Enhancing lateral flow assay analysis through the integration of artificial intelligence and image processing

Abstract

Cardiovascular diseases and thyroid disorders are major contributors to global mortality, necessitating early and accurate diagnosis for effective patient treatment. This thesis focuses on harnessing artificial intelligence (AI) to enhance the accuracy and efficiency of lateral f low assays (LFAs) for point-of-care (POC) diagnostics, aiming to empower healthcare professionals with better decision-making tools and ultimately improve patient care.

The research begins by highlighting the importance of accessible POC diagnostics and proposes the integration of AI with POC biosensors, specifically LFAs, to enable rapid and sensitive screening. Three distinct studies are conducted to investigate novel strategies for enhancing LFA sensitivity using AI. The first study employs a convolutional neural network to classify LFA images based on pixel intensity patterns of C-reactive proteins (CRP) biomarkers observed in the test lines and the control lines, achieving an impressive accuracy of 98.87%. In the second study, custom texture features of LFA images are utilised to detect low levels of thyroid-stimulating hormone (TSH), yielding an accuracy of 97.29% using support vector machines (SVM). The third study utilises hyperspectral imaging combined with machine learning to characterise and classify LFAs by thyroid-stimulating hormone levels, achieving an accuracy of 98.54% with another SVM model.

Overall, the integration of AI significantly improves LFA image classification accuracy, validating computational analysis for quantitative diagnostic interpretation of LFAs. Despite inherent limitations, this thesis underscores the potential of integrating biosensing with AI to facilitate affordable POC diagnostics. By offering a quantitative approach to LFA analysis, this research paves the way for more effective disease detection and management at the POC, ultimately improving healthcare accessibility and outcomes.

Thesis is embargoed until 31 August 2026

Date of Award	Aug 2024
Original language	English
Sponsors	Northern Ireland Connected Health Innovation Centre
Supervisor	Dewar Finlay (Supervisor), Mark Ng (Supervisor) & Jim McLaughlin (Supervisor)