Lens-Free Holographic Microscopy with Deep Learning Image Classification for Biosensing and Disease Diagnosis

Abstract

Novel biosensing approaches are essential to the improvement and advancement of health and medicine. Many diseases have unique biomarkers that can be used to diagnose, guide treatment, and monitor treatment progress for the disease. In addition, biosensing for in- fectious diseases is essential for mitigating the impact of pandemics like the COVID-19 pan- demic. To this end, lens-free holographic microscopy (LFHM) has emerged in recent years as a novel biosensing platform for a variety of targets and diseases. In this dissertation, a novel LFHM-based biosensor was developed to diagnose COVID-19 in response to the pandemic. Additionally, deep learning algorithms were explored to improve the sensor’s performance in real-world and point-of-care conditions. A variety of these deep learning approaches were investigated and characterized, providing an in-depth resource for construction, fine-tuning, and debugging of similar deep learning algorithms. Finally, a second novel LFHM device was constructed that utilizes polarization and localized surface plasmon resonance of gold nanorods to detect single rods. This device was characterized using a resolution test target as a first step towards a biosensing study utilizing this novel device. This work provides a strong foundation for the use of this LFHM-based biosensor in the field for clinical diagnostic procedures, and represents an advancement in LFHM imaging.