Abstract: Single molecule localization microscopy (SMLM) has become an essential tool in imaging nanoscale biological structures. It breaks the diffraction limit by utilizing photo-switchable or photo-convertible fluorophores to obtain isolated single molecule emission patterns (i.e. PSFs) and subsequently localize the molecule’s position with a precision down to ~ 20 to 80 nm laterally-axially. However, optical aberrations compromise its spatial resolution. Additionally, conventional SMLM algorithms require sparse activation to reduce emission pattern overlap, which restricts imaging speed and temporal resolution, thus limiting its utility in dynamic live cell imaging. In this study, we first conducted a comprehensive quantitative analysis of the theoretical precision limits for position and wavefront distortion measurements in the presence of aberrations, which enhances our understanding of aberration effects in SMLM and lays the groundwork for developing more effective aberration correction methods. To improve temporal resolution, we developed a high-density single molecule localization algorithm that utilizes deep learning to analyze molecule blinking sequences. This approach allows us to achieve high localization precision and resolve structures at tens of nanometers resolution, even with highly overlapped blinking data. Validated by both simulated and high-density experimental data, our algorithm successfully resolves the complex structures of various cellular organelles and captures rapid dynamic movements in live cells. This work addresses the knowledge gap about aberrations in SMLM and expands its applications to more dynamic and detailed studies of cellular processes.