Automated ELISA and Machine Learning-Enhanced Precision in Biosensing and Colorimetric Measurements

PASSCODE: 04242025

Access to reliable, affordable, and rapid diagnostics remains a critical challenge in many healthcare settings. This thesis addresses that need by introducing a miniaturized, low-cost, and automated platform for enzyme-linked immunosorbent assay (ELISA), designed for point-of-care use without dependence on centralized laboratories. Interleukin-6 (IL-6) detection was used as a model system to validate the platform’s performance. In addition, several machine learning–based studies are presented to enhance optical biosensing precision. A multi-resonance optical structure is examined in which multiple resonance peak shifts are combined using Ridge Regression, achieving up to three orders of magnitude improvement in refractive index detection compared to traditional single-peak methods. In another study, full-spectrum modeling further reduces mean squared error (MSE). A plasmonic structure based on titanium is also analyzed, where intensity-based detection achieves over 300-fold MSE reduction using machine learning models. An experimental chapter focuses on colorimetric biosensing, where spectrometer-acquired data—along with preliminary chemiluminescence image analysis—is modeled with ML to improve precision. Overall, this thesis contributes both a standalone, practical diagnostic platform and a set of data-driven methodologies that expand the accuracy, flexibility, and reach of modern biosensing technologies.

CHAIR: Professor Xudong Fan

CO-CHAIR: Professor L. Jay Guo