"Early HIV Detection with a Single Drop of Saliva"... Korea University Develops AI-Based Ultra-Sensitive Diagnostic Technology [Reading Science]

Early-Stage Infection Detection Up to 8 Days Faster Than Conventional Rapid Kits
Nanotechnology and Smartphone AI Combined
"Expected to Benefit Medically Underserved Regions Such as Africa"

A domestic research team has developed a technology capable of detecting early-stage human immunodeficiency virus (HIV) infection—often missed by conventional rapid diagnostic kits—using just a single drop of saliva. By combining nanotechnology and artificial intelligence (AI), the team succeeded in detecting early infection signals in oral fluid (saliva), where antibody concentrations are significantly lower than in blood.

Development of a BE-Nanotrap-Based Preprocessing Module for HIV-1 Antibody Specific Detection in Saliva and a Smartphone App AI-Based Lateral Flow Assay Integration Platform (BE-SMART-HIV), Improving HIV Diagnostic Accuracy through Transfer Learning of Large-Scale COVID-19 Lateral Flow Assay Training Data. Provided by the Research Team

Korea University announced on May 7 that the research team led by Professor Junghoon Lee of the KU-KIST Graduate School of Convergence has developed a saliva-based, ultra-sensitive HIV diagnostic platform named "BE-SMART-HIV."

HIV is most transmissible during the early stages of infection; however, during this period, antibodies are not sufficiently formed in the body, making detection difficult with conventional blood-based rapid diagnostic kits. In particular, saliva-based kits face limitations in field diagnosis because antibody concentrations are approximately 1,000 times lower than in blood and are subject to greater influence from proteolytic enzymes.

To address these challenges, the research team combined a bioengineering-based concentration (BE) nanotrap with an AI-driven interpretation system (SMART). The nanotrap selectively captures and concentrates high-molecular-weight substances such as HIV antibodies and filters out unnecessary low-molecular-weight components.

When saliva is dropped onto the diagnostic kit developed by the research team, the HIV antibody signal is amplified by about 20 times, enabling detection even at the earliest stages of infection. A notable feature is that the portable device can be used on-site without any external power supply.

In addition, the technology incorporates AI-based interpretation via smartphone. Like COVID-19 diagnostic kits, the HIV diagnostic kit utilizes a visible red test line to indicate results, but in early infection stages, this line may be faint and difficult to distinguish with the naked eye.

The research team employed transfer learning techniques using existing COVID-19 diagnostic image data, allowing the AI to analyze subtle differences in color intensity. Users simply need to take a photo of the diagnostic kit with a smartphone and transmit it for analysis.

As a result, the system achieved a diagnostic accuracy of 98.6% even with a limited amount of HIV data. This far exceeds the accuracy rates of layperson interpretation (67.1%) and clinical expert interpretation (78.3%).

Notably, the platform detected the emergence of antibodies up to eight days earlier than conventional rapid diagnostic kits and successfully identified early immune response patterns previously observable only through advanced blood tests.

Photo of the research team. (From left) Jeonghun Lee, Professor at Korea University KU-KIST Graduate School of Convergence Science and Technology (corresponding author), Jungsu Park, Ph.D. candidate in the Department of Mechanical Engineering at Korea University (first author), Seungmin Lee, Postdoctoral Researcher at Korea University KU-KIST Graduate School of Convergence Science and Technology (first author). Courtesy of Korea University

The research team emphasized that, beyond simply increasing diagnostic sensitivity, this technology demonstrates that oral fluid—previously considered unsuitable for precision diagnostics—can be utilized as a highly accurate diagnostic sample.

They added that, because the concentration module and detection module are structured independently, the platform has great potential for expansion to diagnose a variety of infectious diseases, such as influenza and COVID-19.

Professor Junghoon Lee remarked, "Highly sensitive HIV analysis is possible with only commercial diagnostic kits and a smartphone. This technology could help address early diagnosis and large-scale infection screening challenges in medically underserved regions, such as Africa."

This research was published in ACS Nano, an international journal in the field of nanobio and AI diagnostics, on April 16.

This content was produced with the assistance of AI translation services.