptom tree diagrams by branching out from different sensory organs as the center points.

[0013] Preferably, the step of extending the disease condition tree diagram by branching out different sensory sites as the center points of the disease symptom tree diagram includes: extending the disease condition tree diagram by branching out different sensory sites as the center points of the disease symptom tree diagram, with each sensory site serving as a center point of the disease symptom tree diagram, and the human body serving as the core point of the disease symptom tree diagram; the next level branch points of the disease symptom tree diagram extending from the center point of the disease symptom tree diagram are more specific sensory site symptoms; each center point of the disease symptom tree diagram is in a parallel relationship, and the next level of the extension of each center point of the disease symptom tree diagram is a diffuse extension, with the number of branch points in the next level of the disease symptom tree diagram being greater than or equal to the number of branch points in the previous level of the disease symptom tree diagram.

[0014] Preferably, each branch point of the symptom tree extending from the center point of the symptom tree is marked as a branch point index; the branch point index is bound to the diagnosis results of the historical patients with corresponding sensory symptoms, and the diagnosis results of the historical patients are repeatedly judged based on the branch route of the symptom tree; the branch route of the symptom tree includes the core point of the symptom tree, a center point of the symptom tree, and one branch point at each level of the symptom tree.

[0015] Preferably, the step of binding the branch point index to the diagnosis results of the historical patients corresponding to the sensory site symptoms, and performing a duplicate judgment on the branch routes of the disease symptom tree diagram for the diagnosis results of the historical patients, includes: if the diagnosis results of the historical patients bound to the branch point index are the same as the diagnosis results of the historical patients in the other branch point indexes of each branch route of the disease symptom tree diagram where the branch point is located, then the diagnosis results of the historical patients in the other branch point indexes are hidden, so that each branch route of the disease symptom tree diagram does not have duplicate diagnosis results of the historical disease.

[0016] Preferably, after binding the branch point index to the diagnostic results of the historical patients corresponding to the sensory site symptoms, and performing a repeat judgment on the branch path of the disease symptom tree diagram of the historical patients' diagnostic results, the method further includes marking one branch path of the disease symptom tree diagram as a diagnostic process record, and judging the rationality of the diagnosis of the diagnostic process record; if the repetition rate between the diagnostic results of the historical patients and the diagnostic process record in the diagnostic process record reaches 95%, then the diagnostic process record is marked as a reasonable diagnostic process record and recorded in the disease symptoms of the historical patients.

[0017] Preferably, the step of statistically analyzing the symptoms of historical patients and the treatment plans for different symptoms in the corresponding treatment plans, and comparing and filtering the treatment plans for the same symptoms using a large language model to obtain the optimal treatment plan includes: statistically analyzing the symptoms of historical patients and the treatment plans for different symptoms in the corresponding treatment plans, as well as the patient recovery process corresponding to the treatment plans for different symptoms; comparing the similarity among the treatment plans for different symptoms, and encapsulating the treatment plan content for different symptoms with a similarity of 100%; comparing the content of the treatment plans for different symptoms with a similarity of more than 60% and the patient recovery process corresponding to the treatment plans for different symptoms, and if the recovery process is effective and the recovery time is fast, then the content of the treatment plan for different symptoms is encapsulated; encapsulating the content of the treatment plan for different symptoms with good recovery process or fast recovery time, and adding a treatment index corresponding to the applicable conditions of good recovery process and fast recovery time; and judging the symptom rejection status of the encapsulated treatment plan content for different symptoms.

[0018] Preferably, the step of judging the symptom rejection status of the encapsulated treatment plan content for the different disease symptoms includes: if there are mutually exclusive or shared treatment plan content that can cause harm to the patient among the treatment plan content for the different disease symptoms, then the corresponding treatment plan content is bound with a non-coexistence tag; if the treatment plan content for the different disease symptoms has the same non-coexistence tag, then a rejection warning is issued, and the treatment plan content cannot coexist.

[0019] Preferably, the step of binding the optimal treatment plan to the symptom tree diagram to complete the training of the symptom tree diagram includes: binding the symptom tree diagram to the treatment plan content of different symptom symptoms; integrating all the treatment plan content of different symptom symptoms on the branch path of the symptom tree diagram, and if a rejection warning is issued, replacing the treatment plan with the same treatment effect to obtain the optimal treatment plan; storing the optimal treatment plan by module according to the branch point of the symptom tree diagram, and updating the optimal treatment plan in real time; encapsulating the optimal treatment plan into the final treatment plan point, and connecting the optimal treatment plan point to the last level branch point of the corresponding symptom tree diagram branch path to complete the training of the symptom tree diagram.

[0020] The beneficial effects of this invention are reflected in the following: By establishing a multi-level tree diagram of disease symptoms based on the patient's different age stages and physical conditions, this invention enables the etiology investigation process and treatment plan to be differentiated for different age groups and physical conditions, making it more adaptable to patients with various conditions. Specifically, the etiology investigation process and treatment plan are an extended tree diagram, which can take into account the combination of multiple different etiologies and identify etiologies with high similarity, thereby planning a more targeted and comprehensive treatment process. Furthermore, the addition of real-time updated plans allows for the adoption of the same plan in various regions based on historical treatment plans, avoiding the problem of insufficient medical knowledge among doctors in some areas. Attached Figure Description

[0021] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0022] Figure 1 This is a flowchart illustrating a large-scale medical diagnostic model training method provided in an embodiment of the present invention.

[0023] Figure 2This is a schematic diagram of the disease symptom tree structure provided in an embodiment of the present invention. Detailed Implementation

[0024] To enable those skilled in the art to better understand the present invention, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or end that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or ends.

[0026] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0027] Currently, doctors in various regions learn about medical treatment through medical books or videos. However, the data and videos are relatively general and do not take into account the rejection of treatment plans when patients have multiple diseases or the different medication situations that arise from similar symptoms. Therefore, there is an urgent need for a method that can consider multiple causes of patients, record multiple scientific treatment methods, and take into account the rejection of treatment methods.

[0028] The present invention aims to provide a method for training a large medical diagnostic model to solve the problems of slow disease etiology investigation, easy misdiagnosis, difficulty for doctors to consider the rejection of multiple etiologies and treatment methods, and lack of pathological knowledge in remote areas in the existing technology.

[0029] This invention provides a method for training a large medical diagnostic model, as follows: Figure 1 and Figure 2 As shown, it includes the following steps:

[0030] S1: Collect information on the patient's symptoms, diagnosis, age, physical condition, and corresponding treatment plan.

[0031] In this embodiment of the invention, the collected historical patient symptoms, different age groups, physical conditions and corresponding treatment plans include: historical patient symptoms, duration of illness, age, physical condition, treatment cycle, daily diet, and differences in treatment plans for patients with the same condition.

[0032] S2: Establish multiple levels of disease symptom tree diagrams based on the patient's different age stages and physical conditions, with different sensory sites as the center points of the disease symptom tree diagrams.

[0033] In this embodiment of the invention, the symptoms of different sensory sites of the patient are used as the central point of the symptom tree diagram. The branches of the symptom tree diagram are extended according to the specific symptoms of the patient. For example, if the throat is uncomfortable, the branches are (1) environmental factors, such as dry air; (2) improper diet, such as cold drinks and spicy and irritating foods; (3) acute pharyngitis, acute epiglottitis, and acute tonsillitis caused by bacterial or viral infection; (4) gastroesophageal reflux, where gastric acid irritates the throat, causing sore throat. Each branch is a data point, and the data point includes the symptom information of that branch. A diagnosis process is recorded as a branch route of the symptom tree diagram, which requires a second judgment. For the rationality assessment, the doctor needs to determine the rationality. If the rationality is high, a diagnostic process is completed, and the process is recorded in the patient's history of symptoms and corresponding treatment plans. Patients are categorized into four age groups: infancy, childhood, adolescence, and old age. Based on their physical condition, patients are categorized into three levels: good, average, and poor. A twelve-level symptom tree diagram is constructed based on the four age groups and three physical condition levels. This symptom tree diagram includes one core point, multiple central points, and extensions of these central points. The process involves creating a multi-level tree diagram of disease symptoms, branching points, and extending the tree diagram of disease conditions using different sensory organs as the central points. This extension includes: branching the tree diagram of disease conditions using different sensory organs as the central points, with each sensory organ serving as a central point and the human body as the core point; the next level of branching points in the disease symptoms tree diagram from the central points represent more specific sensory organ symptoms; and each central point of the disease symptoms tree diagram... The points are in a parallel relationship. The next level of branches extending from the center point of each symptom tree diagram is a diffuse extension, and the number of branch points in the next level of the symptom tree diagram is greater than or equal to the number of branch points in the previous level of the symptom tree diagram. Each branch point extending from the center point of the symptom tree diagram is marked as a branch point index. The branch point index is bound to the diagnosis results of historical patients with symptoms of the corresponding sensory parts, and the diagnosis results of historical patients are re-evaluated for the branch routes of the symptom tree diagram. The branch route of the symptom tree diagram includes the core point of the symptom tree diagram, a center point of the symptom tree diagram, and one branch point of each level of the symptom tree diagram.The process of binding branch point indices to the diagnostic results of historical patients with symptoms corresponding to sensory sites, and performing a duplicate judgment on the branch routes of the symptom tree diagram for the historical patients' diagnostic results, includes: if the diagnostic results of historical patients bound to branch point indices are the same as the diagnostic results of historical patients with other branch point indices in each branch route of the symptom tree diagram where the branch point is located, then the diagnostic results of the historical patients with the other branch point indices are hidden, so that each branch route of the symptom tree diagram has no duplicate diagnostic results of historical patients; after binding branch point indices to the diagnostic results of historical patients with symptoms corresponding to sensory sites, and performing a duplicate judgment on the branch routes of the symptom tree diagram for the historical patients' diagnostic results, the process further includes: marking a branch route of the symptom tree diagram as a diagnostic process record, and judging the rationality of the diagnosis of the diagnostic process record; if the duplication rate between the diagnostic results of historical patients and the diagnostic process records in the diagnostic process records reaches 95%, then the diagnostic process record is marked as a reasonable diagnostic process record and recorded in the historical patients' symptoms.

[0034] S3: Collect statistics on the symptoms of patients in history and the corresponding treatment plans for different symptoms. Use a large language model to compare and select treatment plans for the same symptoms to obtain the optimal treatment plan.

[0035] In this embodiment of the invention, the symptoms of historical patients and the corresponding treatment plans for different symptoms are statistically analyzed, along with the patient recovery process corresponding to each treatment plan. The similarity of treatment plans for different symptoms is compared, and treatment plans with 100% similarity are encapsulated. For treatment plans with more than 60% similarity, the patient recovery process corresponding to each treatment plan is compared. If the recovery process is effective and the recovery time is fast, the treatment plan content for each symptom is encapsulated. The treatment plans for different symptoms of the disease are encapsulated, and a treatment index is added to indicate the applicable conditions for good recovery process and fast recovery time. The encapsulated treatment plans for different symptoms are then assessed for symptom rejection. The steps for assessing symptom rejection of the encapsulated treatment plans for different symptoms include: if there are mutually exclusive or shared treatment plans that could harm the patient among the treatment plans for different symptoms, then the corresponding treatment plans are labeled as incompatible; if the treatment plans for different symptoms have the same incompatible label, then a rejection warning is issued, and the coexistence of the treatment plans is not permitted.

[0036] S4: Bind the optimal treatment plan based on the disease symptom tree diagram to complete the training of the disease symptom tree diagram.

[0037] In this embodiment of the invention, a disease symptom tree diagram is bound to the corresponding treatment plan content for different disease symptoms; the treatment plan content for all different disease symptoms on the branch path of the disease symptom tree diagram is integrated, and if a rejection warning is issued, a treatment plan with the same treatment effect is replaced to obtain the optimal treatment plan; the optimal treatment plan is stored by module according to the branch point of the disease symptom tree diagram, and the optimal treatment plan is updated in real time; the optimal treatment plan is encapsulated into the final treatment plan point, and the optimal treatment plan point is correspondingly connected to the last level disease symptom tree diagram branch point of the corresponding disease symptom tree diagram branch path to complete the training of the disease symptom tree diagram.

[0038] The beneficial effects of this invention are reflected in the following: By establishing a multi-level tree diagram of disease symptoms based on the patient's different age stages and physical conditions, this invention enables the etiology investigation process and treatment plan to be differentiated for different age groups and physical conditions, making it more adaptable to patients with various conditions. Specifically, the etiology investigation process and treatment plan are an extended tree diagram, which can take into account the combination of multiple different etiologies and identify etiologies with high similarity, thereby planning a more targeted and comprehensive treatment process. Furthermore, the addition of real-time updated plans allows for the adoption of the same plan in various regions based on historical treatment plans, avoiding the problem of insufficient medical knowledge among doctors in some areas.

[0039] The above descriptions are merely embodiments of the present invention. Commonly known technical solutions or characteristics are not described in detail here. It should be noted that those skilled in the art can make various modifications and improvements without departing from the present invention, and these should also be considered within the scope of protection of the present invention. These modifications and improvements will not affect the effectiveness of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A method for training a large-scale medical diagnostic model, characterized in that, The method includes: S1: Collect information on the patient's symptoms, diagnosis results, age, physical condition, and corresponding treatment plan. S2: Based on the different age stages and physical conditions of the patients, establish multiple different levels of disease symptom tree diagrams, with different sensory parts as the center points of the disease symptom tree diagrams; S3: Statistically analyze the symptoms of the patients in the history and the treatment plans for different symptoms in the corresponding treatment plans. Use a large language model to compare and filter the treatment plans for the same symptoms to obtain the optimal treatment plan. S4: Bind the optimal treatment plan according to the disease symptom tree diagram to complete the training of the disease symptom tree diagram.

2. The method for training a large medical diagnostic model according to claim 1, characterized in that: The data collected includes the patient's symptoms, diagnosis, age, physical condition, and corresponding treatment plan. This includes the patient's symptoms, duration of illness, age, physical condition, treatment cycle, daily diet, and differences in treatment plans for patients with the same condition.

3. The method for training a large medical diagnostic model according to claim 1, characterized in that, The step of establishing multiple hierarchical disease symptom tree diagrams based on the patient's different age stages and physical condition, with different sensory organs as the center points of the disease symptom tree diagrams, includes: Based on the different age stages of the patients, the patients are divided into four age stages: infancy, childhood, adolescence, and old age. Based on the described physical condition, patients are divided into three levels of physical condition: good physical condition, average physical condition, and poor physical condition. Based on the four age stages and three physical condition levels, a twelve-level disease symptom tree diagram is established. The disease symptom tree diagram includes a disease symptom tree diagram core point, multiple disease symptom tree diagram center points, and multiple disease symptom tree diagram branch points extending from the multiple center points. The disease condition tree diagram expands by branching out different sensory organs as the central points of the disease symptom tree diagram.

4. The method for training a large medical diagnostic model according to claim 3, characterized in that, The step of expanding the disease condition tree diagram by branching the tree diagram with different sensory sites as the center points of the disease symptom tree diagram includes: The disease condition tree diagram extends the disease symptom tree diagram by taking different sensory parts as the center point of the disease symptom tree diagram, and each sensory part is a center point of the disease symptom tree diagram, with the human body as the core point of the disease symptom tree diagram. The next level of the disease symptom tree branch extending from the central point of the disease symptom tree represents more specific sensory symptoms. The center points of each of the disease symptom tree diagrams are in a parallel relationship, and the next level of branches extending from the center point of each of the disease symptom tree diagrams is a diffuse extension. The number of branch points in the next level of the disease symptom tree diagram is greater than or equal to the number of branch points in the previous level of the disease symptom tree diagram.

5. The method for training a large medical diagnostic model according to claim 4, characterized in that: Each branch point of the disease symptom tree diagram extending from the center point is marked as a branch point index; The branch point index is bound to the diagnosis results of the historical patients corresponding to the symptoms of the sensory parts, and the diagnosis results of the historical patients are subjected to repeated judgment of the branch route of the disease symptom tree diagram. The disease symptom tree diagram branch route includes the core point of the disease symptom tree diagram, a center point of the disease symptom tree diagram, and one branch point of the disease symptom tree diagram at each level.

6. The method for training a large medical diagnostic model according to claim 5, characterized in that, The step of binding the branch point index to the diagnostic results of the historical patients corresponding to the sensory site symptoms, and performing a repeat judgment on the branch path of the disease symptom tree diagram of the historical patients' diagnostic results, includes: If the diagnosis result of the historical patient bound to the branch point index is the same as the diagnosis result of the historical patient in the other branch point indexes of each branch route of the disease symptom tree diagram where the branch point is located, then the diagnosis results of the historical patients in the other branch point indexes are hidden, so that each branch route of the disease symptom tree diagram has no duplicate diagnosis results of the historical disease.

7. The method for training a large medical diagnostic model according to claim 5, characterized in that: The method further includes binding the branch point index to the diagnosis results of the historical patients corresponding to the symptoms of the sensory parts, and after making repeated judgments on the branch routes of the disease symptom tree diagram of the diagnosis results of the historical patients, marking one branch route of the disease symptom tree diagram as a diagnosis process record, and judging the rationality of the diagnosis on the diagnosis process record. If the diagnostic results of the historical patients recorded in the diagnostic process have a repetition rate of 95% with the diagnostic process records, then the diagnostic process records are marked as reasonable diagnostic process records and the symptoms of the historical patients are recorded.

8. The method for training a large medical diagnostic model according to claim 7, characterized in that, The step of statistically analyzing the symptoms of historical patients and corresponding treatment plans for different symptoms, and comparing and filtering treatment plans for the same symptoms using a large language model to obtain the optimal treatment plan includes: The statistics include the symptoms of the patients in the history of the disease, the treatment plans for different symptoms, and the recovery process of the patients corresponding to the treatment plans for different symptoms. By comparing the similarity among the treatment plans for the different symptoms, the treatment plans for the different symptoms that have a 100% similarity are encapsulated. Compare the treatment plans for different symptoms that have a similarity of more than 60% to the patient recovery process corresponding to the treatment plans for different symptoms. If the recovery process is effective and the recovery time is fast, then the treatment plan content for the different symptoms will be packaged. The treatment plans for different symptoms with good or fast recovery processes are packaged and an index is added to indicate the applicable conditions for good or fast recovery processes. The treatment plan content for the different disease symptoms described in the encapsulation is used to determine the symptom rejection status.

9. The method for training a large medical diagnostic model according to claim 8, characterized in that, The step of determining symptom rejection status of the encapsulated treatment plans for different disease symptoms includes: If the treatment plans for different symptoms contain mutually exclusive or shared treatment plans that could harm the patient, then the corresponding treatment plan content will be labeled as incompatible. If the treatment plans for different symptoms contain the same incompatible label, then an exclusion warning will be issued, and the treatment plan content cannot coexist.

10. The method for training a large medical diagnostic model according to claim 9, characterized in that, The step of binding the optimal treatment plan to the disease symptom tree diagram to complete the training of the disease symptom tree diagram includes: The disease symptom tree diagram is then linked to the corresponding treatment plans for the different disease symptoms; The treatment plans for all the different symptoms on the branch path of the disease symptom tree are integrated. If a rejection warning is issued, a treatment plan with the same treatment effect is replaced to obtain the optimal treatment plan. The optimal treatment plan is stored in sections according to the branch points of the disease symptom tree diagram, and the optimal treatment plan is updated in real time. The optimal treatment plan is encapsulated into a final treatment plan point, and the optimal treatment plan point is connected to the last level branch point of the corresponding symptom tree diagram branch path to complete the training of the symptom tree diagram.

Images