Machine learning assisted wavefront sensor.

Area of Science:

• Optical Engineering
• Machine Learning Applications
• Biomedical Imaging

Background:

• Adaptive optics (AO) systems correct optical aberrations to improve imaging resolution.
• Traditional AO systems rely on dedicated wavefront sensors to measure aberrations.
• Accurate wavefront measurement is crucial for effective aberration compensation.

Purpose of the Study:

• To investigate the feasibility of estimating wavefront error using machine learning directly from guide star images.
• To develop an AO system that bypasses the need for conventional wavefront sensors.
• To demonstrate aberration compensation using a machine learning-based wavefront sensor.

Main Methods:

• A two-photon laser spot was generated in fluorescein solution.
• Known aberrations were introduced using a deformable mirror to create a training dataset.
• A machine learning model was trained to predict wavefront error from guide star images.
• The trained model was integrated into an AO feedback loop with a deformable mirror.

Main Results:

• The machine learning model successfully estimated wavefront errors from simulated guide star images.
• The integrated AO system demonstrated proof-of-principle compensation of sample-introduced optical aberrations.
• This novel approach shows potential for simplifying AO system design.

Conclusions:

• Machine learning can effectively estimate wavefront errors, offering an alternative to traditional wavefront sensors.
• This method facilitates the development of more compact and integrated AO systems.
• The study validates the potential of AI-driven wavefront sensing for aberration correction in imaging.