April 3, 2020, The Federal Service for Surveillance in Healthcare (Roszdravnadzor) registered CDSS Webiomed as a medical device. Webiomed became the first software product with artificial intelligence that passed technical and clinical trials and received permission for medical use in the Russian Federation. The main focus of Webiomed is disease risk management.
Chronic non-communicable diseases (NCDs) are the leading cause of morbidity and mortality in all countries around the world. They place the main burden on national health systems. Cardiovascular diseases (CVD) are the largest category in the structure of NCDs. About 850 thousand people die from CVD annually in Russia. It is every 3rd death accounts for CVDs.
A set of preventive measures against NCDs is implemented in Russia, including conducting an annual medical examination, provision of a standardized set of surveys and inspections specialists, early detection of NCD risk factors. During screening some problems worsen the final result: obtaining and accounting for the completeness of medical data requires a lot of time, the accuracy of the applied risk assessment methods is low and their correct interpretation can be difficult. Moreover, medical appointments are often overloaded which makes it difficult to correctly interpret complete and detailed medical information. As a result, the full risk stratification of the patient is not carried out and the available medical data is not used for the prevention of diseases. Therefore, the health care system begins to deal with the patient only when the disease has already manifested itself, which often makes the prognosis heavier and requires higher treatment costs.
Webiomed, a Russian clinical decision support system (CDSS) is a predictive analytics and risk management software for patients. It is the product that will help to bring the effectiveness of disease prevention and diagnosis to a new level. The system is trained to analyze various medical data of the patient, identify risk factors and suspected cases of diseases, form forecasts based on them, containing a comprehensive assessment of the probability of developing various diseases and death of the patient from them. It offers personal recommendations for doctors and patients. The system is based on its own comprehensive methodology for the determination of risks of disease and complications development, including the implementation of machine learning models and several auxiliary algorithms based on clinical recommendations and risk assessment scores.
«We have the aim to create an effective service that can help healthcare providers and doctors automatically identify high-risk patients for prediction of various health conditions and development of diseases in the future. Our team truly believes that the future of medicine is in application of the best possible ways to prevent diseases rather than fighting their menifestations, when there is little that can really be changed. We rely on artificial intelligence and deep machine learning technologies as a key way to succeed in this goal. For this purpose, a cool team of experts in the field of medicine, machine learning, software development and market launch was assembled,», said Alexander Gusev, Chief Business Development Officer.
The scientific part of the project is supervised by the General Director of "SMRC Cardiology" Ministry of Health of Russia, academician of RAS, Professor, Ph.D. in Medical Science Sergey Boytsov.
Scientific consultants are the leading scientists of Petrozavodsk State University. Tatiana Kuznetsova, Ph.D., Head of the Department of faculty therapy, phthisiology, infectious diseases, and epidemiology, is an advisor of the project on medical issues. Alexander Rogov, Ph.D., Professor, Head of the Department of probability theory and data analysis is a consultant of the project as well.
"
The implementation of modern information technologies in daily work of doctors is very important today. First of all the doctors needs to communicate with patients who asked for help and а computer should be a useful assistant for him. We are already used to keeping that doctors save medical records in electronic form but developments in the field of artificial intelligence have not yet been widely introduced into our clinical practice. At the same time, we have no idea what huge opportunities opens this technology for us.
Webiomed clinical decision support system helps doctors in assessment of the risk of cardiovascular disease at early stage when it has not yet been clinically manifested as well as helps in assessment of the risk of existing cardiovascular diseases, without additional research and time-consuming analysis of information. A phycisian needs to ask Webiomed the question by clicking on the button. After that, the CDSS is searching for information itself and gives the results according to which it will decide which prevention and treatment program to assign to the patient. The very important aspect of using artificial intelligence in medicine is analytical work such as the ability to analyze the epidemiological prevalence of specific risk factors. This is important for improving efficiency of regional health care, " said Tatyana Kuznetsova, Research Medical Director of the project.
Assessing the goal of the system creation and development, the chief freelance specialist in radiation and instrumental diagnostics, director of the Scientific and Practical Clinical Center for Diagnostics and Telemedicine Technologies of the Moscow Department of Health, professor at the Department of Radiation Diagnostics and Radiation Therapy, First Moscow State Medical University named after I.M. Secheno» of Ministry of Health of Russia, PhD Sergey Morozov reported :
“Clinical decisions support systems are irreplaceable assistants to a phycisian. In fact, they are safety systems for healthcare, a kind of autopilot. CDSS are not directories that are referred to if necessary. On the contrary, they work in the background, scan medical data and attract doctor's attention to patients with a high risk of stroke, with signs of cancer on CT skan, with errors in prescribing drugs. In contemporary medicine, with its huge streams of digital data, automation is indispensable and the key to ensuring timely diagnosis. At the same time, conducting clinical trials of CDSS allows us to guarantee the reliability of their operation and safety for use in any clinics. The development team of this system acted as a pioneer and paved the way for many manufacturers of CDSS which will make healthcare more accessible, transparent, reliable and useful for all patients.”
Webiomed can be integrated into medical information systems and other software products for healthcare management. It helps physicians to get rid of manual identification of the necessary information in electronic health records and provides a ready-made disease risk assessment, such as the total risk of developing atherosclerosis and its complications, the risk of thromboembolic complications in heart rhythm disorders, the risk of cardiac arrest in hospitalized patients, the severity of community-acquired pneumonia, etc. Based on these assessments, the system makes recommendations for determining patient management tactics.
Webiomed uses machine learning models:
1. Prediction of an individual probability of CVD development based on machine learning for 10 years;
2. Prediction of the individual probability of death from coronary heart disease and stroke based on machine learning for 10 years.
To build the "Forecast of the individual probability of developing CVD" model, the features used in the Framingham score were taken. An artificial neural network was used as a machine learning method. As a result, the model estimates the probability of developing CVD cases in a patient over the next 10 years. The model provides the following parameters: accuracy: 78%, area under the ROC curve (AUC): 0.77. As a result of training, the predictive accuracy of the Webiomed model was 19 positions higher compared to the accuracy of the Framingham score..
We collect the signs used in the most common SCORE risk score which were supplemented with 2 more clinical indicators (body mass index, heart rate) to build the "Forecast of the individual probability of death from CHD and stroke" model. An artificial neural network was used as a machine learning method. As a result, the model estimates the probability of developing CVD cases in a patient over the next 10 years. The model provides the following parameters: accuracy: 78%, area under the ROC curve (AUC): 0.77.
During preparation for registration of the system as a medical device, technical and clinical trials were successfully completed. During technical tests, it was confirmed that the medical device Clinical Decision Support System Webiomed according to technical SPECIFICATIONS 62.01.29-001-12860736-2019 performs the stated functions in accordance with the operational documentation when used for the intended purpose provided by the manufacturer and meets the requirements of current national standards: GOST R ISO/IEC 12119-2000, GOST R ISO/IEC 9126-93, GOST R 51188-98, GOST R IEC 62304-2013, requirements for technical and operational documentation of the manufacturer.
Successful clinical trials have proven the safety and effectiveness of the Webiomed based on real medical data obtained in the Russian Federation and not used for machine learning. For this purpose, experts in cooperation with the developer company created a special test method. As a result of this work, it was concluded that the principles of operation, clinical action, ergonomics, and safety, as well as artificial intelligence technologies used in Webiomed, correspond to the modern level of medical devices and meet the needs of clinical specialists, harmless for patients.
After comprehensive inspections, Webiomed was registered by Roszdravnadzor as a medical device in the 1st class of potential risk of use in accordance with the nomenclature classifier of medical devices approved by order No. 4n of the Ministry of the health of the Russian Federation dated June 06, 2012. Webiomed safety class in accordance with GOST R IEC 62304-class A. The system received a Registration certificate no. "RZN 2020/9958", a unique number of the registry entry 41741 in the state register of medical devices, is available at Roszdravnadzor (in Russian).
"The registration of our product Webiomed as a medical device is the result of systematic and well-coordinated work of the entire project team and a large number of partners in pilot projects and, of course, doctors, who helped us in every way with advice, criticism and the desire to make a sound and useful tool for healthcare. We were looking forward to this event and finally we can say with confidence that now our team has a very good reason to be proud of the excellent result. Now we are preparing to attract significant investments in order to focus our efforts on the development and promotion of Webiomed implementation projects in practical healthcare, " said Roman Novitsky, CEO of Webiomed.
Pilot implementation of the system took place in the Yamal-Nenets Autonomous district, the Kirov region, and several medical organizations in the Republic of Karelia. During this time, it was used to analyze more than 1 million electronic health records, including a retrospective analysis of medical examination data, comparative studies of the accuracy of risk assessment by doctors and artificial intelligence and collected, analyzed feedback from doctors who asked for a "second opinion" from CDSS. Pilot projects have confirmed the practical significance of the service: the system is really able to relieve the physician from routine analysis of medical data and more accurately identify risk factors and high-risk patients, which helps to focus the work of medical organizations on preventive personalized medicine and thus reduce the burden of cardiovascular diseases.