From ancient tools to artificial intelligence for TB detection and intelligence
Read the first seminal article on artificial intelligence (AI) software for TB by the Stop TB Partnership's TB REACH
23 October 2019, Geneva, Switzerland. Artificial intelligence (AI), a buzzword in the tech world, has steadily marked its presence in the field of healthcare, including in tuberculosis (TB). Recently, AI is being employed to read and screen chest x-rays for TB. These new solutions are providing hope to tackle TB, which kills more people worldwide than any single infectious disease. The Stop TB Partnership’s TB REACH is leading evaluations of multiple AI software packages to screen and triage people with TB using chest x-rays. The team published the first-ever scientific article, evaluating the accuracy of multiple AI software for detecting TB from chest x-rays now available in Nature Scientific Reports.
The AI software evaluated in the paper makes use of deep
learning neural networks, the cutting edge of AI research.
According to Dr. Jacob Creswell, Head of TB REACH and senior
author of the paper, “This collaboration with partners in
Nepal and Cameroon demonstrates how the TB REACH platform is at
the forefront of identifying emerging technological solutions
that will modernize TB care. TB REACH provides the mechanisms to
evaluate and translate new, innovative approaches for scale-up
and uptake. Although chest x-ray has a potential to identify
people with undetected TB, skilled radiologists are often scarce
in many high TB burden countries. This automated technology can
address current shortcomings.”
The paper examined three deep learning software systems with
stable version control:
CAD4TB (v6) developed by Delft Imaging Systems (Netherlands),
qXR (v2) by Qure ai
(India), and
Lunit INSIGHT (v4.7.2) developed by Lunit (South Korea) using
bacteriological evidence (Xpert®) as a reference standard.
The results are very promising. All three deep learning
software performed significantly better than human radiologists
in detecting TB abnormalities and can greatly reduce costs - by
more than half – of follow-up Xpert testing.
Dr. Lucica Ditiu, Executive Director of the Stop TB
Partnership, said, “In Stop TB this is what we
do the best: push the boundaries on TB, ambition, commitment
and persistence. Since 2010 when TB REACH was created, we
contributed hugely to the change in the landscape for TB
diagnosis, active case finding, finding
“missing” people with TB, and adherence
technology. We are now looking into x-ray, wearables,
and many other forms of AI platforms, with the
ultimate goal of raising the standard of TB care and
identifying all people with TB.”
The Stop TB Partnership is working with partners and
developers on several other studies and evaluations of different
AI products and will share them as they become available.
Source:
Stop TB Partnership
