Diagnosis, prescription support system using large-scale language model

Abstract

To provide a diagnosis and a prescription support system: having a large-scale language model additionally learn patient records of a great amount of diagnosed cases; having the additionally learned large scale language model propose an effective inspection plan for determining diagnosis, from symptom remarks obtained from a visiting diagnosis undetermined patient; support obtaining an established diagnosis by integrating the obtained testing results; and furthermore propose a suitable treatment plan on the basis of the established analysis.SOLUTION: An electronic medical chart system also combined with a large scale language model includes: "possible diagnosis term, additional symptoms, body remarks, and test answer means" for responding to at least one of a possible diagnosis name and a list of the probability, recommended additional symptoms, body remarks, and a test; and sequential diagnosis improvement means repeating "symptoms, physical remarks, and test result data provision means" and "diagnosis name with a possibility, additional symptoms, body remarks, test answer means" until the list of the diagnosis name and its probability has sufficient accuracy as a definite diagnosis.SELECTED DRAWING: Figure 5

Description

[Technical field]

[0001] The present invention relates to a diagnosis and prescription support system that uses a large-scale language model to develop an examination plan based on a patient's symptoms and physical findings, supports the diagnosis of a disease based on the results of the examination, and further supports the development of an appropriate treatment plan. [Background technology]

[0002] Computer systems (electronic medical records) for efficiently recording medical instructions (orders) and records are becoming widely used in medical facilities. Moreover, there has been remarkable progress in artificial intelligence (AI) in recent years, with particular attention being paid to large-scale language models (LLMs), or generative AI, which are trained on large volumes of literature and data to enable inquiries and responses in natural language. Prior art documents relevant to this application include the following: [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Special Publication No. 2023-523644 [Patent Document 2] Patent Publication No. 2023-73095 [Patent Document 3] Patent No. 5484317 [Patent Document 4] Patent No. 7313757 [Patent Document 5] Patent No. 6792750 [Patent Document 6] Patent No. 7284970 [Patent Document 7] Patent No. 7145367

[0004] [Non-Patent Document 1] Inside the Mind of ChatGPT Stephen Wolfram Hayakawa Shinsho 009 Summary of the Invention [Problem to be solved by the invention]

[0005] When providing medical care, examination plans are made based on the patient's symptoms and examination findings, and treatment plans are made based on the examination results, and treatment is then implemented. Since a wide range of medical professionals, including doctors and nurses, are involved, it goes without saying that electronic medical records are useful for efficient medical activities. It is necessary to examine the patient, narrow down the disease name to a certain extent based on symptoms and physical findings, conduct tests to confirm the disease name, and then create a treatment plan based on this disease name. Here, medical instructions (orders) for tests, treatments, etc. are compiled manually by doctors. This requires a huge amount of work, which leads to doctors working long hours. In addition, differences in the knowledge and experience levels of each doctor can lead to problems such as misdiagnosis, unnecessary tests, and inconsistent treatment results.

[0006] The present invention has been made to solve such conventional problems, and its purpose is to provide a diagnosis and prescription support system that additionally trains a large-scale language model with patient records of a large number of diagnosed cases (additionally trained large-scale language model), and has the additionally trained large-scale language model propose an efficient test plan for confirming a diagnosis based on symptoms and findings obtained from undiagnosed patients who visit the hospital, and helps to obtain a definitive diagnosis by synthesizing the obtained test results, and further to propose an appropriate treatment plan based on the definitive diagnosis. [Means for solving the problem]

[0007] As a means for achieving the above object, the diagnosis and prescription support system according to claim 1 includes, in an electronic medical record system using a large-scale language model, (1) a diagnostic additional training data providing means for additionally training the large-scale language model using diagnosis names, symptoms, physical findings, and test result data of diagnosed cases extracted from the electronic medical record to generate an additionally trained large-scale language model (an additionally trained large-scale language model generating means using diagnosed case data), and (2) a diagnostic additional training data providing means for additionally training the large-scale language model using diagnosis names, symptoms, physical findings, and test result data of diagnosed cases extracted from the electronic medical record to generate an additionally trained large-scale language model, (3) the additionally trained large-scale language model is provided with a "possible diagnosis, additional symptoms, physical findings, and test answering means" that answers at least one of a list of possible diagnoses and their probabilities, and recommended additional symptoms, physical findings, and tests; and (4) a sequential diagnosis improvement means is provided that repeats the "symptom, physical findings, and test result data providing means" and the "possible diagnosis, additional symptoms, physical findings, and test answering means" until the list of diagnoses and their probabilities has sufficient accuracy to be used as a definitive diagnosis.

[0008] The diagnosis and prescription support system of claim 2 is characterized in that in the diagnosis and prescription support system of claim 1, the "possible diagnosis name, additional symptoms, physical findings, and test response means" includes a recommended test order creation means for creating an order for a recommended test in the electronic medical record.

[0009] The diagnostic and prescription support system of claim 3 is characterized in that in the diagnostic and prescription support system of claim 1 or 2, the diagnostic additional learning providing means includes not only "symptoms, physical findings, and test findings" to advance the diagnosis of diagnosed cases, but also treatment items such as administered medications, injections, rehabilitation, surgery, and other procedures.

[0010] In the diagnosis and prescription support system described in claim 4, in the diagnosis and prescription support system described in claim 1 or 2, the additionally trained large-scale language model is characterized by being equipped with a treatment plan proposal means for proposing a treatment plan including at least one treatment item such as administered medication, injection, rehabilitation, or surgery for the definitive diagnosis obtained by the sequential diagnosis improvement means.

[0011] The diagnosis and prescription support system of claim 5 is characterized in that the diagnosis and prescription support system of claim 4 is provided with a recommended treatment order creation means for creating individual treatment orders, such as administration of drugs, injections, rehabilitation, surgery, etc., proposed by the treatment plan proposal means, in an electronic medical record.

[0012] The diagnosis and prescription support system according to claim 6 is characterized in that in the diagnosis and prescription support system according to claim 1 or 4, it further comprises a progress log creation means for recording a progress log of the successive diagnosis improvement means and the treatment plan proposal means. Effect of the Invention

[0013] The diagnosis and prescription support system described in claim 1 is equipped with a diagnostic additional learning data providing means, which performs additional learning of the large-scale language model using the diagnosis name, symptoms, physical findings, and test result data of diagnosed cases extracted from electronic medical records to create an additionally trained large-scale language model (an additionally trained large-scale language model creation means using diagnosed case data). Since the system is equipped with a "symptom, physical findings, and test result data providing means," the symptom, physical findings, and test result data from the electronic medical records obtained for undiagnosed cases are provided to the additionally trained large-scale language model. The device is equipped with a "means for answering possible diagnoses, additional symptoms, physical findings, and tests," which prompts the user to answer at least one of a list of possible diagnoses and their probabilities, and recommended additional symptoms, physical findings, and tests. The system is equipped with a sequential diagnosis improvement means, which repeats the "symptoms, physical findings, and test result data providing means" and the "possible diagnoses, additional symptoms, physical findings, and test response means" until the list of diagnoses and their probabilities is accurate enough to be considered a definitive diagnosis.

[0014] The diagnosis and prescription support system according to claim 2 includes a recommended test order creating means, which creates an order for a recommended test for the electronic medical record.

[0015] The diagnosis and prescription support system described in claim 3 includes not only "symptoms, physical findings, and test findings" to advance the diagnosis of diagnosed cases, but also treatment items such as administered medications, injections, rehabilitation, and surgery.

[0016] The diagnosis and prescription support system according to claim 4 is provided with a treatment plan proposing means, which proposes a treatment plan including at least one treatment item, such as an administered drug, an injection, rehabilitation, or surgery, for a definitive diagnosis obtained by the sequential diagnosis improving means.

[0017] The diagnosis and prescription support system according to claim 5 includes a recommended treatment order creating means, which creates individual treatment orders for the administration of drugs, injections, rehabilitation, surgery, and other procedures proposed by the treatment plan proposing means for the electronic medical record.

[0018] The diagnosis and prescription support system according to claim 6 further comprises a progress log creating means, which records progress logs of the sequential diagnosis improving means and treatment plan proposing means. [Brief description of the drawings]

[0019] [Figure 1] FIG. 2 is a diagram illustrating a hardware configuration of the present invention. [Diagram 2] This is an explanatory diagram of how a large-scale language model (LLM) is additionally trained using a large amount of diagnosed case data to create an additionally trained large-scale language model. [Diagram 3] The relationship between basic training operations and incremental training of a large-scale language model is shown. [Figure 4] FIG. 2 is an explanatory diagram of additional learning in the present invention. [Diagram 5] This is an explanatory diagram showing how a diagnosis and treatment plan is created while receiving suggestions by utilizing an additionally trained large-scale language model. [Figure 6] This shows an example of the diagnostic and treatment planning process. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0020] FIG. 1 is a diagram showing an example of a hardware configuration according to the present invention. Large-scale, additionally trained language models require huge amounts of data and computational resources, so they are usually stored on the cloud and connected to a local area network (LAN) within the medical institution via a router over the Internet. The medical institution has an electronic medical record server. Doctors, nurses, and other staff access electronic medical records and large-scale language models via terminals connected to the LAN. In addition, part or all of the electronic medical record system may be built on the cloud. In addition, with the lightweight version, it is possible to set up a configuration in which part or all of the additionally trained large-scale language model is installed within the medical institution (edge ​​computing). If this is possible, it will be possible to speed up the response to inquiries and further reduce the risk of information leakage.

[0021] FIG. 2 is an explanatory diagram of a large-scale language model (LLM) that is additionally trained using a large amount of data on diagnosed cases to create an additionally trained large-scale language model. Case data such as symptoms, physical findings, and test results are extracted from the electronic medical records (case data extraction means), and "diagnosed case data" is created together with the definitive diagnosis that has already been obtained, and this is used for additional training of a large-scale language model (additional training large-scale language model creation means using diagnosed case data). Large-scale language models can sometimes return fictitious responses, known as "hallucinations," which are a problem when you ask for something that is not in the training data. This problem can be prevented by limiting the scope of the search to the large amount of diagnosed case data used for training. If possible, accuracy can be expected to improve if data on diagnosed cases from other hospitals, not just our own, can be used. For example, even if the same symptom of "chest pain" is present, a hospital specializing in cardiovascular disease may suspect "angina pectoris" or "myocardial infarction," whereas a hospital specializing in orthopedic surgery may suspect "rib fracture." For this reason, it may be useful to group the diagnosed case data by hospital as appropriate.

[0022] Large-scale language models are currently being developed rapidly, and many models have been developed, including ChatGPT (a registered trademark of OpenAI), Bard, LaMDA (a registered trademark of Google), and LLaMA (a registered trademark of Meta). In large-scale language models, a large amount of data is learned to construct a general-purpose language model (base learning model), but knowledge of a specific field is not necessarily deep. For this reason, a dedicated model is constructed by intensively learning additional knowledge about a specific field (the present invention is part of this). In addition, it is known that the large-scale language models of general-purpose substrate learning models are weak in numerical calculations and logical operations. To compensate for this weakness, there are dedicated modules such as Wolfram Alpha (Wolfram Alpha Non-Patent Document 1), both of which respond to prompt queries from the LLM and return answers.

[0023] Another possibility is to query a web search engine and import the response. Function calling has been proposed as one method to solve these problems. This is a method in which a large-scale language model uses the functions of other search engines, databases, or LLMs specialized for specific purposes via an API (Application Programming Interface). If dedicated large-scale language models can be used in parallel, it is expected that response times will be reduced. Furthermore, as described below, if it is possible to reference information from and write to the electronic medical record system, it will be possible to control the electronic medical record system directly from the large-scale language model using Function Calling.

[0024] FIG. 3 shows the relationship between the basic training operation and the additional training operation of a large-scale language model. A general-purpose large-scale language model collects as much document data as possible from sources such as the web. Any document element (e.g., a word) contained in the collected document data is masked, and the remaining document elements are used to learn how to make predictions. During this learning process, the relationships between document elements are grasped using a method called the Attention mechanism. In the document generation task, the next document element that is likely to appear is predicted based on highly related document elements and the document elements that have been generated up to that point, and the document element with the highest probability is adopted. These are achieved by multiplying the vector representation of each document element by an attention weight, accumulating the result for each partially created document element sequence, and calculating the context vector. When a document element is added, the next document element is predicted, and this procedure is repeated, making it possible to generate long documents.

[0025] Conventional learning models require a large number of documents (corpora) with teacher labels manually annotated. The costly labeling process was a bottleneck, limiting the size of the corpus. In large-scale language models, parts of a document are masked and the masked document elements are predicted using the remaining parts of the document. The masked document elements become the correct answer (teaching label), making it possible to use large corpora from unsupervised documents without human intervention. Conventionally used RNNs (Recurrent Neural Networks) only use document elements close to the document element to be output, and document elements at distant positions are "forgotten", so it was known that accuracy drops rapidly when the number of words exceeds several dozen, even with some ingenuity. LLM uses an attention mechanism to incorporate document elements at distant positions into the prediction, making it possible to generate long sentences.

[0026] FIG. 4 is an explanatory diagram of additional learning in the present invention. Large-scale language models are trained by collecting a large amount of documents from all fields, so they can understand and generate general documents without any problems. However, their accuracy in specific fields is not necessarily high. For this reason, documents related to the specific field can be additionally trained, and the two can be combined to create an additionally trained large-scale language model suited to the specific field. In the reinforcement learning of the present invention, the symptoms, physical findings, and test results of cases with a confirmed diagnosis are used as supervised additional learning data with the confirmed diagnosis as the teacher labels. In this case, information extraction using tag information from electronic medical records as described in Patent Document 7: Japanese Patent No. 7145367 may be useful. In addition, there are two types of additional learning: "transfer learning," which adds only the output layer without changing the parameters of the substrate learning model, and fine-tuning, which re-learns the parameters of the substrate learning model using additional learning data. Either method is possible, but in the present invention, fine-tuning is more preferable.

[0027] Through learning using the attention mechanism, relationships can be obtained between individual symptoms, physical findings, test findings, etc. and the training labels, which are disease names. In addition, students will learn the relationships between individual symptoms, physical examination findings, and laboratory findings. By using the attention weights for each of the initially obtained symptoms and physical findings, and the context vector obtained by accumulating them, the probability of each disease name can be calculated at the intermediate stage when several symptoms, findings, test results, etc. have been obtained. That is, in a typical large-scale language model, the document element with the highest probability is added one after another to the end of the already generated document element string, but in the present invention, the Attention mechanism is used to calculate the probability of each disease name at a certain point in time from the symptoms, physical findings, and test findings observed in a patient. This corresponds to the posterior probability of each possible disease name after the symptoms, physical findings, and test findings are partially obtained.

[0028] In Figure 4, for ease of explanation, "fever" and "abdominal pain" are aligned, but the order in which individual symptoms, physical findings, and test findings appear is arbitrary, since the attention mechanism calculates individual document element vectors and their attention weights. Although it is desirable to have all the data, it is not essential. Even incomplete data contains some information and can contribute to the accuracy of the output. Also, there are variations in notation, for example, "+" in Figure 4, but it can also be "Yes" or "Positive". Furthermore, LLM is good at translation, and it is possible to mix multiple languages ​​such as Japanese, English, French, and German.

[0029] If all that is needed is to proceed with diagnosis, symptoms, physical examination findings, and test results are sufficient; however, to receive further treatment plan suggestions from the additionally trained large-scale language model, it is necessary to include treatment elements such as administered medications, injections, rehabilitation, surgery, and other procedures in the additional training data (means for providing additional learning for diagnosis and treatment). The attention mechanism learns the relationship between each treatment element and the disease name, making it possible to suggest each treatment element using an additionally trained large-scale language model. Typically, disease names and treatment patterns have a simple 1:1 correspondence; however, when an additionally trained large-scale language model is used, the attention mechanism can obtain not only the relationship between disease names and each treatment element, but also the relationships between individual symptoms, physical findings, and test findings, making it possible to propose treatment plans that are individualized to the circumstances of each case.

[0030] Figure 5 is an explanatory diagram of how a diagnosis and treatment plan is created while receiving suggestions by utilizing a large-scale language model that has undergone reinforcement learning. For undiagnosed patients, symptoms are interviewed and physical findings are obtained and entered into the electronic medical record. The symptoms and physical findings are then fed into a large-scale, additionally trained language model. In this case, information extraction using tag information from electronic medical records as described in Patent Document 7: Japanese Patent No. 7145367 may be useful. The additional large-scale trained language model suggests possible diagnoses and their probabilities, as well as suggestions for additional symptoms, physical exam findings, and tests that can help improve the certainty of a diagnosis. The doctor enters the suggested additional symptoms and physical findings, and then selects the necessary tests from the suggested tests to create a test order in the electronic medical record.

[0031] As additional symptoms, physical findings, and test results become available, they are presented to the additionally trained large-scale language model, which again provides a possible diagnosis and its probability, as well as suggestions for additional symptoms, physical findings, and tests that could help improve the certainty of the diagnosis. If a sufficiently high probability is obtained for the possible diagnosis and its probability, and the doctor is satisfied with the diagnosis, the diagnosis is confirmed and registered in the electronic medical record. If the probability is still insufficient and the doctor is not satisfied, the same procedure is repeated until the diagnosis is confirmed (successive diagnosis improvement means). Once a diagnosis has been confirmed, the additionally trained large-scale language model will suggest appropriate medications, injections, rehabilitation, surgery, and other procedures. The doctor will then select the treatment he or she deems appropriate and create a treatment order in the electronic medical record. Here, the diagnosis names and individual examination and treatment orders proposed by the additionally trained large-scale language model are merely suggestions, and the final adoption is left to the doctor. This is because the doctor bears the final responsibility in the unlikely event of a medical accident or other problem. If there were a progress log creation means for recording the progress logs of the sequential diagnosis improvement means and treatment plan proposal means, it would be possible to verify on what basis the examination and tests were conducted and the diagnosis was reached, as well as how the treatment plan was made, providing "explainability," a problem that was lacking in conventional AI.

[0032] Figure 6 shows an example of the diagnosis and treatment process in Figure 5. An undiagnosed patient was admitted to the hospital with fever and abdominal pain. Since there are many possible illnesses based on this alone, it is suggested that additional symptoms, such as coughing and vomiting, and physical findings be checked. In addition, blood tests and X-ray CT scans are recommended. If the doctor deems necessary, an order for the test is issued. The obtained additional symptoms, physical findings, and test results are returned to the additional trained large-scale language model, which responds that the probability of a diagnosis of gallstones is high. If the doctor determines at this point that the diagnosis is definitive, the diagnosis is registered in the electronic medical record and a treatment suggestion for gallstones is requested from the additional trained large-scale language model. If the doctor judges the suggestion of painkillers or antibiotics to be appropriate, he or she issues the corresponding treatment order in the electronic medical record. Finally, a progress log is created.

[0033] When examining a patient, a doctor will first interview the patient to understand the symptoms they are complaining about, whether there was a trigger for the symptoms, and the time course of the symptoms, then take a physical examination to determine the general diagnosis of the disease, and order blood tests and X-rays to confirm the diagnosis, or to differentiate if multiple diseases are suspected. After looking at the results of the examinations, if necessary, they will check for additional symptoms and physical findings, order additional tests, and repeat the process as necessary until they are confident in the diagnosis of the disease. When the possibility of a particular disease becomes high enough, the disease is registered in the electronic medical record as a confirmed disease and an order for treatment required for that disease is issued. When "roughly guessing the name of the disease," doctors often recall cases with similar symptoms and patterns of physical findings from a large number of cases they have experienced in the past. However, as things stand, diagnosis is subjective and experiences vary from doctor to doctor, so it will be difficult to improve the accuracy rate through objective diagnosis. Furthermore, no matter how skilled a doctor is, his or her knowledge will be lost upon death. If knowledge continues to accumulate on a permanent basis, as in the present invention, it is expected that diagnostic accuracy will continue to improve.

[0034] In order to make a diagnosis based on objective criteria, it is necessary to collect a large amount of case data with confirmed diagnoses. The method of collection is optional, but in practice this would be a discharge summary or an outpatient summary at the end of the outpatient visit. In addition, information extraction using tag information from electronic medical records as described in Patent Document 7: Japanese Patent No. 7145367 may also be useful. In a situation where there is no prior information, such as when an unknown patient enters the examination room for their first visit, the probability of a disease name is the ratio of the frequencies of each disease name (prior probability). When information on symptoms or physical findings is added, the probability of the disease name having the symptoms or physical findings (posterior probability) increases, and the probability of the disease name not having the symptoms or findings decreases. When multiple disease names remain as candidates, it is necessary to decide which disease name is the correct diagnosis by integrating additional symptoms, physical findings, and test results (differential diagnosis). Here, it goes without saying that the symptoms, physical findings, and tests that have a large difference in the frequency of positive / negative results between diseases are effective as the basis for differential diagnosis. By collecting symptoms and physical findings that have a large difference in the frequency of positive / negative results between diseases, conducting tests, and providing these results to an additionally trained large-scale language model, the posterior probability for each disease name will change significantly, bringing us closer to a definitive diagnosis. Taking into account the degree of difficulty and the cost of the test, the next symptoms, physical findings, and tests can be performed. This process can be repeated until the probability of the disease name is sufficiently high (successive diagnosis improvement method).

[0035] Patent Document 5: Japanese Patent No. 6792750 describes a diagnostic support system that collects a large amount of diagnosis names, symptoms, physical findings, and test results (prior probability distribution) and narrows down the diagnosis names from set operations of the symptoms, physical findings, and test results of undiagnosed patients (posterior probability). Furthermore, Patent Document 6: Japanese Patent No. 7284970 describes a technology called a key-value store, which records and searches a large amount of records at high speed. However, there were problems such as complicated procedures for inquiries, complicated adjustments to variations in notation when collecting large amounts of data, and the enormous amount of computing resources required to record and reference large amounts of data, which resulted in considerable costs. Large-scale language models make it possible to make inquiries in everyday language, greatly reducing the complexity of inquiries. The attention mechanism used in the large-scale language model learns the relationships between document elements through inference learning of masked document elements, but an additionally trained large-scale language model is created by learning the same relationships for the data provided by the diagnostic reinforcement learning data provision means. Using this mechanism, it is possible to calculate the posterior probability for each disease name after the symptoms, physical findings, and test results of observed undiagnosed patients are obtained. Since a huge language system has already been built, by utilizing this language model, it is possible to record and process large amounts of case data, even for cases where the diagnosis has been confirmed, at a relatively low cost.

[0036] Although the embodiments have been described above, the specific configuration of the present invention is not limited to the above-described embodiments, and the present invention includes design changes and the like that do not deviate from the gist of the invention. For example, although the medical field has been used as an example, a similar system can be used in nursing care. ChatGPT and the like have been given as an example of LLM, but the same is true for other systems, and the present invention also includes systems that will be newly developed in the future. In a typical large-scale language model, a query statement called a prompt is often entered and an answer is obtained from the large-scale language model. In the present invention, this method is acceptable, but the prompt / answer may also be executed via an API between the electronic medical record and the large-scale language model.

Claims

1. In an electronic medical record system that uses a large-scale language model, (1) A diagnostic additional training data providing means is provided for additionally training the large-scale language model using diagnosis names, symptoms, physical findings, and test result data of diagnosed cases extracted from electronic medical records to generate an additionally trained large-scale language model (an additionally trained large-scale language model generating means using diagnosed case data), (2) A "symptom, physical findings, and test result data providing means" for providing symptoms, physical findings, and test result data of an electronic medical record obtained for an undiagnosed case to the additionally trained large-scale language model; (3) The additionally trained large-scale language model includes a "possible diagnosis, additional symptoms, physical findings, and examination answering means" for answering at least one of a list of possible diagnoses and their probabilities, and recommended additional symptoms, physical findings, and examinations; (4) A diagnostic and prescription support system comprising a sequential diagnosis improvement means for repeating the "symptom, physical findings, and test result data providing means" and the "possible diagnosis, additional symptoms, physical findings, and test response means" until the list of diagnosis names and their probabilities is accurate enough to be a definitive diagnosis.

2. In the "Possible diagnosis, additional symptoms, physical findings, and test response means" 2. The diagnosis and prescription support system according to claim 1, further comprising a recommended test order creating means for creating an order for a recommended test for an electronic medical record.

3. In the diagnostic additional learning providing means, 3. The diagnosis and prescription support system according to claim 1 or 2, characterized in that it includes not only "symptoms, physical findings, and test findings" for advancing the diagnosis of diagnosed cases, but also treatment items such as administered medications, injections, rehabilitation, and surgery.

4. The diagnosis and prescription support system of claim 1 or 2, wherein the additionally trained large-scale language model is provided with a treatment plan proposal means for proposing a treatment plan including at least one treatment item such as administered medication, injection, rehabilitation, or surgery for the definitive diagnosis obtained by the sequential diagnosis improvement means.

5. 5. The diagnosis and prescription support system according to claim 4, further comprising a recommended treatment order creation means for creating individual treatment orders, such as administration of drugs, injections, rehabilitation, surgery, etc., proposed by said treatment plan proposal means, for an electronic medical record.

6. 5. The diagnosis and prescription support system according to claim 1, further comprising a progress log creating means for recording a progress log of said successive diagnosis improving means and said treatment plan proposing means.

Images