ROSA Virtualization provided virtualization of video cards for AI in Research Institute of ENT

2024: Introduction of AI for setting up Esculap-KI cochlear implants

The St. Petersburg Research Institute of Ear, Throat, Nose and Speech (SPb NII LOR) has developed and implemented an artificial intelligence (AI) system for setting up Esculap-KI cochlear implants. Using neural networks, this technology automatically predicts the settings of the implant audio processor, providing more than 95% accuracy. To effectively allocate computing resources during the work of AI, the institute deployed a domestic virtualization system ROSA Virtualization, which made it possible to increase the number of users who simultaneously use computing power. This was announced by STC IT ROSA on January 29, 2025.

In many AI solutions, the entire memory of one video card is allocated to only one user, even if his tasks occupy only 10% of the total. Our system allows you to allocate resources so that several users can simultaneously work with one video card. This significantly increases the efficiency of equipment use and speeds up tasks. Without such flexibility, the process would be more complicated and costly, "said Vyacheslav Kadomsky, director of strategic development at STC IT ROSA, a virtualization platform developer.

The hardware and software complex (PAC) of St. Petersburg Research Institute LOR, on the basis of which AI operates, includes server equipment from the Russian manufacturer Delta Computers, as well as two Nvidia A100 video cards with 80 GB of video memory each. Thanks to ROSA Virtualization, flexible distribution of graphics resources between users has become possible, which is critical for working with medical AI algorithms.

The Esculap-KI AI system greatly simplifies the process of setting up cochlear implants. Previously, specialists took several days or even weeks to manually select the optimal parameters of the audio processor for patients. Now neural networks analyze telemetry data and form accurate settings in 10-15 minutes, providing faster adaptation of patients to hearing devices. This reduces rehabilitation time, reduces the burden on specialists and increases the quality of auditory adaptation of patients.

The use of artificial intelligence makes it possible to significantly speed up the process of setting up cochlear implants, which is critical for patients. Now all testing and forecasting of processor parameters take no more than 10-15 minutes instead of several days of traditional tuning, "said Vladimir Dvoryanchikov, director of St. Petersburg Research Institute of ENT, professor, doctor of medical sciences.

Thus, the use of ROSA Virtualization ensured the efficient allocation of resources and increased the performance of the AI system, which allowed not only to speed up the process of setting up cochlear implants, but also to expand the possibilities of multi-user work with medical neural networks.