Roszdravnadzor suspended the use of a medical system with AI due to possible harm to the health of patients

For the first time, Roszdravnadzor suspended the use of a medical product with artificial intelligence (AI) – the Botkin.AI analysis system, which allows doctors to detect pathologies in computed tomography images. The department considered that the development, created through investments from a pool of venture funds, including Rosatom, the Ministry of Industry and Trade, R-Pharm, Tashira, could “cause harm to the health” of patients. Three years ago, the developer of the system, Intellogik, announced its integration into medical institutions in Moscow, but the mayor’s office claims that they do not use it. To train AI to identify pathologies, serious technical solutions are required, which clinics are not yet ready for, experts say.

Roszdravnadzor suspended the use of Botkin.AI, a system for analyzing computed tomography images using AI, which received a registration certificate in 2020, “due to the threat of harm to the life and health of citizens.” Kommersant found the corresponding letter dated November 8 in the department’s register. Roszdravnadzor told Kommersant that the manufacturer is currently being checked. Intellogik, the developer of Botkin.AI, did not respond to Kommersant’s request.

Back in early November, the head of Roszdravnadzor, Alla Samoilova, stated at the Novamed forum that the agency for the first time suspended the circulation of medical devices with AI due to the “lack of clinical effect and repeatedly identified defects in functioning.” She did not specify the name of the product then.

Intellogik raised 100 million rubles for the development of the Botkin.AI project in 2019. from a number of investors, including Digital Evolution Ventures, created with the participation of Rosatom, RBV Capital (the R-Pharm fund and the Russian Venture Company), and in 2020 – another 160 million rubles, including from Tashir Medica » Samvel Karapetyan and Unicorn Capital Partners, the management company of the venture fund of the Ministry of Industry and Trade.

Tashir Medica reported that they are aware of the suspension of Botkin.AI and are looking into the reasons for this. Unicorn Capital Partners explained that Intellogic is “in the process of technical reorganization,” which explains the suspension of the regulatory certificate. They stated that any threat of harm to the life and health of citizens by the system is “completely excluded.”

Products with AI belong to a high, third class of risk, which is why Roszdravnadzor pays such close attention to them, explains Medleveltech CEO Mikhail Vinogradov. According to him, in fact, the suspension of use may last until the manufacturer eliminates all the regulator’s comments.

Meanwhile, back in 2020, Intellogik stated that Botkin.AI was integrated into the city’s Unified Radiological Information System created by the Moscow City Hall to help identify lung cancer.

The Moscow Department of Health (DZM) assured Kommersant that they do not use Botkin.AI in the capital’s clinics. They added that today, as part of the experiment, over 50 AI services are working, identifying signs of pathologies in 28 clinical areas. In particular, DZM uses the Celsus medical image analysis system, the Third Opinion platform, and AIDiagnostic.

The Ministry of Health of the Tula Region, where Botkin.AI was tested in three medical institutions, told Kommersant that the system is not currently used. “During testing, a number of shortcomings were identified, feedback was passed on to the developers,” the regional Ministry of Health reported. In the Open Clinic network, according to its founder Philip Mironovich, the use of Botkin.AI “has been paused pending clarification from the developer.”

Third Opinion, Carementoryay and Cyberia have similar AI-based products registered by Roszdravnadzor. General Director of Third Opinion Anna Meshcheryakova, citing government procurement data, says that since the beginning of 2023, more than 15 regions have already purchased medical products with AI to help decipher computed tomography of the chest organs, but Botkin.AI is not among the supplied systems.

In medicine, only the first steps are being taken to train AI; this is the most complex process with a large number of parameters for analysis, notes Irina Ostroukhova, director of partner relations at Rekfaces. To implement AI in practice, serious technical solutions are needed, which not all medical institutions are ready for, and the shortage of personnel in this area also affects, adds Dmitry Roslyakov, general director of the online image interpretation service “Honest Medicine”. Mikhail Vinogradov sees the main problem as the fact that AI has to learn, among other things, from images where the doctor did not highlight the pathology, did not comment on the data in any way, and did not confirm the diagnosis.

Polina Gritsenko, Timofey Kornev