AI app “improves accuracy of HIV tests”

[NAIROBI] Scientists have developed technology that uses artificial intelligence (AI) to interpret HIV test results, raising hopes of improving the quality of diagnoses, especially in low-income countries and intermediate, according to a study.

With more than 100 million HIV tests performed worldwide each year, better quality assurance could improve the lives of millions of people by reducing the risk of inaccurate tests, the study adds.

The AI-based mobile app deployed in a pilot field study in KwaZulu-Natal, South Africa accurately identified 97.8% of positive results and 100% of negative HIV test results.

“The overall performance with an accuracy of 98.9% is significantly better than the traditional visual interpretation of the study participants. “

Rachel McKendry, University College London

Trained community health workers accurately predicted 95.6% of positive tests and 89% of negative tests, showing the superior accuracy of AI technology. Workers used the mobile app to record their interpretation of 40 HIV test results, as well as to capture an image of the tests to be automatically read by the machine learning classifier.

Artificial intelligence, or deep learning, refers to the imitation of human intelligence in machines programmed to think like humans and mimic their actions.

“The results demonstrate the potential of deep learning for accurate classification, whether positive or negative, of rapid diagnostic tests. The overall performance with an accuracy of 98.9% is significantly better than the traditional visual interpretation of the study participants, ”says Rachel McKendry, study co-author and professor of biomedicine and nanotechnology at the London Center. for Nanotechnology, University College London in the United Kingdom.

McKendry tells SciDev.Net that the study began in 2017 with the aim of developing inexpensive, user-friendly and mobile-connected HIV diagnostic tools and to assess the feasibility of introducing these tools to improve access to HIV testing and after care, under resource-constrained conditions. The settings.

According to McKendry, other devices have agreed to send the results of the rapid HIV test to an online database in real time. Many small-scale approaches piloted on these have shown good performance, but most require a physical accessory such as a portable drive.

As part of the study, 60 field workers trained at the Africa Health Research Institute, based in South Africa, helped create a library of more than 11,000 images of HIV tests that were taken in various field conditions in KwaZulu-Natal. From this library, the AI ​​app was trained to classify tests as positive or negative, according to the study published in Natural medicine last month.

“We believe our real-world image library of 11,000 field-acquired images… is the first of its kind at this scale and our study demonstrates that deep learning models can be deployed with mobile devices on the web. field, without needing any other attachments, ”adds McKendry.

The app could be adapted to other locations, says Kobus Herbst, co-author of the study and director of the South African Population Research Infrastructure Network funded by the Department of Science and Technology.

“Although the study focuses on the interpretation of HIV tests, this tool could be adapted to interpret the results of rapid diagnostic tests for other infectious diseases,” said Herbst, specialist public health doctor. in health and research information systems. SciDev.Net.

According to Herbst, the app is not yet available for general use, but has been co-developed with end users in mind and to be inexpensive and affordable for healthcare facilities in low- and middle-income countries.

McKendry calls on African decision-makers to promote the use of digital technologies to solve health problems “Historically, public health decision-makers have been slower to adopt digital innovations than other sectors,” she explains .

Bernard Langat, director of the HIV, tuberculosis, malaria and noncommunicable diseases program at Amref Health Africa, says the app has great potential to minimize errors in reading HIV test results beyond health facilities.

Ensuring accurate results, he adds, is an important intervention in HIV programs. “There is a need to expand the library of images to include those from different contexts in sub-Saharan Africa and to work closely with HIV programs in these countries to further refine innovations in these settings,” says -he.

This article was produced by the English Sub-Saharan Africa office of SciDev.Net.