Physics-informed multi-encoder adaptive optics enables rapid aberration correction for intravital microscopy of deep...

Deep tissue imaging often suffers from significant light scattering that obscures fine biological structures. Conventional methods for correcting these distortions frequently rely on guide-stars or require lengthy processing times. This limitation prevents real-time observation of rapid physiological processes in living specimens. No prior work had resolved the trade-off between correction speed and image quality in complex environments. Researchers have long sought ways to bypass the need for external reference points during optical correction. That uncertainty drove the development of new computational frameworks for adaptive optics. Prior research has shown that neural networks can assist in image restoration tasks. However, existing models often struggle with the noise levels inherent in deep biological imaging.

The authors aim to develop a method for rapid, guide-star-free aberration correction in deep tissue imaging. This study addresses the fundamental constraints imposed by tissue-induced optical distortions on intravital microscopy. The researchers seek to overcome the limitations of conventional adaptive optics, which often require slow processing or external reference points. They propose a multi-encoder network to handle large-amplitude aberration modes. This gap motivated the creation of a noise-resilient, structure-independent feature extraction model. The team intends to balance prediction accuracy with temporal efficiency for dynamic biological observation. They aim to validate this platform in living organisms to prove its versatility. This work addresses the need for robust imaging tools in low-signal and scattering environments.

The investigators designed a multi-encoder network to perform aberration correction. They employed a structure-independent feature extraction model to process biological image data. This approach relies on training the system with wavefront-modulated image pairs. The team validated their framework using living zebrafish and mouse models. They performed fluorescence imaging in the zebrafish brain and eye to test performance. The researchers also conducted experiments in the mouse visual cortex through thinned-skull windows. Their protocol emphasizes temporal efficiency during the decoding of large-amplitude distortion modes. This review approach highlights how the model handles low-signal and scattering environments.

The researchers report that their method enables rapid, guide-star-free aberration correction in complex biological specimens. They observed improved fluorescence imaging quality in both zebrafish and mouse models. The system successfully enhanced the detection of neuronal calcium transients in the visual cortex. Their findings show that subcellular-resolution microglial imaging is achievable through thinned-skull windows. The model effectively decodes multiple large-amplitude aberration modes from image pairs. This technique reveals spatiotemporally heterogeneous signaling patterns that were previously hidden by skull-induced distortions. The authors demonstrate robustness in low-signal and high-scattering conditions. Their results establish a balance between prediction accuracy and processing speed.

The authors propose that their multi-encoder architecture effectively balances prediction accuracy with temporal efficiency. This approach enables rapid aberration correction without relying on external guide-stars. The researchers demonstrate that their method improves fluorescence imaging quality in zebrafish models. They report enhanced visualization of neuronal calcium transients within the mouse visual cortex. The study suggests that subcellular-resolution microglial imaging becomes possible through thinned-skull windows using this platform. The team claims that their model reveals signaling patterns previously hidden by skull-induced distortions. They conclude that the system remains robust under low-signal and high-scattering conditions. This versatile framework supports dynamic subcellular observation deep within native tissue environments.