Measuring biological signals with high precision is crucial for early disease diagnosis and personalized medicine. Digital enzyme-linked immunosorbent assay (dELISA) in particular offers 1000x improved sensitivity over conventional immunoassays by counting single molecules partitioned into millions of compartments. This improvement in sensitivity has enabled the measurement of previously undetectable levels of protein biomarkers for early detection and treatment monitoring. In addition to targeting proteins, dELISA can target extracellular vesicles (EVs), nanoscale vesicles shed from cells that carry molecular cargo that can be traced back to their cell of origin. EVs have become an attractive biomarker for diagnostics because of their involvement in intracellular communication and disease progression. Because EVs are heterogeneous, they should be studied at the single EV level, making dELISA a suitable tool for their analysis. However, no single platform has been able to achieve the following requirements for EV analysis: 1) the sensitivity to detect individual EVs, 2) the specificity and multiplexing to detect multiple rare EV subtypes against the high concentration of background EVs in blood (10^10 EVs/mL), and 3) the throughput necessary to scan enough EVs to detect rare subtypes. This thesis addresses these technological challenges. In Chapter 2, we describe methods to achieve the necessary throughput (~20 million droplets/min) to perform ultrasensitive rare EV detection. We introduce droplet-based extracellular vesicle analysis (DEVA), a droplet dELISA platform that overcomes the throughput limitations of conventional microfluidics by parallelizing droplet generation and detection, and by interrogating the fluorescent signals using a time-domain encoded excitation approach previously developed in our lab. We detect CD81+ EVs with a limit of detection (LOD) of 9 EVs/µL, a >100x improvement over conventional EV characterization methods. We show that DEVA's sensitivity is related to the assay’s throughput to quantify enough false positive droplets to overcome Poisson noise. Chapter 3 demonstrates our ability to measure specific EV subpopulations from clinical samples. We specifically target PD-L1+ EVs, which have been shown to be a promising biomarker for melanoma diagnosis and immunotherapy response prediction. In a direct comparison with conventional ELISA, we show a 360× improvement in the LOD when targeting dual PD-L1+/CD81+ EVs. Moreover, we measure PD-L1+/CD81+ EVs in melanoma patient plasma (n=14) and match conventional ELISA measurements without the need for prior EV purification and with reduced plasma volumes. Chapter 4 expands on our time-domain modulation technology to perform high-throughput multiplexed measurements for digital assays. We perform quantitative fluorescence measurements on five populations of dual-encoded fluorescent beads with different ratios of blue and green dyes typically designed for flow cytometry. We combine time-domain modulation, high dynamic range imaging, and large field of view optics to measure droplets at a rate of 6x10^6/minute, and accurately classify five populations of dual-encoded beads (accuracy >99%). In addition, we detect bead-bound single enzyme molecules in a third fluorescence channel, as a proof-of concept for performing digital immunoassays. Taken together, this work addresses some of the key sensitivity, specificity, throughput, and multiplexing requirements for digital assays that can be applied towards a wide range of applications for the study of biology, disease pathology, and medicine.