Building a National AI Radiology Ecosystem: The Moscow Model

Moscow’s health system embeds AI in radiology through digitalisation and URIS, linking imaging equipment, staff and workflows. Use spans from COVID-19 chest CT support to broader detection services and mammography double reading under compulsory insurance. The Center for Diagnostics and Telemedicine provides evaluation, trials and registration, and develops standards, datasets and a maturity matrix. MosMedAI extends access nationwide, including autonomous chest X-ray triage.

Key Points

• Moscow built URIS to connect imaging, staff and workflows for AI-enabled radiology.
• AI use expanded from COVID-19 chest CT to multi-pathology services and screening support.
• Mammography double reading combines AI triage with radiologist review in compulsory insurance.
• The Center runs trials, registration and standards for safe, scalable medical AI adoption.
• MosMedAI scales Moscow-developed AI services nationwide, including autonomous CXR sorting.

Moscow ranks among the world's first cities to integrate artificial intelligence (AI) into practical healthcare. In just a few years, the capital has progressed from pilot testing in computer vision to creating of a large-scale infrastructure that unites dozens of clinical modalities, hundreds of medical facilities and millions of imaging studies. This experience is now informing new diagnostic standards not only within the city but also at the federal level, while attracting international professional interest.

The advancement of AI in radiology represents not merely the targeted deployment of a software product, but a comprehensive transformation of the entire system, including infrastructure, regulatory frameworks, assessment methodologies, educational processes and clinical practice. Effective AI integration is feasible only with deep digital maturity across the sector and a robust quality-control system.

From Equipment Modernisation to the Creation of the Unified Radiological Information Service

The implementation of artificial intelligence projects in radiology became possible thanks to the consistent digitalisation of the healthcare sector. The first stage involved upgrading the material and technical base: outdated equipment was replaced with modern digital systems, ensuring high-quality images and the ability for centralised processing.

The next key step was the creation of the Unified Radiological Information Service (URIS), a large-scale information system that unites all digital diagnostic equipment, radiographer workstations and radiologists across medical organisations within the Moscow Healthcare Department. URIS ensures the centralised storage, transmission and interpretation of medical images, as well as the integration of artificial intelligence into daily clinical practice.

Following the full connection of all diagnostic devices in outpatient departments to URIS, a Reference Center was established in March 2019 at the Center for Diagnostics and Telemedicine of the Moscow Healthcare Department. This is the world’s first teleradiology centre in the public healthcare system, providing high-quality radiology services. Its expert team comprises leading radiologists, including PhD holders, who are active members of international professional associations. Operating 24/7, the centre interprets over 130,000 CT, MRI, mammography and X-ray studies each week. In addition, specialists provide methodological support and consult colleagues from other regions of Russia.

Radiologists have online access to the Unified Radiological Information Service within the Unified Medical Information and Analytical System (UMIAS), a digital repository that aggregates all medical images acquired in Moscow’s hospitals. Following each examination, patient images are automatically uploaded to the system, where they become available for review by Reference Center specialists.

Healthcare is a key sector for implementing innovative technologies, particularly neural networks. Based on the medical images stored in the system, artificial intelligence algorithms were developed that now help doctors reliably detect pathologies in internal organs.

Moscow has been actively deploying AI-based technologies in its healthcare system since 2020. This work is conducted within the framework of an experiment focused on integrating computer vision into radiology.

The goal of the Moscow experiment was to explore the potential of applying a clinical decision support system based on medical data analysis, using advanced innovative technologies within the healthcare system.

In fact, the key task was a comprehensive assessment of the applicability, safety and quality of artificial intelligence technologies, primarily computer vision algorithms, in radiology. An additional strategic goal was to create a new market segment for digital medical technologies and stimulate the development of in-house research and development.

The Moscow experiment became a unique intersection of several key directions in the development of healthcare and digital technologies.

It is important to emphasise that the project was not an isolated initiative. Its implementation was based on the long-term activities of the Center for Diagnostics and Telemedicine. At the same time, the scientific foundation of the experiment was shaped by the work of Russian researchers and by accumulated practical experience in medicine.

The Moscow Social Development Complex and the Department of Information Technology launched the project jointly. The Center for Diagnostics and Telemedicine has become a platform for developing artificial intelligence technologies in Russia, as well as a source of support for Russian developers.

AI in COVID-19 Diagnostics

The first large-scale application of artificial intelligence was the interpretation of CT images to detect signs of COVID-19. From January to April 2020, specialists at the Center for Diagnostics and Telemedicine conducted accelerated testing of computer vision algorithms capable of recognising signs of lung damage in CT scans from patients with suspected coronavirus infection. By May 2020, the use of AI had become the standard for diagnosing COVID pneumonia in Moscow (Morozov et al. 2021).

The process of integrating the technology into clinical practice was designed to be as seamless as possible. For example, a patient undergoes a chest CT scan at a clinic; the resulting image is automatically uploaded to URIS and sent to artificial intelligence algorithms for analysis. Neural networks generate an additional image series with colour-coded areas of possible pathological change, along with a report. This data is returned to the system and becomes available to the radiologist, who uses it to interpret the examination. The completed report is saved in URIS and is immediately available to the treating physician and the patient in the electronic medical record.

Thus, AI acts as an assistant, increasing the accuracy and speed of diagnosis, while the final decision always rests with the physician.

Expanding Clinical Applications

Despite the ongoing demands of the pandemic, technological progress continued apace. Between 2020 and 2022, the application of computer vision expanded to include the detection of breast cancer, lung pathologies and other diseases.

A major milestone was the development of the first comprehensive service for chest CT, capable of simultaneously identifying multiple pathologies in a single scan. This multifunctional approach aligns more closely with the diagnostic needs of radiologists in clinical practice, offering a holistic view of a patient’s condition in a single analytical pass.

Currently, Moscow radiologists have access to more than 60 artificial intelligence services across 43 clinical modalities, including pneumonia, lung cancer, coronary heart disease, osteoporosis, hydrothorax, breast cancer and other conditions. The healthcare system also employs more than 20 comprehensive solutions that identify 2 to 14 pathologies per image, including early signs of cancer. Since the project’s inception, these AI algorithms have processed nearly 18 million imaging studies, affirming the technology’s reliability and scalability.

Clinical Trials and AI Registration

In 2020, the Center for Diagnostics and Telemedicine was authorised to conduct clinical trials of medical devices under the national certification system. In 2022, the Center expanded into the Eurasian Economic Union (EAEU) market, gaining the capacity to perform clinical and technical trials of medical artificial intelligence systems for their registration across the EAEU.

Today, the Center offers developers comprehensive, full-cycle services, from technical and clinical trials to assistance in obtaining registration certificates under both national procedures and EAEU regulations. These registration certificates substantially enhance the capabilities of medical device manufacturers by granting access to a wider market.

It is important to note that, before any medical device can be implemented in clinical practice, it must undergo state registration, which includes rigorous technical and clinical trials. These requirements also apply to artificial intelligence systems, ensuring their safety, efficacy and readiness for integration into routine medical care.

AI in Compulsory Health Insurance: Double Reading of Mammography Screening

In 2023, artificial intelligence advanced beyond experimental use and was integrated into practical healthcare as a service within the compulsory health insurance system (Vasilev et al. 2023).

One of the most significant applications has been the double reading of breast cancer screening. Images are now analysed in two stages, first by AI, followed by a radiologist. Over the past five years, specialists have analysed more than 3 million mammograms within the compulsory health insurance system, with over 1 million undergoing double reading assisted by artificial intelligence. This approach substantially increases the likelihood of detecting tumours at an early stage, when treatment is most effective.

This technology serves as a vital assistant to radiologists, accelerating report preparation and enhancing diagnostic quality while preserving free access for patients. According to research conducted by Moscow scientists, the integration of AI accelerates mammogram interpretation by more than eightfold. This efficiency gain has significantly reduced the workload on specialists and minimised the risk of missed pathologies, as the algorithms demonstrate diagnostic accuracy comparable to that of experienced radiologists.

Equipment upgrades have further boosted efficiency: all adult clinics now have modern mammography machines. Examination volumes have risen accordingly, with early-stage breast cancer detection reaching 80%. This achievement stems directly from process digitalisation and intelligent deployment.

Market and Scientific Ecosystem Growth

The computer vision implementation experiment laid the groundwork for the emergence of a fully operational market for AI services in radiology. This initiative fostered a competitive landscape in Moscow, where developers must now navigate rigorous, multi-stage evaluations of their solutions’ quality and clinical efficacy.

The project has developed unique scientific methodologies for evaluating algorithms. Over 300 benchmark datasets have been prepared, and the first official library of healthcare datasets in the Russian Federation has been created.

The development of the regulatory framework is of particular importance. The Center for Diagnostics and Telemedicine has initiated and implemented standards governing the use of artificial intelligence in medicine. These standards define protocols for the technical and clinical trials of neural networks in radiology. Only upon successful completion of these rigorous evaluations does an AI service receive a registration certificate, authorising its deployment in medical organisations. Subcommittee 01 of Technical Committee 164, “Artificial Intelligence in Healthcare”, operates within the Center for Diagnostics and Telemedicine. Its activities encompass developing strategies, procedures and rules for standardisation under SC01/TC164, including organising and conducting work within national, interstate, regional and international standardisation frameworks, in accordance with legal norms.

Moscow experts’ contributions have earned national recognition: the Center received the “Standardiser of the Year” award twice, in 2022 and 2025, for its practical role in developing standards of high social significance. To date, the Center for Diagnostics and Telemedicine of the Moscow Health Department has developed 28 national standards regulating the use of artificial intelligence in healthcare, six of which entered into force in early 2026. These documents establish a coherent regulatory framework for AI in medicine, setting industry benchmarks. They ensure transparency in technology application and form the basis for patient safety.

Work in this area continues actively. The next frontier involves developing national standards for generative AI. As generative AI is a focal point for scientists worldwide, the creation of a national standard in this domain signifies a substantial advance in both technological development and implementation regulation.

Notably, Moscow has been developing standards for medical AI algorithms for several years, a task that remains challenging for many countries. Documents from the International Organization for Standardization (ISO) lack direct equivalents to the solutions proposed by Moscow specialists. Thus, Moscow experts are shaping not only the Russian but also the global agenda in medical artificial intelligence. Patient health and safety remain paramount: neural networks have become reliable tools in physicians’ hands, with standards helping to guarantee their quality.

Maturity Matrix: An Assessment Tool

The primary challenge in deploying artificial intelligence in healthcare remains selecting mature, reliable solutions. Drawing on performance monitoring during the Moscow experiment, researchers developed a methodology for evaluating AI service maturity, the “Maturity Matrix” (Center for Diagnostics and Telemedicine 2026).

Scientists at the Center for Diagnostics and Telemedicine analysed the performance of participating services and published a maturity matrix incorporating technical stability metrics (technological defect rates) and diagnostic efficacy (area under the ROC curve). This tool enables medical organisations to objectively identify optimal AI-powered algorithms, while allowing developers to monitor their progress and market competitiveness (Tyrov et al. 2022).

Federal Expansion: The MosMedAI Platform

In 2024, following a presidential directive, the federal digital platform MosMedAI was established to provide medical organisations across the country with access to Moscow-developed artificial intelligence services for radiology.

The platform supports the interpretation of CT, MRI and other scans: algorithms identify areas of potential abnormality using colour coding, generate reports and perform required measurements. Throughout this process, the physician’s fundamental role remains unchanged, as the final clinical decision rests entirely with the specialist.

By 2025, the platform had processed 8.6 million studies. Currently, over 2,000 health facilities across 74 regions are connected to the system, providing access to 17 leading AI services that underwent rigorous, multi-stage expert evaluation and demonstrated effectiveness in Moscow’s clinical practice.

The implementation of the platform has enhanced diagnostic access in the regions, helping to bridge the technological divide between the federal centre and Russia’s constituent entities.

Autonomous AI in Moscow Healthcare

Since May 2024, autonomous artificial intelligence algorithms have analysed chest X-rays in Moscow, following a decree signed by the Mayor of Moscow.

Under this protocol, AI automatically sorts chest X-ray (CXR) images into two categories: “normal” and “abnormal”. The accuracy of determining normality approaches 100%. This implementation represents the world’s first example of large-scale autonomous artificial intelligence deployment within a real-world healthcare system.

The process operates as follows: the attending physician issues a prescription for a chest X-ray. The patient undergoes a scan at their local clinic, and the image is automatically routed to the AI for classification as “normal” or “abnormal”.

For “normal” results, the report is uploaded directly to the EHR and is accessible to both physician and patient. “Abnormal” images proceed to a radiologist for review, with a final report delivered within 24 hours. Notably, an “abnormal” classification does not confirm the presence of disease: the radiologist makes the definitive determination following detailed image analysis.

Driving Innovation: Advancing Radiology and Instrumental Diagnostics in Clinical Practice

The Center for Diagnostics and Telemedicine, a leading research and practical organisation within the Moscow Department of Healthcare, serves as the coordinating and scientific hub for the digital transformation of radiology.

The Center's responsibilities encompass management of radiology and instrumental diagnostics departments, implementing artificial intelligence technologies, developing regulatory standards, conducting scientific research and training healthcare professionals. Since 2013, the Center’s staff have produced over 800 scientific publications, guidelines, monographs and manuals, and have registered more than 200 intellectual property rights.

Strategic Development Pathway

Key projects, the computer vision implementation experiment, the MosMedAI platform and the AI service maturity matrix, form a unified ecosystem that enables not only the application of technology but also the systematic evaluation of its effectiveness, safety and impact on the quality of medical care.

The Moscow model demonstrates that AI in healthcare functions as a practical tool integrated into everyday clinical practice. The sequential progression, from infrastructure digitalisation to the creation of a regulatory framework and federal-level scaling, serves as a blueprint for advancing medical technologies in Russia and internationally.