Over the past century, our understanding of life has been focused on its most fundamental building blocks: molecules. Biological molecules, such as proteins found in blood, other body fluids, or tissues, are excellent guides to identifying a normal or abnormal process, condition, or disease. Nonetheless, their minuscule scale presents challenges in accurately quantifying protein molecules. Photonic biosensors provide a convenient method for probing analytes using light. Nanophotonics, the study of light's interaction with nanoscale structures, has emerged as a promising platform to tightly confine the light and overcome the scale barrier between minuscule biomarkers and electromagnetic fields. Metasurfaces have recently garnered considerable attention as an auspicious platform for biosensing applications. These two-dimensional engineered structures or artificial electromagnetic media are made of unit cells much smaller than their operating wavelength. High-index dielectric nanostructures are excellent candidates because of their ability to confine and scatter light strongly with relatively low absorption losses. Due to their unique optical characteristics, dielectric metasurfaces can realize sharp resonances that are highly sensitive to the minute changes of refractive index in the vicinity of the nanoresonators, making them favorable for label-free refractometric sensing and providing reliable quantitative information. This doctoral thesis aims to demonstrate state-of-the-art metasurface technology for imaging-based refractometric biosensing in both end-point and real-time modalities and investigate dielectric metasurface resonance features and design parameters for optimal functionalities. Novel applications of high-quality-factor (high-Q) dielectric metasurfaces supporting modes rooted in the physics of bound-states-in-the-continuum are exploited. The high-Q metasurface sensors coupled with an imaging-based optical setup and advanced data processing methods have enabled the construction of an end-point sensing platform with superior sensitivity. Our hyperspectral imaging setup provides spectral information from the sensing area with pixel resolution, which, combined with advanced data processing methods, allows for detecting highly diluted samples. Additionally, a real-time imaging-based biosensor employing an optofluidic chip comprising high-Q dielectric metasurfaces and microfluidics has been demonstrated. The method implements an aided imaging approach based on novel data processing to extract spectral shift information from time-resolved single-wavelength intensity images for highly reliable performance. This approach is suitable for multiplexed detection of biomarkers in real-time for high-throughput monitoring. As a proof-of-concept, we performed real-time in-flow experiments using the optofluidic device to detect extracellular vesicles secreted from breast-cancer tumors with clinically relevant results. The developed sensing devices rely on a wide range of engineering technologies. Innovative biosensing platforms have been presented by incorporating dielectric metasurfaces with imaging, microfluidics, micro- and nano-fabrication techniques, and novel data processing strategies. These methods hold great promise to overcome the challenges of biomedical diagnostics and pave the way for futuristic point-of-care devices enabling early diagnosis, treatment monitoring, personalized medicine, and democratized healthcare systems.