Olga Magomedova, Researcher of the International Best Practices Analysis Department at the Gaidar Institute, commented for Nezavisimaya Gazeta on the problem of AI introduction to medicine. In an article devoted to a World Economic Forum study on the rise in medical errors related to use of AI, the expert identified the main cause of the challenges, i.e. poor-quality data for training algorithms.
According to Olga Magomedova, the effective use of AI-based technologies is possible "only if developers are provided access to large, representative health databases."
"The data needed to train the AI consists of individuals' health information collected by medical organizations over a specific period. This information is protected by personal data laws and medical confidentiality regulations. In practice, this means that developers can only request medical data anonymously and only if patients have consented to use them," explained Olga Magomedova.
The expert noted that developers are forced to collect databank from each medical organization separately and spend resources on additional dataset preparation. However, "even with such efforts, the collected information is often of insufficient quality, incomplete, irrelevant, or poorly structured (for example, lacking age stratification of patients)," she noted.
Olga Magomedova proposed creating a mechanism for centralized access to medical data that would function on a permanent basis to address a challenge. She emphasized that Russia already possesses the technical basis due to the Uniform State Health Information System (USHIS).
"Therefore, the infrastructure is essentially ready for Russian AI developers to obtain the high-quality data required for training algorithms," the expert concluded.