
Dr. Kirill Arzamasov
Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies of the Moscow Healthcare Department, Russia
Title: Modular Workflow Platform for Radiological Data Processing and Annotation
Abstract:
Background: Artificial intelligence (AI) has significantly advanced radiological ap-plications. However, the lack of standardized and systematic processes for dataset preparation and curation continues to pose a critical challenge. The heterogeneity of dataset formats, inconsistencies in annotation methodologies, and fragmented lifecycle management hinder the predictability of AI-based diagnostic service per-formance, complicating the identification of vulnerabilities and the strategic en-hancement of these services.
Objective: To address these limitations, we developed a modular platform designed to standardize and simplify radiological dataset management, from data retrieval to annotation and presentation.
Methods: The developed platform employs a modular architecture, allowing inde-pendent operation of specialized modules that can be easily integrated as needed. Each module sequentially processes outputs from preceding modules, covering data search and extraction, anonymization, annotation, and generation of annotated radi-ological datasets with standardized documentation. The platform directly integrates with radiology information systems (RIS) via standardized data exchange protocols, maintains a detailed registry of dataset metadata and textual study protocols, and provides reliable storage for annotated images and corresponding reports.
Results: Since implementation, the platform has successfully facilitated the creation of 70+ publicly accessible datasets specifically designed to support AI algorithm developers. Furthermore, 160+ annotated datasets were utilized in evaluating 100+ AI diagnostic services. Currently, the system manages 696 comprehensive datasets encompassing CT, MRI, and X-ray imaging modalities, totaling approximately 70 000 validated and annotated radiological studies. Additionally, the platform supports dataset preparation for quality assurance in AI solutions within the world's largest experiment, the "Experiment on the use of innovative computer vision technologies for medical image analysis and subsequent applicability in the healthcare system of Moscow."
Conclusion: The key innovation of our platform lies in its integrated approach to the complete lifecycle of radiological dataset curation (fig 1). This structured, modular system significantly enhances data quality and reliability, directly enabling more accurate evaluation and targeted improvements of AI-driven radiological diagnostics. This development can also be scaled and adapted to accommodate diverse medical imaging dataset preparation tasks. Future work will focus on enhancing functionali-ty, improving user experience, and making the platform more intuitive for working with datasets.
Biography:
Dr. Kirill Arzamasov (b. 1984) is a medical informatics expert with over 15 years of experience in functional diagnostics, healthcare digitalization, and artificial intelligence applications in radiology. He graduated with distinction from the Russian State Medical University in 2007 with a degree in Medical Cybernetics and completed a clinical residency in Functional Diagnostics in 2009.
In 2012, he defended his PhD thesis in Radiology Diagnostics and Therapy, focusing on arterial blood flow assessment under physical stress in patients with impaired cardiac pump function.
From 2009 to 2019, he served as a physician in functional diagnostics at the RZD Clinical Scientific Center. Between 2019 and 2021, he worked as a Senior Research Fellow, and from 2021 to 2022 as Head of the Sector for Medical AI Technologies Implementation at the Center for Diagnostics and Telemedicine. Since 2022, he has been leading the Department of Medical Informatics, Radiomics, and Radiogenomics at the same institution.
Dr. Arzamasov plays a key role in supervising scientific research within his department and is the principal investigator or leading executor in several government-funded R&D and experimental development projects.
In 2024, he successfully defended his doctoral dissertation in Medical Informatics and Public Health on the topic: "Artificial Intelligence Technologies in Mass Preventive and Diagnostic Imaging."
He is actively involved in academia: mentoring PhD students at Center for Diagnostics and Telemedicine and teaching graduate-level students in the Master’s program on Data Mining and AI in Healthcare at MIREA — Russian Technological University.
Dr. Arzamasov is the author of over 60 scientific publications, and has received 5 software copyright certificates and 40 database registration certificates.
His research interests include medical AI systems, radiomics workflows, digital diagnostics infrastructure, and preventive imaging strategies using intelligent decision support.