Rapid advancements in artificial intelligence (AI) and deep learning have allowed computers to learn and derive novel insights from vast datasets for transformative breakthroughs. In the field of medical diagnostics, the most widely available health monitoring tools are confined to measuring vital signs like temperature, blood pressure, and heart rate without the ability to conveniently detect complex biochemistry of the human body. This dissertation bridges the gap by developing simple, affordable, and highly sensitive biosensors tailored for precise, individual-specific health monitoring based on optical methods via label-free, highly multiplexed sensing with nanometric-scale resolution.
A colorimetric biosensing approach for at-home protein detection is developed using localized surface plasmon resonance (LSPR) on the aggregation of antigen-coated gold nanoparticles (GNPs) to detect SARS-CoV-2 Nucleocapsid (N) proteins. Experiments show this technique can produce results observable by the naked eye in 5 mins with a LOD (Limits of Detection) of 150 ng/ml for the N proteins. A comprehensive numerical model of the LSPR effect on the aggregation of GNPs has been developed to identify the key parameters in the reaction processes. An enhanced sensing mechanism that leverages the coffee-ring effect to amplify sample concentration is combined with deep neural network models. Using the evaporation of two sessile droplets and the formation of coffee-rings with asymmetric nanoplasmonic patterns, disease-relevant proteins as low as 3 pg/ml can be detected in less than 12 mins. The evaporation of the first droplet pre-concentrates proteins at the coffee ring, while the second plasmonic droplet produces a visible asymmetric plasmonic pattern by the naked-eye due to different aggregation mechanisms. The detection process is enhanced by incorporating a deep neural model that combines generative and convolutional networks for the quantitative diagnosis of biomarkers from smartphone-captured photos. Four different proteins, Procalcitonin (PCT) for sepsis, SARS-CoV-2 Nucleocapsid protein (N-Protein) for COVID-19, Carcinoembryonic antigen (CEA) and Prostate-specific antigen (PSA) for cancer diagnosis have been tested for a wide working concentration range over 5-order of magnitude.
A more sensitive platform by arranging plasmonic nanoparticles into 2D lattices to achieve exceptional points (EPs) is established to enable ultra-sensitive detections of nucleic acids. To enhance the robustness and stability of these EP biosensors, an AI-assisted phase-sensing methodology is incorporated with tunable parameters for post-fabrication calibration, achieving LOD of 1 fM in 30 min for nCoVE gene of SARS-CoV-2 diagnosis, about order of magnitude better than the state-of-the-art rapid photonic sensors. Finally, the integration of single-particle EP biosensing with photonic integrated circuit (PIC) interferometers is accomplished to achieve a compact, scalable design, enabling single-molecule detection. In conclusion, this research combines advanced plasmonic techniques with deep learning and phase sensing to address critical challenges in medical diagnostics. By bridging the gap between sensitivity, affordability, and accessibility, these innovations lay the groundwork for more personalized and precise health monitoring solutions.