Method for selecting a patient treatment method

Abstract

The invention relates to selecting a patient treatment method. The method, implemented using a processor, includes the steps of: obtaining depersonalized patient medical data; extracting information about patients with a similar type of illness, the treatment method used, the treatment outcome and data about the clinic and the treating physician; determining from the information obtained criteria relating to the indications and contraindications of the treatment methods used, according to each clinic and treating physician, and generating a list of criteria for an algorithmic search according to medical approaches for curing the given type of illness; determining the type of medical care; generating a rating of the clinics and physicians and selecting a method for treatment with a radical treatment method or a palliative treatment method; sending to the patient the selected treatment method and a list of clinics and physicians in the patient's region, rated in descending order. The invention is intended to allow automated selection of a patient treatment method.

Description

[0001] METHOD OF SELECTING A TREATMENT METHOD FOR PATIENTS

[0002] AREA OF TECHNOLOGY

[0003] This technical solution relates to the field of computer technologies used in medicine, in particular to the method of selecting a treatment method for patients.

[0004] LEVEL OF TECHNOLOGY

[0005] To select the most successful treatment strategy for a patient, it is necessary to consider many factors: the characteristics of the disease, concomitant diseases, constitutional and genetic factors, etc.

[0006] In addition, a specialist must be familiar with all the specifics of modern treatment methods, both surgical and conservative. Typically, over the course of years of practice, accumulating experience and knowledge from their own practice, literature, and the experience of other specialists, a physician develops a treatment algorithm for each individual patient. The main limitation for a specialist when processing large volumes of data is time [1].

[0007] The integration of artificial intelligence (AI) technologies into medicine largely determines its future. The development of modern technologies such as AI makes it possible to mimic human cognitive functions, reducing the time required to process large volumes of data and enabling accurate predictions for personalized patient treatment [2].

[0008] These advantages have enabled the widespread application of AI in medicine and other fields. The term artificial intelligence is a branch of computer science, first coined by computer scientist John McCarthy in 1956 [3]. AI refers to the use of automated or computer modeling of human behavior to solve complex problems [4].

[0009] A well-known project in the technology field is Prodoctorov.ru (owned by MedRating LLC), which provides information on medical clinic and doctor ratings based on patient reviews. The relevance of reviews is verified by the presence of a receipt or contract for services rendered by the clinic or doctor. In this project, the system allows patients to independently make a choice based on the review rating. The service also provides pricing information.

[0010] The disadvantages of this project are that the patient makes their own choice based on reviews, which are usually left immediately after the initial consultation, which does not allow for an assessment of the treatment results. Reviews usually do not indicate the reasons for the visit, and positive reviews are based on the doctor's personal perception, not the treatment effectiveness, which the system verifies.

[0011] Additionally, a solution known from the prior art is DocDoc.ru (owned by DocDoc LLC), an online platform where patients can receive a specialist consultation online (telemedicine) or schedule an in-person clinic appointment. Patients choose a clinic / doctor based on reviews and location.

[0012] Another disadvantage of this solution is that the patient makes their own choice based on reviews, which are usually left immediately after the initial consultation. This makes it impossible to judge the treatment results. Reviews usually do not indicate the reasons for the visit, and positive reviews are based on the doctor's personal perception, not the effectiveness of the treatment, which the system verifies.

[0013] Medicare.gov, a well-known American doctor search system, is an official government portal listing all licensed clinics and doctors, with insurance companies providing coverage. Patients select a doctor based on several filters: insurance type, doctor's gender, language spoken, location, treatment plan, type of medical facility, and more.

[0014] The disadvantage of this system is that the patient himself chooses the treatment method and doctor based on a huge number of options; the system does not automatically help make a choice, and does not provide information on the effectiveness of the selected methods / clinics / doctors in treating a specific disease.

[0015] The prior art discloses information source US20200350072A1, published on November 5, 2020, which discloses a method for medical diagnosis, treatment, or evaluation of patient tests, using the input of patient symptoms into a computer program and evaluation, and the software matching the information with a medical code protocol and displaying correlated codes during a patient appointment, including:

[0016] (a) entering a description of the patient's symptoms into a device that has access to the medical code protocol database;

[0017] (b) evaluating the patient's symptom description against the medical code protocol and matching the medical code with the patient's symptom description;

[0018] (c) observation of correlated medical code;

[0019] (d) creating a patient diagnosis or treatment plan, or testing a patient with a correlated medical code; and (e) performing steps (a) through (d) during the patient's appointment. A disadvantage of this solution is that it lacks the ability to automatically select a treatment method and provide the patient with a rating of clinics and doctors.

[0020] The claimed solution differs from the solutions known from the prior art in that:

[0021] 1) The patient is offered the most effective automated treatment selection based on their personal data, including gender, age, current symptoms, duration of illness, comorbidities, their severity, information from previous medical visits, and treatment outcome. The selection is performed using machine learning methods based on collected datasets of completed treatments and achieved treatment outcomes.

[0022] 2) They rank clinics and doctors and offer the patient the best choice of clinic / doctor based on their effectiveness and location. The effectiveness of the treatment proposed for a given problem is verified at a specific clinic and with a specific doctor. Each completed treatment is monitored at every stage, with results recorded.

[0023] 3) In the declared solution, the service offers information with a brief description / demonstration of the selected procedure by the patient; a detailed target description of the procedure is available to medical workers.

[0024] ESSENCE OF THE INVENTION

[0025] The declared technical solution proposes a new approach to selecting a treatment method for patients.

[0026] The technical result is to provide the ability to automatically select a treatment method for patients.

[0027] The said technical result is achieved by implementing a method for selecting a treatment method for patients, performed with the aid of a processor and containing the following stages:

[0028] - receive depersonalized medical data of the patient such as:

[0029] • patient’s gender, age, current symptoms, duration of the current illness, concomitant diseases, their severity, information from previous visits to medical care, treatment outcome;

[0030] - process the obtained anonymized medical data of the patient using a trained machine learning model and, based on the processed data, extract information from a database with anonymized medical records about patients with a similar type of disease, the treatment method used, the achieved treatment result and data about the clinic and the attending physician;

[0031] - determine, from the information received about patients with a similar type of disease and data about the clinic and the attending physician, the criteria related to the indications and contraindications of the treatment methods used for each clinic and attending physician, and form a list of criteria for an algorithmic search in medical areas for the elimination of this type of disease;

[0032] - determine what type of medical care was provided for a given type of disease - palliative or radical;

[0033] - based on the identified criteria and type of medical care, as well as data on the clinic and the attending physician, a rating of clinics and doctors is formed and a treatment method is selected with priority towards a radical direction; in the event that radical treatment cannot be carried out due to non-compliance with the criteria, palliative treatment methods are offered;

[0034] - send the patient the selected treatment method and a list of clinics and doctors in the patient's region in descending order of rating.

[0035] DETAILED DESCRIPTION OF THE INVENTION

[0036] The following detailed description of the invention includes numerous implementation details intended to provide a clear understanding of the present invention. However, one skilled in the art will readily appreciate how the present invention may be utilized with or without these implementation details. In other instances, well-known methods, procedures, and components have not been described in detail to avoid unnecessarily obscuring the features of the present invention.

[0037] Furthermore, it will be clear from the foregoing description that the invention is not limited to the embodiment described. Numerous possible modifications, changes, variations, and substitutions, while preserving the spirit and form of the present invention, will be apparent to those skilled in the art.

[0038] As follows from the description below, the stated method for selecting a treatment method for patients consists of several stages.

[0039] In the stated decision, at the first stage, depersonalized medical data of the patient is obtained, such as:

[0040] • patient gender, age, current symptoms, duration of the current illness, comorbidities, their severity, information from previous medical visits, and treatment outcome. At the next stage, the obtained anonymized patient medical data is processed using a trained machine learning model (MLM), and based on the processed data, information on patients with a similar type of illness, the treatment method used, the achieved treatment outcome, and information about the clinic and treating physician is extracted from a database of anonymized medical records.

[0041] Machine learning (ML) is a set of methods that automatically discover patterns in pre-existing data and then use them to make predictions. ML is a class of artificial intelligence methods.

[0042] A machine learning model is an artificial intelligence method that solves a problem based on experience gained from training on a data set.

[0043] The proposed solution develops the content and structure of a medical database with anonymized patient data transmitted through the service's web interface. The database is compiled for all the aforementioned disease types, including addiction medicine, urology, gynecology, and proctology, for subsequent use in machine learning of the search model. The database includes patients' personal information and their medical problems, specific treatment methods used, the selected clinic and physician, and the treatment outcomes achieved.

[0044] The MMO is trained using data generated from real questionnaires completed by doctors. The data is generated in CSV format and verified by expert specialists for accuracy.

[0045] At the first stage of training, MMOs collect data:

[0046] • Data sources: medical records, patient questionnaires, clinic data.

[0047] • Data types: demographic data, medical history, current symptoms, socioeconomic data, clinic ratings, treatment method used, treatment outcomes achieved, and clinic and physician information.

[0048] Next, the data is cleared:

[0049] • Missing value removal: If a significant portion of data is missing, rows or columns are removed.

[0050] • Filling in missing values: with mean values, median or most frequent values.

[0051] After which the data is pre-processed:

[0052] • Numerical data normalization: bringing numerical data to a uniform scale (e.g., Min-Max Scaling). Input patient data is converted to a numerical format and pre-processed - cleaning from anomalies, filling in missing values, normalization, etc.

[0053] • Encoding categorical features: using methods such as one-hot encoding.

[0054] The next step is data separation:

[0055] • Training and testing sets: split the data into training (e.g. 80%) and testing (20%) sets to evaluate the model.

[0056] Tutorials and examples:

[0057] • Input data: patient data (patient gender, age, current symptoms, duration of the current illness, comorbidities, their severity, information from previous medical visits, treatment outcome). These features are encoded using the one-hot method.

[0058] • Training: based on historical data of patients, their diagnoses, treatment methods used, and treatment results achieved.

[0059] The MMO is trained on historical data using methods such as logistic regression or a neural network model. The MMO is also trained using loss function (cross-entropy) estimates and measurements of the Accuracy, Fl-score, and ROC-AUC metrics.

[0060] Finding the optimal treatment method:

[0061] • Model: trained on data about treatments and their outcomes.

[0062] • Process: The model predicts the likelihood of success for each treatment.

[0063] Selecting a clinic:

[0064] • Model: trained on data about clinics and their ratings.

[0065] • Process: The model selects the clinic with the highest probability of successful treatment.

[0066] Choosing a doctor:

[0067] • Model: trained on data about doctors and their ratings.

[0068] • Process: The model selects the physician with the highest probability of successful treatment.

[0069] The next step in the proposed method is to use the information obtained on patients with a similar disease type, as well as data on the clinic and treating physician, to determine criteria related to the indications and contraindications for the treatment methods used for each clinic and treating physician. This is then used to generate a list of criteria for an algorithmic search across medical specialties for the treatment of this disease. This also determines whether the type of medical care provided for this disease was palliative or curative.

[0070] Data on clinics and doctors is collected from open sources and stored with parameters such as: type (organization or private doctor), location, medical service and its parameters (medications, dosage, duration of treatment), and cost.

[0071] Information on patient treatment outcomes is collected at intervals sufficient to ensure positive results with the selected treatment methods. Information on treatment methods is entered into a database for subsequent training of the medical staff.

[0072] Data tagging for treatment outcomes involves identifying the type of result achieved (information is collected through feedback from clients who completed the treatment program as recommended by the service). Specifically, the service outcome may be:

[0073] 1) Radical - complete recovery;

[0074] 2) Palliative - relief (improvement of condition);

[0075] 3) Stabilization (the condition has not changed);

[0076] 4) Deterioration (the condition has changed for the worse).

[0077] Based on the labeled and prepared data, the classification MMO is trained, which implements the following tasks.

[0078] Based on the patient's anonymized medical data, the most appropriate treatment method is predicted. Input: patient data. Output: the most effective treatment method.

[0079] The proposed solution uses three approaches to data analysis.

[0080] 1) Classification. The required labeled data is in the "patient data - best treatment" format. Machine learning algorithms used range from classical ones (catboost) to neural networks (such as the standard multilayer perceptron (MLP) or convolutional networks (such as ResNet).

[0081] Neural network for classification - ResNet34:

[0082] • Quality metrics - Accuracy, Precision, Recall, Fl - more than 80%

[0083] • Network inference time (processing time for 1 data packet) - 34 ms

[0084] • Number of trainable parameters - 63.5M.

[0085] 2) Recommender system. The required data includes data on patients, treatment types, and outcome measures (outcome type). A Deep Learning Recommendation System (DLRM) can be used as the architecture for the recommender system.

[0086] Neural network for recommendation - DLRM:

[0087] • Quality metric - Micro average precision (MAP) - more than 60%, root mean square error (RMSE) - less than 20%

[0088] • Network inference time (processing time of 1 data packet) - 20 ms

[0089] • Number of trainable parameters - 540M.

[0090] 3) Regression. The target treatment outcome feature is the percentage effectiveness of the treatment. The neural network will predict the percentage effectiveness of a given treatment method. The algorithm uses a network architecture similar to that used for the "Classifying" approach, but with a different final layer.

[0091] Neural network for regression:

[0092] • Quality metric - root mean square error (RMSE) - less than 20%, MAE

[0093] - less than 20%

[0094] • Network inference time (processing time for 1 data packet) - 3 / 4 ms

[0095] • Number of trainable parameters - 63.5M.

[0096] Based on the identified criteria and type of medical care, as well as data on the clinic and treating physician, a ranking of clinics and doctors is created and a treatment method is selected, with a priority on radical treatment. If radical treatment is not possible due to non-compliance with the criteria, palliative treatments are offered. The selected treatment method is then sent to the patient, along with a list of clinics and doctors in the patient's region, ranked by rating. The proposed neural network approach is to use recommender systems such as DLRM based on the Pytorch framework.

[0097] Based on the collected data, aggregate metrics are regularly calculated for various data segments: by clinic / physician, by treatment method, and by patient. The resulting metrics are used both to generate analytical reports and as additional parameters for MMOs, such as:

[0098] - the number of successful treatment results in this clinic;

[0099] - percentage of cured patients from all those who applied;

[0100] - percentage of success for each treatment method;

[0101] - the number of patient visits with this problem. To improve the predictive qualities and generalization capacity of the MMO, it is continuously retrained using newly accumulated data. The aforementioned data preparation and training stages are repeated regularly.

[0102] The proposed technical solution addresses a critical issue facing any patient: how to best achieve the best treatment results, which clinic, and which specialist to consult. The proposed solution, which leverages deep machine learning on collected datasets from completed cases, is designed to provide patients with the most effective treatment selection, the most qualified clinic, and the most competent physician.

[0103] The declared technical solution performs the following functions:

[0104] 1) Analytical search function: machine learning based on collected datasets from implemented cases for:

[0105] • automated selection of the most effective treatment method / service for a specific type of disease;

[0106] • compiling a rating of clinics / doctors based on the effectiveness of treatment results;

[0107] • automated selection of the most effective clinic / doctor in the patient’s region to achieve the result of a radical / palliative treatment method, providing a list of clinics / doctors in descending order of rating, indicating location and cost.

[0108] 2) Information function: an overview of modern methods of treating a specific medical problem, including radical and palliative ones, a detailed schedule of targeted procedures for medical workers.

[0109] 3) Control and coordination function:

[0110] • proposal of an optimal plan for treatment stages using the selected method and online support of these stages;

[0111] • control of the quality of the service provided and control of the declared price of the service;

[0112] • monitoring of treatment results (radical / palliative);

[0113] • formation of a dataset for each completed case indicating the treatment results.

[0114] The proposed solution utilizes machine learning based on labeled, anonymized data from patients treated using the proposed service. Data will be stored using the PostgreSQL object-relational database management system and the Redis NoSQL resident database management system. Neural network algorithms written in Pytorch will be used to solve the tasks.

[0115] The relevance of the proposed solution is due to the lack of a digital solution in the state of the art that would offer the patient an optimal treatment method and an accessible clinic / doctor based on an intelligent (computer) analysis of the effectiveness of various treatment methods, by various clinics and doctors.

[0116] The doctor selection offered by medical information systems is currently based on ratings compiled from patient reviews, which often do not contain information about the patient's illness or treatment success, but are based on the patient's subjective impressions of the initial consultation. Repeatedly visiting doctors of various specialties, depending on the qualifications of randomly selected specialists for treatment, and wasting time significantly reduce the prospects for treatment and recovery.

[0117] This technical solution can be implemented on a computer, in the form of an automated information system (AIS) or a machine-readable medium containing instructions for performing the above-mentioned method.

[0118] The technical solution may be implemented in the form of a distributed computing system that ensures the implementation of the claimed method or is part of a computer system, for example, a server, a personal computer, part of a computing cluster that processes the necessary data to implement the claimed technical solution.

[0119] In general, the system comprises one or more processors, memory such as RAM and ROM, input / output interfaces, input / output devices, and a device for network interaction, all connected by a common information exchange bus.

[0120] The processor (or several processors, a multi-core processor, etc.) can be selected from a range of devices that are widely used today, for example, from manufacturers such as: Intel™, AMD™, Apple™, Samsung Exynos™, MediaTEK™, Qualcomm Snapdragon™, etc. Under the processor or one of the processors used in the system, it is also necessary to take into account the graphics processor, for example, NVIDIA GPU or Graphcore, the type of which is also suitable for the full or partial implementation of the method, and can also be used for training and application

[0121] ■ 35 machine learning models in various information systems. RAM is random access memory (RAM) and is designed to store machine-readable instructions executed by the processor to perform the necessary logical data processing operations. RAM typically contains executable instructions from the operating system and corresponding software components (applications, software modules, etc.). The available memory of a graphics card or graphics processor can also serve as RAM.

[0122] ROM is one or more permanent storage devices, such as a hard disk drive (HDD), solid-state drive (SSD), flash memory (EEPROM, NAND, etc.), optical storage media (CD-R / RW, DVD-R / RW, BlueRay Disc, MD), etc.

[0123] Various types of I / O interfaces are used to organize the operation of system components and organize the operation of external connected devices. The choice of appropriate interfaces depends on the specific design of the computing device, which may include, but are not limited to: PCI, AGP, PS / 2, IrDA, FireWire, LPT, COM, SATA, IDE, Lightning, USB (2.0, 3.0, 3.1, micro, mini, type C), TRS / Audio jack (2.5, 3.5, 6.35), HDMI, DVI, VGA, Display Port, RJ45, RS232, etc.

[0124] To ensure user interaction with the computing system, various I / O information means are used, such as a keyboard, display (monitor), touch display, touchpad, joystick, mouse, light pen, stylus, touch panel, trackball, speakers, microphone, augmented reality tools, optical sensors, tablet, light indicators, projector, camera, biometric identification tools (retinal scanner, fingerprint scanner, voice recognition module), etc.

[0125] A network communication tool enables data transfer via an internal or external computer network, such as an intranet, internet, LAN, etc. One or more of these tools may include, but are not limited to, an Ethernet card, GSM modem, GPRS modem, LTE modem, 5G modem, satellite communication module, NFC module, Bluetooth and / or BLE module, Wi-Fi module, etc.

[0126] A program is a sequence of instructions intended for execution by a computer control unit or command processing device.

[0127] In these application materials, a preferred disclosure of the implementation of the claimed technical solution was presented, which should not be used as limiting other, particular embodiments of its implementation that do not go beyond the scope of the requested scope of legal protection and are obvious to specialists in the relevant field of technology. and Literature

[0128] 1. Mintz Y, Brodie R. Introduction to artificial intelligence in medicine. Minimum Invasive

[0129] Ther Allied Technol 2019;28(2):73-81. https: / / doi.org / l 0.1080 / 13645706.2019.1575882.

[0130] 2. Hameed BMZ, Shah M, Naik N, Rai BP, Karimi H, Rice P., et al. The ascent of artificial intelligence in endourology: a systematic review over the last 2 decades. Curr Urol Rep 2021 ;22(10):53. https: / / doi.org / 10.1007 / sl l934-021-01069-3.

[0131] 3. Andresen SL. John McCarthy: father of Al. Intell Syst IEEE 2002;17(5):84-5. https: / / doi.org / ! 0.1109 / MIS.2002.1039837.

[0132] 4. Klot CA, Kuczyk MA. Kunstliche intelligenz und neuronale netze in der urologie. [Artificial intelligence and neural networks in urology]. Die Urologie 2019;58(3):291- 9. https: / / doi.Org / l 0, 1007 / s00120-018-0826-9.

Claims

Invention formula 1. A method for selecting a treatment method for patients, performed using a processor and containing the following stages: - receive depersonalized medical data of the patient such as: • patient’s gender, age, current symptoms, duration of the current illness, concomitant diseases, their severity, information from previous visits to medical care, treatment outcome; - process the obtained anonymized medical data of the patient using a trained machine learning model and, based on the processed data, extract information from a database of anonymized medical records about patients with a similar type of disease, the treatment method used, the achieved treatment result, and data about the clinic and the attending physician; - determine, from the information received about patients with a similar type of disease and data about the clinic and the attending physician, the criteria related to the indications and contraindications of the treatment methods used for each clinic and attending physician, and form a list of criteria for an algorithmic search in medical areas for the elimination of this type of disease; - determine what type of medical care was provided for a given type of disease - palliative or radical; - based on the identified criteria and type of medical care, as well as data on the clinic and the attending physician, a rating of clinics and doctors is formed and a treatment method is selected with priority towards a radical direction; in the event that radical treatment cannot be carried out due to non-compliance with the criteria, palliative treatment methods are offered; - send the patient the selected treatment method and a list of clinics and doctors in the patient's region in descending order of rating.