Vietnam’s VinBrain gives healthcare sector a leg-up with AI-infused solutions

[HO CHI MINH CITY] For any country’s healthcare system, the availability and reliability of data-driven tools can go a long way towards easing the strain on hospitals that are short on manpower, and help them diagnose and treat their patients in a faster and more accurate manner.

At VinBrain – a four-year-old healthcare startup backed by Vietnam’s largest private conglomerate, Vingroup – researchers and scientists have developed artificial intelligence (AI) solutions based on health data from millions of Vietnamese people.

These tools have helped doctors and medical professionals in more than 100 hospitals in Vietnam, Myanmar, the US and New Zealand detect abnormalities in a patient’s heart, lungs and bones, via an AI analysis of chest X-rays.

VinBrain’s flagship product, a software called DrAid, supports radiologists in detecting diseases in scans of patients in South-east Asia and the US, where it has the most representative data to train AI models.

Industry players say there is lots of merit to having a wider data set when it comes to medical diagnoses. A 2020 study by a group of British researchers found that almost all the images in data sets used to train AI algorithms to diagnose eye conditions came from patients in North America, Europe and China.

This disparity in the source of data sets, dubbed “data poverty”, in the development of AI tools might cause biases and widen health inequalities, they said.

Nguyen Do Trung Chanh, the director of VinBrain’s applied science division, said it is important to acquire representative data from diverse population groups with varying biological and physiological characteristics.

Similar to other AI tools in the market today, DrAid’s algorithms were also initially trained on publicly available data sets from the US and Europe.

But the team at VinBrain went one step further and fed the system with chest X-ray images of thousands of Vietnamese adults. These images were acquired from Vingroup-owned hospital chain Vinmec and the National Lung Hospital.

VinBrain chief executive officer Steven Truong said in January that nearly 2,000 doctors are using DrAid at hospitals across Vietnam. The data sets of some 2.5 million images are updated and enlarged daily, with 70 per cent coming from Vietnamese citizens.

Dr Chanh said VinBrain’s strategy is to develop machine-learning models that are trained with large data sources in Vietnam, and then improving the accuracy by adding in representative samples from elsewhere.

Going beyond Vietnam

In September 2022, DrAid became the first AI tool for chest X-ray diagnosis in South-east Asia to be approved by the US Food and Drug Administration (FDA), with this certificate enabling medical equipment distributors and manufacturers to export the software around the world.

Apart from Vietnam, it has been reported that only the US, the UK, Israel, Australia and South Korea have companies that successfully developed FDA-approved AI products of this type.

In February, Stanford University signed an agreement with VinBrain to share the former’s 240,000 anonymised medical images and reports, to help improve VinBrain’s AI model training and testing.

A month before that, VinBrain announced a partnership with Microsoft in data sharing, product cross-validation, as well as research and development.

Dr Chanh said VinBrain aims to evolve into a comprehensive platform through partnerships with other AI projects and companies in Vietnam and overseas.

“Given the wide range of AI solutions for diagnostic imaging, it is not feasible for any company to address every challenge independently in a short amount of time,” he said.

“It is anticipated that different AI models and firms will come together and integrate within a marketplace or app store, offering hospitals a comprehensive range of products in the near future.”