Graph-based modeling of optical system enables adaptive optics on dynamic samples with self-calibration

Area of Science:

• Optical imaging
• Computational microscopy
• Biomedical optics

Background:

• Sensorless adaptive optics (SAO) offer advantages over hardware-based wavefront sensing.
• Existing SAO methods face limitations with system imperfections, spatially invariant aberrations, and dynamic samples.

Purpose of the Study:

• To introduce a novel computational adaptive optics framework, GRAPHYCS, addressing key limitations of current SAO techniques.
• To enable accurate wavefront sensing and correction in challenging imaging scenarios, including dynamic biological samples.

Main Methods:

• Development of GRAPHYCS, a differentiable graph-based modeling framework incorporating self-calibration.
• Implementation of spatially variant wavefront sensing by modeling local aberrations.
• Application to dynamic live-sample imaging and zebrafish brain imaging.

Main Results:

• GRAPHYCS achieves up to a 9-fold improvement in wavefront sensing accuracy in simulations compared to analytic phase diversity.
• Experimental results demonstrate consistent outperformance of GRAPHYCS over phase-diversity-based methods.
• Successful simultaneous wavefront sensing and neuronal activity detection in live zebrafish brain imaging.

Conclusions:

• GRAPHYCS overcomes limitations of traditional SAO, offering self-calibration for system non-idealities and enabling spatially variant sensing.
• The framework supports dynamic live-sample imaging, expanding the capabilities of computational adaptive optics.
• GRAPHYCS facilitates advanced applications like simultaneous neuronal activity detection and wavefront sensing in live biological systems.