WO2020046173A1

WO2020046173A1 - Automated system to support medical decisions in combined pathology

Info

Publication number: WO2020046173A1
Application number: PCT/RU2019/000589
Authority: WO
Inventors: Михаил Борисович БОГДАНОВ
Original assignee: Михаил Борисович БОГДАНОВ
Priority date: 2018-08-27
Filing date: 2019-08-19
Publication date: 2020-03-05
Also published as: RU2698007C1

Abstract

The invention relates to the field of medical diagnostics. A medical decision support system comprises an interface unit, a data input unit, a decision storage unit for each of the therapeutic fields, a computer unit, a unit that checks for contradictions in the input data, a decision output unit and a treatment recommendations unit. The interface unit receives data and exchanges data with the data input unit, the decision output unit and the treatment recommendations unit. The data input unit transmits data to the unit that checks for contradictions in the input data, and the computer unit calculates the weighting coefficients assigned to the symptoms for each disease and the scores in favor of specific diseases. The decision output unit produces information on all of the reference medical history, highlighting signs identified during examination of the patient and allowing each reference disease model to be compared and diseases to be grouped by human body organs or systems. Moreover, the treatment recommendations unit is capable of receiving data from the data input unit and the decision output unit and displaying an optimum treatment plan on the interface unit.

Description

Automatic system for supporting medical decisions with combined pathology

Description of the invention

FIELD OF THE INVENTION

The invention relates to the field of medicine, medical diagnostics, maintaining patient histories and to the category of medical information systems for supporting clinical decisions. More specifically, the invention relates to an automatic system for supporting medical decisions in combination pathology.

State of the art

Known computer systems for supporting clinical decisions (medical expert systems, medical artificial intelligence).

The first development of MYCIN, Stanford University, (Rule-Based Expert Systems: The MYCIN Experiments of the Stanford Heuristic Programming Project) dates back to the late 60s and early 70s. Similar systems were later adapted to personal computers (Naylor K. How to build your own expert system. (Build Your Own Expert System, 1987) Moscow: Energoatomizdat, 1991). Now the developments are transferred to the Internet, for example, they are developing the domestic Internet system MEDAI (http://medai.ru). Such systems are characterized in that

• The result is calculated based on the prevalence of diseases and symptoms and the result is presented in the form of a list of diagnoses ranked by probability. In reality, the doctor works with patients who have several diseases at the same time, each of which has an actual probability equal to unity for a given clinical situation, for example, “pulmonary edema due to hypertensive crisis and exacerbation of glomerulonephritis in pneumonia in a patient with bronchial asthma and rheumatism”.

• Collect clinical information using an automatic algorithm based on the prevalence of symptoms and diagnoses. This leads to cutting off unambiguous, but rare signs, markers, ignoring less common diseases, and does not allow to treat combined pathology.

• Gathering information about the most common, but not significant in this clinical situation signs according to a computer probabilistic algorithm delays the examination time, which is critical in emergency patients and at a mass reception. • It does not include the processing of clearly indicated information “not known / no data”, which is typical for initial patient admission when there are no results of analyzes, fluoroscopy and other instrumental examination methods.

• Weights that the system uses to calculate the diagnosis are automatically generated during the training of the system and cannot be interpreted by clinicians who are responsible for the consequences of the decision, including the legal one.

• Probability-based education requires information on the prevalence of symptoms and diseases in the patient category to which the patient belongs, for example, “What is the likelihood of pneumonia in long-term and often ill children in the age group of 3 to 5 years attending kindergarten against the background of the influenza epidemic in the Arctic in December, from families with incomes below the poverty line, and what are the most likely clinical signs? ” The absence of such statistics leads to the use of “fuzzy logic” with surrogate indicators such as “conviction coefficients” (MYCIN system), or requires the collection of “big data” that is currently missing, with the deployment of expensive information infrastructure and time consuming. In addition, there is a vicious circle of unresolved problems referring to each other, when for making diagnoses based on probabilities it is necessary to know which category the patient belongs to, and to determine the likelihood of a patient being assigned to this category, one needs to know the diagnoses. Systems based on "big data" are also vulnerable to the quality of input information "millions of general practitioners cannot be wrong, this is really the diagnosis and treatment of the level of general practitioners."

Clinical decision support systems are also known that use probabilistic pattern recognition methods to solve narrow clinical problems. For example, systems are known which predict:

- complications of thrombophlebitis (Method for verifying the diagnoses of diseases of veins of the lower extremities according to combined thermometry based on the Bayesian classifier and posterior probabilities, RU 2 358 644. Date of publication of the application: 27.06.2008 Bull. ”18);

- complications in obstetrics (Artificial intelligence and a device for the diagnosis, screening, prevention and treatment of conditions of the mother and fetus system. RU 2006 123 425. Date of publication of the application: 10.01.2008); - the course of a known disease in the intensive care unit based on the readings of several sensors that determine the pulse, blood pressure, temperature, partial pressure of oxygen (Method for detecting critical trends in multi-parameter patient monitoring and clinical data using clustering. RU 2008 122 936. Date of publication of the application: 12/20/2009).

Such systems do not aim to support the diagnosis and treatment of multiple pathologies in the absence of clinical data.

Also known is a system for searching for similar clinical situations in a bank of computer case histories according to a sample set by a clinician (Clinical Decision Support System with an external context. RU 2,541,198. Date of publication of the application: 06/20/2013 Bul. Ms 17), but it does not directly the goal is to support the diagnosis and treatment of multiple pathologies with gaps in clinical information, and is focused on the accumulation of input information for subsequent statistical processing in the technology of "big data".

At the same time, the technology of "big data" means the amount of data sufficient to identify, verify statistical significance, and to study phenomena and patterns that were not available on previously studied statistical samples of smaller volumes. The volume of "big data" is specific for each subject area and can reach exabytes, which requires an appropriate information infrastructure.

SUMMARY OF THE INVENTION

The invention is characterized in that the system: processes combined pathology with a simultaneous indication of all diseases in favor of which there are clinical signs; provides the user with explicit clinical signs that were taken into account when forming the decision by the system, with the ability of the user to adjust the amount of information provided; works with clearly indicated gaps in the examination of patients “no data / impossible to evaluate”; can work autonomously, without access to the Internet; can use, but does not require "big data" in training and work.

The technical result of the claimed system is aimed at improving the quality of diagnosis and treatment due to the fact that the system:

1. Provides support for clinical decisions in patients with multiple diseases; 2. Provides transparency for clinicians of recommendations issued by the system, including for clinical consultations and reviews, and in administrative and legal conflicts;

3. Provides support for clinical decisions in the absence of clinical information.

Also, the claimed system is aimed at increasing the efficiency of medical personnel in extreme situations, namely, when there is no stable Internet access, for example, when doctors work on calls, in remote areas, and in emergency situations.

The system can be trained and work without access to statistical information and "big data", which allows you to quickly implement the system without deploying a massive information infrastructure, and to minimize the impact of low-quality input data during training.

Allows you to save time on the mass reception of patients with typical pathology, and if necessary, analyze in detail complex and rare cases.

The claimed technical result is achieved due to the fact that the automatic medical decision support system for combined pathology contains an interface unit, an input data unit made in the format of an electronic medical history, a solution storage unit for each of the therapeutic areas, a computing unit, an input data inconsistency check unit , a unit for issuing decisions, a block of recommendations for treatment, while the interface unit is configured to receive data from a clinician, or from databases with electronic medical histories, or from the repositories of "big data", while the interface unit is configured to exchange data with the data input unit, with the decision-making unit and with the treatment recommendations unit, while the input data unit is made in the format of an electronic medical history with the possibility transmitting input information to the computing unit and with the ability to store reference models of various diseases corresponding to different organs and systems of a person, while the input data block is configured to transmit data to the block checks for inconsistency of the input data, and the computing unit is configured to recalculate weighting factors assigned to the symptoms or signs for each disease of a particular organ or system of the human body, and points in favor of specific diseases, while the interface unit, input data blocks and inconsistency checks , decision storage unit, computing unit, decision issuing unit, the block of treatment recommendations is made with the ability to work on calls and / or in remote areas and / or in emergency zones without stable Internet access, the block for issuing decisions is made with the ability to display information about all reference case histories with highlighting signs or symptoms identified when examining a patient, or displaying information about those reference diseases in which there are common signs or symptoms with the data obtained during the examination, and with the possibility of comparing each reference mode and diseases with the entered patient data survey; analysis of the sufficiency of the survey data to identify diseases; requesting additional examinations if it is impossible to determine the disease or prescribe a treatment plan; grouping diseases according to organs or systems of the human body, while the block of treatment recommendations is made with the possibility of obtaining data from the input data block and the block for issuing decisions and outputting the optimal treatment plan to the interface block.

These technical results are achieved due to the fact that:

1. To support clinical decisions in patients with several diseases, the system provides the clinician with a list of all possible diseases in favor of which there are clinical signs, ranked in accordance with the closeness in the multidimensional space of signs to the reference disease models created during the training of the system.

2. For the transparency of computer recommendations when issuing the result to the clinician for each of the alleged diagnoses, the system displays the patient’s clinical signs that are available (i) that support the particular disease, (i) the symptoms that are against it, and (w) the symptoms, for which there is no data, but which must be (further examined by the patient) to confirm / refute each of the alleged diseases.

3. To support clinical decisions in the absence of information, the system allows you to leave gaps in the examination of patients, based on the assessment by the clinician of the clinical situation and / or the availability of diagnostic capabilities, as well as explicitly indicate "No data / Unable to evaluate." When calculating the proximity coefficients of available clinical data to reference disease models in a multidimensional symptom space, the system takes into account data gaps, and when the clinician provides results for each of the probable diseases, the system indicates signs for which there is no data, but which must be (to examine the patient) to confirm / refute the disease.

4. For the system to work in situations where there is no stable access to the Internet, decision storage units formed according to the results of the system’s training on reference disease models in a multidimensional feature space are available directly in each working instance of the system and are available on any computing device on which it is deployed system.

5. For training and operation of the system without access to statistical information and “big data”, reference models of diseases are created by medical experts for training the system on the basis of normative documents of the Ministry of Health, guidelines and recommendations of scientific societies of clinicians for each of the medical specialties. The proximity of the available clinical data to the reference model of the disease in a multidimensional space of signs is calculated by the system using any of the known mathematical and / or algorithmic methods (see, for example, I. Chernorutsky, Decision Making Methods. - St. Petersburg: BHV-Petersburg, 2005. - 416 pp., Fomin Ya. A. Pattern Recognition: Theory and Applications. - M .: FAZIS, 2014.), including but not limited to methods for pattern recognition and object classification, such as distance estimates, clustering, k-means, algorithm k-medians, minimizing mathematical about expectations, “deep learning” of artificial neural networks.

6. To save time on the mass admission of patients with typical pathology, and if necessary for a detailed analysis of complex and rare cases, the priority of collecting information is determined by the clinician, in accordance with his assessment of the clinical situation, without automatically cutting off the system of rare signs, and without automatically imposing relevant signs. The system also presents the result to the clinician both in a compact form, with cutting off the gaps in the collected information, which makes it possible to conduct mass reception of patients with typical cases, and in a detailed form across the entire combination of clinical signs and suspected diseases that are necessary for system training and in the analysis of complex clinical cases.

List of figures of drawings and other materials

In FIG. 1 is a block diagram of an automatic medical decision support system. In FIG. 2 shows the structure of the system for processing information at the level of organs and systems of the body.

In FIG. 3. The structure of the system for processing information at the pathology level of an individual organ and / or system is presented.

In FIG. 4 shows an example of ternary input of information in the interface of an electronic medical history.

In FIG. 5 presents a detailed presentation of the results issued by the system for training the system and analysis of complex cases.

In FIG. 6 presents an abbreviated presentation of the results issued by the system for standard situations in various diseases.

The implementation of the invention

According to figure 1, the automatic medical decision support system for combined pathology, contains an interface unit 1, an input data unit 2, made in the format of an electronic medical history, a decision storage unit 5 for each of the therapeutic areas, a computing unit 4, an inconsistency checker 3 input data, decision issuing unit 6, treatment recommendations block 7. Interface unit 1 is configured to receive data from a clinician, or from databases with electronic medical records, or from repositories of “big data” s "and the ability to communicate with the data input unit 2, a dispensing unit used solutions and recommendations for the treatment unit 7. The input unit 2 is configured in the format of the electronic medical records to transmit the input information to the computing unit

4 and with the ability to store reference models of various diseases corresponding to different organs and systems of a person. The input data block 2 is configured to transmit data to the block for checking the inconsistency of 3 input data. Computing unit 4 is configured to recalculate weights assigned to symptoms or signs for each disease of a particular organ or system of the human body, and points in favor of specific diseases. Interface block 1, input data blocks 2 and inconsistency checks 3, decision storage block 5, computational block 4, decision issuing block 6, treatment recommendation block 7 are configured to work on calls and / or in remote areas and / or emergency areas without sustainable internet access. Block issuing decisions 6 is configured to display information about all the reference histories of diseases with highlighting signs or symptoms identified during the examination of the patient, or display information about those reference diseases in which there are common signs or symptoms with data obtained during the examination, and with the possibility of comparing each reference model of the disease with the entered patient examination data; analysis of the sufficiency of the survey data to identify diseases; requesting additional examinations if it is impossible to determine the disease or prescribe a treatment plan; with the possibility of grouping diseases into organs or systems of the human body. The block of treatment recommendations 7 is made with the possibility of obtaining data from the input data block 2 and the block for issuing decisions 6 and outputting the optimal treatment plan to the interface block 1.

Information confirming the possibility of carrying out the invention

The system operates in two modes: in training mode and in support of medical decisions.

In training mode, the system works as follows:

The training data for each disease is entered into the system through the interface unit 1 into the input information unit 2. The input information unit 2 can be made in a format that includes, but is not limited to a standard computerized workplace of a clinician with an electronic medical history (Fig. 4), and / or in the input format from the database of electronic case histories, and / or in the input format of processed "big data".

Input data (symptoms, syndromes, examination indicators) are structured

(figure 5, 6) in accordance with the organs and systems of the body and their diseases

(respiratory system, cardiovascular, digestive and others).

When training the system, the input data are reference models of diseases in the form of a set of clinical signs characteristic of the disease, which were created by medical experts on the basis of regulatory documents

Ministry of Health, guidelines and recommendations of scientific societies of clinicians for each of the diseases, including with the involvement of "big data".

For example, for pneumonia, the key signs are acute, during the day, onset of the disease, fever, leukocytosis, physical changes in the lungs (wheezing during auscultation, dullness during percussion), and blackout on the radiograph. The medical expert enters these symptoms in the interface of the electronic medical history (Fig. 6). An alternative way to create reference models of the disease is to identify the most common clinical signs characteristic of specific diseases, based on the processing of databases, containing computer medical records, including "big data." To enter the reference models of diseases generated by the processing of databases, the same format of the electronic case history can be used.

The system in the inconsistency check unit 3 checks the input data for consistency according to the criteria previously set by the medical expert for each organ and system. For example, “vesicular respiration in both lungs” cannot be present simultaneously with “crepitus in the right lower lobe of the lungs.”

The training input data specific to the reference model of a specific disease is transmitted to computing unit 4 to generate weighting coefficients of clinical signs characteristic of the disease.

For example, for the formation of weighting coefficients of symptoms and scoring, reflecting a digital assessment of the proximity of actual symptoms to the reference models of diseases, when training the system, you can use the linear separation method of objects (below is a conditional example):

• Step 1. Weighting factors of all signs are equal to zero. In the input data block 2, through the interface block 1, a training reference model of the disease "pneumonia" is entered. The values of the signs of the reference model “pneumonia” (1 for fever and leukocytosis) are multiplied by the values of the coefficients of symptoms for each disease obtained from the decision storage unit 5, multiplication by zero gives zero.

• Step 2. The medical expert who conducts the training of the system, in the input data block 2 indicates the correct diagnosis of “pneumonia” for which training is carried out, after this the attribute fields for “pneumonia”, which in the reference model are not equal to 0, increase by 1, for other diagnoses, the fields of similar signs decrease by 1. When the reference model “pneumonia” is re-introduced and the values of the signs of the reference model “pneumonia” are multiplied (1 for fever and leukocytosis) by new values of the coefficient of symptoms for each disease s pick up maximum points diagnosis "pneumonia". The modified weighting coefficients of the signs are stored in the storage unit solutions 5 for subsequent use. • Step 3, similar to step 1. Training in the diagnosis of bronchial asthma. The weights of the reference model of the disease “bronchial asthma” (1 for suffocation on exhalation and eosinophilia) are multiplied by the values of the coefficients of the signs of the disease formed in the previous steps (after learning the diagnosis of pneumonia) obtained from the decision storage unit 5. Multiplication of any number by zero gives zero.

• Step 4, similar to step 2. The medical expert who conducts the training of the system indicates the correct diagnosis of bronchial asthma for which training is being carried out, after this the attribute fields for bronchial asthma, which are not 0 in the reference model, increase by 1 , for other diagnoses, the fields of similar signs decrease by 1. When the reference model “bronchial asthma” is repeated and the values of the signs of the reference model “bronchial asthma” are multiplied (1 for suffocation on inspiration and eosinophilia) by new values, the coefficient a sign for each of the diseases attains maximum points diagnosis "bronchial asthma". The modified weighting coefficients of the signs are stored in the storage unit solutions 5 for subsequent use.

• Step 5, similar to steps 1 and 3. Training in the diagnosis of chronic obstructive pulmonary disease. The weights of the reference model of the disease “chronic obstructive pulmonary disease” (1 for fever, suffocation on exhalation and leukocytosis) are multiplied by the values of the coefficients of the signs of the disease formed in the previous steps (after learning the diagnoses of “pneumonia” and “bronchial asthma”) obtained from the block storing solutions 5. Multiplying any number by zero gives zero.

• Step 6, similar to steps 2 and 4. The medical expert who conducts the training of the system indicates the correct diagnosis of “chronic obstructive pulmonary disease” for which the training is carried out, after this field of signs for “chronic obstructive pulmonary disease”, which are not in the reference model equal to 0, increase by 1, for other diagnoses, the fields of similar signs decrease by 1. When the reference model “chronic obstructive pulmonary disease” is re-introduced and the values of the signs of the reference model “chronic obstruction” are multiplied structural lung disease ”(1 for fever, suffocation on exhalation, and leukocytosis) for the new values of the coefficient of symptoms for each of the diseases, the diagnosis“ chronic obstructive pulmonary disease ”will get maximum points. Modified weight feature coefficients are stored in decision storage unit 5 for later use.

The steps are cyclically repeated until the system is fully trained in all diseases, taking into account all clinical signs.

When using the system outside the training mode, for the diagnosis of diseases, the steps are used to calculate points reflecting the closeness of the actual values to the reference models of diseases (in the conditional example with the diagnoses of “pneumonia”, “bronchial asthma”, “chronic obstructive pulmonary disease” steps 1, 3 and 5), exclude the steps associated with the modification of the coefficients of signs (in the conditional example with the diagnoses of “pneumonia”, “bronchial asthma”, “chronic obstructive pulmonary disease” steps 2, 4 and 6), and instead of the reference models Evan used actual clinical symptoms of the patient.

Thus, training input data specific to the reference model of the disease are transferred to the computing unit 4 from the input data block 2, and from the decision storage unit 5 in the form of weighting factors assigned to the symptoms / syndromes / clinical indicators specific to each disease of a particular organ / body systems, calculated, for example, by linear separation, as described above.

Based on the training data of the reference model of the disease and previously generated solutions in the form of weighting factors (empty values at the beginning of training), the system calculates the degree of proximity of the training data to previously formed solutions in the form of a digital coefficient, and issues a solution.

Calculations in a multidimensional feature space are performed by any of the known mathematical and / or algorithmic methods for classifying objects and pattern recognition, including, but not limited to linear separation, clustering, k-means, minimization of mathematical expectation, gradient descent, and methods for implementing artificial neural networks. For example, in terms of “deep learning” of artificial neural networks, the basic elements are clinical signs (small bubbling rales in the lungs from the bottom right and dullness when percussing lungs from the bottom), they form higher-order elements (small bubbling rales + dullness during percussion = physical signs of pneumonia) which in combination with other elements (physical changes in the lungs + fever + sudden onset + blackout on the radiograph) give the diagnosis of "pneumonia". Several diagnoses form a picture of the patient's condition "pneumonia in a patient with bronchial asthma against the background of hypertensive crisis and renal failure." An embodiment of the dispensing form is shown in FIG. 5, 6, with the standardization of a digital coefficient for assessing proximity to a reference disease in the range from 0 to 1; other ways of providing an assessment are possible, for example, normalization from -1 to 1. When the result is output, the internal weighting coefficients of clinical signs calculated by the system and not subject to interpretation by the clinician, the system replaces the results in block 6 with digital and color indicators of signs for interpretation by the clinician, which evidence in favor of a specific disease, or against it, or for which there is no data, but which must be (additionally examine the patient) to confirm / refute each of the alleged diseases. For example, in FIG. 5, 6, the field of clinical signs contains a digital indication in the ternary system “-1” / “0” / ”1” (No sign / Unable to evaluate / There is a sign) and a color indication in a gradation of gray tones. Indication of clinical signs can be implemented in any other form depending on the particular implementation of the system, for example, in the form of pictograms and symbols (W, [],?, -, + and other options) and color indication (for example, red alarming tones for pathological symptoms, calming light green tones for normal conditions, gray tones for lack of information and other options).

Diagnostic solutions calculated by the system can be combined with a block of treatment recommendations 7, implemented as part of the system, or independently of it. Likewise, training takes place on additional layers of information processing. This is the formation of syndromes from symptoms, followed by the use of syndromes in the computing unit for the diagnosis of diseases, and the assessment of risk groups / stage of severity of the disease. For example, the intraday divergence of body temperature of more than 1.5 degrees for several days is interpreted as “hectic fever”, which testifies in favor of bacterial purulent-destructive processes.

In the support mode for medical decisions, the system works as follows:

The examination data of a particular patient is entered into the system through the interface unit 1 into the input information unit 2. The input information unit 2 can be made in a format that includes, but is not limited to a standard computerized workplace of a clinician with an electronic medical history (Fig. 4), and / or in format input from a database of electronic case histories, and / or in the input format of processed "big data".

The input data (symptoms, syndromes, examination indicators) are structured in accordance with the organs and systems of the body and their diseases (respiratory system, cardiovascular, digestive and others).

The clinician has the opportunity to explicitly indicate the presence of a symptom / clinical sign, its absence, the impossibility of evaluating / lacking data, or not filling out fields that, from his point of view, are not relevant in this clinical situation. In the example of FIG. 4 designations are used in the ternary system "-b> /" 0 "/" 1 "(Sign Absent / Unable to evaluate / Sign present); Information can be entered in any other form depending on the specific implementation of the system, for example, in the form of pictograms and symbols (□, П,?, -, + and other options).

The system checks the entered data for consistency (3 in Fig. 1) as described in the "training mode" section.

Then the patient input is transferred to the computing unit 4. There, from the decision storage unit (5 in Fig. 1 and Fig. 3), previously formed decisions are transmitted in the form of weighting factors assigned to the symptoms / syndromes / clinical parameters specific to each disease of a particular organ / system of the body.

Based on the available patient data and previously generated solutions in the form of weighting coefficients for the reference disease models, the system calculates the degree of proximity of the patient data to each of the reference disease models and issues a solution, for example, by linear separation method, as described in the “training mode” section. Calculations in a multidimensional space of attributes are performed by any of the known mathematical and / or algorithmic methods (see, for example, Chernorutsky

I. G. Methods of decision making. - St. Petersburg: BHV-Petersburg, 2005. - 416 p., Fomin Ya. A.

Pattern recognition: theory and applications. - M .: FAZIS, 2014.), classification of objects and pattern recognition, including, but not limited to linear separation, clustering, k-means, minimization of mathematical expectation, gradient descent, methods for implementing artificial neural networks.

Comparison of patient data with reference disease models occurs for each of the available reference disease models in the therapeutic category

(pulmonology, cardiology, endocrinology, etc.) independently of other reference disease models. This allows diagnosis in patients with multiple diseases. The doctor has the opportunity to choose which therapeutic category to evaluate, which saves time in typical cases, as well as evaluate all organs and systems of the body in complex cases.

An embodiment of the dispensing form is shown in FIG. 5 and 6, with normalization of the digital coefficient for assessing proximity to a reference disease in the range from 0 to 1. In the example of FIG. 6, diseases were cut off for which the proximity coefficient between the patient data and reference disease models is below 0.4 (when normalizing the proximity coefficients from 0 to 1), which allows to cut off excess information in typical cases. In the example of FIG. 5 diseases are displayed in full for the analysis of complex cases.

In this case, the output tables in FIG. 5 and 6 contain the following fields:

8- fields of diagnoses in favor of which there is data;

9- field of digital assessments of the proximity of actual symptoms to the reference model of the disease;

10 - field of actual symptoms identified during the examination;

11- symptom fields with digital and color indication of coincidences with disease standards;

12- Fields of diagnoses for which no data are available.

Upon delivery of the result, the internal weighting coefficients of clinical signs, calculated by the system and not subject to interpretation by the clinician, replace digital signs and color indicators of signs for interpretation by the physician, which indicate in favor of a particular disease or against it, or for which there is no data, but which must be ( to examine the patient) to confirm / deny each of the alleged diseases, similar to that described above in the "training mode" section.

The volume of the output of symptoms / clinical signs can be adjusted to cut off excess information in typical cases, or output in full for analysis of complex cases. In FIG. Figure 6 shows an example of issuing only that symptomatology for which there is a coincidence between the data of a particular patient and reference models of diseases, with cutting off that symptomatology of reference models of diseases for which there is no data for a particular patient.

When reviewing the results calculated by the system, the clinician can agree with them and proceed to the choice of treatment (can be implemented as an additional unit integrated with the system, or independently of it), or enter additional data that was skipped during the initial input of patient information, and / or to conduct an additional examination of the patient, and restart the system with updated data.

Claims

Formula

An automatic medical decision support system for combined pathology, containing an interface unit, an input data unit made in the format of an electronic medical history, a solution storage unit for each of the therapeutic areas, a computing unit, an input data inconsistency check unit, a decision issuing unit, a recommendation unit for treatment, while the interface unit is configured to receive data from the clinician, or from databases with electronic medical histories, or from repositories of "big data", while m the interface unit is capable of exchanging data with the input data unit, with the decision-making unit and with the treatment recommendations unit, while the input data unit is made in the format of an electronic medical history with the possibility of transmitting input information to the computing unit and with the ability to store reference models of various diseases corresponding to different organs and systems of a person, while the input data block is configured to transmit data to a block for checking for inconsistency of input data, and The th block is made with the possibility of recalculating the weight coefficients assigned to the symptoms or signs for each disease of a particular organ or system of the human body, and points in favor of specific diseases, while the interface block, input data block and inconsistency check block, decision storage block, computational unit, unit for issuing decisions, unit for recommendations on treatment made with the ability to work on calls, and / or in remote areas and / or in emergency areas without sustainable access to the Internet, to the issuance of decisions, it is possible to display information on all reference medical histories with highlighting signs or symptoms identified during the examination of the patient, or display information on those reference diseases in which there are common signs or symptoms with data obtained during the examination, and with the possibility comparing each reference model of the disease with the entered patient examination data; analysis of the sufficiency of the survey data to identify diseases; requesting additional examinations if it is impossible to determine the disease or prescribe a treatment plan; grouping diseases according to organs or systems of the human body, while the block of treatment recommendations is made with the possibility of obtaining data from the input data block and the block for issuing decisions and outputting the optimal treatment plan to the interface block.