Abstract Optical microscopy’s power to resolve nanoscale structures deep within living tissues is limited by light diffraction and sample-induced aberrations. While super-resolution microscopy (SRM) has broken through the diffraction limit, its performance rapidly deteriorates in thick, heterogeneous specimens. Adaptive optics (AO) has emerged as a suitable solution to this challenge. By measuring and correcting wavefront distortions, AO restores imaging conditions, enabling SRM to achieve its full potential deep within complex biological samples like living brain tissue and developing organoids. The integration of AO with SRM (AO-SRM) is unlocking volumetric imaging capabilities, enhancing imaging depth and restoring nanoscale resolution. A recent advance has been the fusion of these hardware corrections with deep learning (DL). This convergence encompasses a dual paradigm of high-fidelity algorithmic restoration to bridge signal-to-noise ratio (SNR) gaps and systemic autonomy through self-driving frameworks for real-time hardware loop optimization. By employing intelligent agents that autonomously navigate aberration landscapes and trigger acquisition during biological events, this synergy maximizes experimental throughput while optimizing the photon budget to prevent sample degradation. DL algorithms enhance imaging capabilities by providing high-speed aberration prediction and robust computational reconstruction. This synergy not only boosts image contrast and resolution but also improves data processing efficiency and mitigates phototoxicity, enabling volumetric imaging with nanoscale precision. This review synthesizes recent developments in AO-integrated super-resolution imaging - encompassing image restoration approaches such as denoising and deblurring as well as wavefront correction techniques - and addresses how this convergence empowers the study of neuronal morphology and subcellular-scale dynamics deep within biological tissues.
Park et al. (Tue,) studied this question.