17-06-2022 | Adlink | Subs & Systems
With its MLB-3000 and MLB-3002 systems, ADLINK Technology presents two new real-time capable medical computing platforms for AI-supported imaging processes. The IEC 60601-1 and IEC 60601-1-2 compliant systems may be configured with 13 Intel Core and Intel Celeron processor configurations and nine high-performance NVIDIA Quadro and RTX processor-based graphics cards up to the NVIDIA RTX GPUs of the Ampere family. AI-supported imaging processes are used as recording, processing and visualisation systems in which image data are increasingly analysed and visualised in real-time. Areas of application range from CT and MRI to X-rays and ultrasound to endoscopy.
The processor boards and the graphics cards come from the company's development and production, so customer support for all components is provided from a single source.
"Imaging procedures contain an increasingly high level of detail. 2D layer models and 3D visualisations increase the performance requirements, and the integration of artificial intelligence further markedly expands the requirements of the image-processing systems. Simultaneously, it is required to provide imaging in real-time. On the other hand, the system design must be balanced concerning costs and energy consumption. With our MLB-300x series, we offer medical device manufacturers, system integrators and hospitals a modular system so that they can optimally design their imaging tasks," explains Matthias Lubkowitz, director of the Healthcare Business Center EMEA at ADLINK Technology in Deggendorf.
In imaging processes in medical technology, the new systems impress with their flexible GPU equipment, which is implemented utilising an MXM slot. Standard configurations with MXM graphics modules are offered with NVIDIA Quadro graphics cores of the Pascal, Turing and Ampere generations. The MLB-3002 systems also have two further PCI Express slots for dedicated frame grabbers or other graphics cards.
In MXM modules based on the Ampere and Turing architecture, ray tracing, which is essential for 3D volumetric imaging, is based on hardware accelerators called RT Cores for the first time. Therefore, rendering image data is possible in real-time for the first time. The onboard Tensor Cores also speed the matrix calculations at the heart of deep learning neural network training and inference operations, enabling artefact removal, contrast adjustment, and sharpness enhancement for clearer medical images.
