A Vietnam-based health-tech startup is using AI analysis of chest X-rays to make automated observations of abnormalities in a patient’s heart, lungs and bones.
Doctors can combine the x-ray analysis with a large language model to read the patient’s health record, perhaps detecting a fever to help make a pneumonia diagnosis.
The AI model work of the startup VinBrain is being deployed in more than 100 hospitals in Vietnam, Myanmar, New Zealand and the US. VinBrain’s AI software, called DrAid, is being applied to medical images for more than 120,000 patients each month, according to an Nvidia blog.
The US FDA has cleared DrAid for detecting x-ray features that could indicate collapsed lungs. It was trained on a dataset of 2.5 million images and relies on Nvidia AI Enterprise, made up of Nvidia Clara and Monai, an open source medical imaging dev tool, as well as Nvidia NeMo, a conversational AI toolkit. VinBrain also uses Nvidia DGX SuperPOD for AI training and Nvidia GPUs for greater efficiency and deployment. Those GPUs include Nvidia Triton inference server and Nvidia TensorRT based on Tensor Core GPUs.
VinBrain also recently entered a formal collaboration with Microsoft in AI healthcare on data sharing, product validation and R&D.
Switching to GPUs from CPUs, engineers were able to speed up inference jobs for medical imaging AI by 3x and video streaming by 30x, according to Nvidia. Use of DGX SuperPod for training has been 100x faster than CPU-only training.
VinBrain is funded by Vingroup of Vietnam, which is also the conglomerate behind an emerging line of electric vehicles from VinFast, which made an appearance at CES 2023 in Las Vegas and has filed to go public. VinGroup was founded by Pham Nhat Vuong, one of the richest men in Vietnam.
VinBrain has developed more than 300 AI models to process speech, text, video and images. The company is a part of the Nvidia Inception accelerator.
RELATED: Vietnam’s VinFast, Canada’s Project Arrow show electric cars at CES 2023.