Artificial intelligence (AI) technology is already beginning to have an impact in medicine, but there’s certainly a lot more that we should be expecting from it in the future. There are a number of issues that prevent a wider adoption of AI in medicine, a lot of them technological in nature. For example, medical systems can produce vast quantities of data and a lot of the data may be coming from old, slow systems that were never expected to interface with modern computer platforms. Moreover, there aren’t a lot of tools available for medical technology developers to harness the power of AI

In order to overcome these limitations, NVIDIA is releasing a specially built platform specifically geared toward implementing AI in medical devices such as imaging systems. It consists of the NVIDIA Clara AGX, an architecture developed from the company’s Xavier AI module and Turing graphics processing units, plus a software development kit (SDK) that lets programmers write artificial intelligence programs to use data coming from legacy medical systems.

In its announcement NVIDIA says that in order to process gigabytes of data per second for AI applications, a combination of general purpose processors, graphical processing units (GPU), and field programmable gate arrays was required. The new Clara platform effectively combines these into a GPU-based system. The company claims that this is the “world’s fastest AI inferencing on NVIDIA Tensor Cores”, and it can be scaled to different developer needs.

Additionally, the Clara can process raw data coming from existing medical instruments, which effectively allows all kinds of medical devices to benefit from AI techniques. NVIDIA says that methods such as iterative reconstruction for CT and X-ray images, beamforming in ultrasound, and compressed sensing in MR, can be run on devices that are many years old already using the Clara.