Vietnamese health tech firm develops AI model for TB diagnosis, screening

The model showed 86% sensitivity and a specificity of up to 96.1%.
By Adam Ang
11:11 pm
Share

Credit: VinBrain

VinBrain, an artificial intelligence healthcare software developer in Vietnam, recently came up with an AI model for the diagnosis and screening of tuberculosis.

The Vingroup-backed company has worked with the German non-profit group Friends for International Tuberculosis Relief in writing a white paper discussing the utilisation of AI in TB diagnosis and screening in the country.

WHY IT MATTERS

The partners wrote the report to help improve the diagnosis and treatment of TB in Vietnam. It was revealed in the joint report that VinBrain's AI model v1, which was launched on 5 January, showed 86% sensitivity and a specificity of up to 96.1% in reading chest x-rays for TB diagnosis.

VinBrain has developed the TB diagnosis model through Microsoft's cloud computing platform Azure. Initially, over 500,000 TB x-ray images were culled from Vietnam, India and China. After filtering poor-quality images, around 185,000 were utilised for model training, evaluation and testing.

The company has since integrated the model into its AI-powered platform called DrAid, which detects 21 abnormalities and diseases of the heart, lungs and bone with an average accuracy of over 88% within 5 seconds. It can also auto-generate medical reports.

Based on WHO's latest report, there were an estimated 170,000 TB cases in Vietnam in 2019 or 176 people out of a 100,000 population contracting the disease.

THE LARGER TREND

Qure.ai reported multiple partnerships last month to help expand the screening of various diseases, such as tuberculosis and lung cancer, in developing Asian countries. The partnerships involve the integration of its AI software into medical systems to analyse radiographs and screen for abnormalities within seconds.

In April, Hong Kong-based AI company SenseTime got a CE mark to market its SenseCare-Chest DR Pro for detecting abnormal findings in chest x-rays. Its diagnostic software can spot and locate abnormalities for multiple diseases and generate precise quantitative analysis and text descriptions.

South Korean medical imaging firm VUNO has recently disclosed its study showing significant improvement in the diagnostic accuracy of its Med Chest X-ray solution with reading time cut by an average of 50%.

ON THE RECORD

"The white paper on utilising AI in TB diagnosis and screening has harnessed new tools and technologies to improve the current situation of the diagnosis and treatment of the disease, which has been considered to be 'incurable'. Technological solutions will help Vietnam to achieve its national goal of ending TB by 2030," VinBrain CEO Steven Truong said in a statement.

"This not only reflects Vingroup's determination to build a better society but also affirms its right strategy of investing in industry and high technology," he added.

Share