Transforming Rapid Diagnostic Tests into Trusted Diagnostic Tools in LMIC using AI

Abstract

In low and middle-income countries (LMICs), Rapid Diagnostic Tests (RDTs) are often the only way to diagnose diseases such as malaria, HIV, and COVID efficiently and cost effectively, especially in rural settings. However, basic RDTs are often misinterpreted, reducing their reliability for medical treatment or official case counts. AI-based mobile solutions are difficult to implement in LMICs due to limited resources available on commonly used phones and unstable Internet connectivity. HealthPulse AI algorithms aim to address these issues by providing a lightweight, yet highly accurate library of Computer Vision (CV) models for the detection and interpretation of common RDTs for conditions such as malaria, HIV, and COVID. The complete system can function end-to-end offline on phones with as little as 1 GB of total device memory. In addition to detecting the RDT type and interpreting the results, the system can flag image quality issues such as bad lighting or blurriness. If required, it can ask the user for a photo retake in real-time, reducing the need for re-testing. The system provides accurate and consistent result interpretation for surveillance or decision support use cases, helping health systems better understand current disease prevalence which may help mitigate the next pandemic. The AI algorithm pipeline uses deep learning to analyze RDT images, with multiple computer vision models working together to confirm the presence of the expected RDT, flag adverse image conditions, and provide accurate and consistent results. HealthPulse AI prioritizes privacy, accountability, and accessibility while aiming to revolutionize care delivery in LMICs by transforming low-cost RDTs into trusted diagnostic tools using computer vision and AI.

Ähnliche Arbeiten