Chromatic aberration correction employing reinforcement learning.

Area of Science:

• Optical microscopy
• Biomedical imaging
• Machine learning

Background:

• Fluorescence microscopy utilizes various labels exciting at different wavelengths, leading to chromatic aberrations.
• These aberrations shift focal positions, reducing spatial resolution in optical systems and biological samples.
• Sample-induced aberrations further complicate imaging quality.

Purpose of the Study:

• To present a novel method for correcting chromatic aberrations in fluorescence microscopy.
• To demonstrate the efficacy of an electrically tunable achromatic lens controlled by reinforcement learning.
• To improve spatial resolution and imaging quality in complex biological samples.

Main Methods:

• An electrically tunable achromatic lens with two chambers of optical oils and deformable glass membranes was developed.
• Reinforcement learning agents were trained to control the lens via four input voltages for non-linear system management.
• The system was tested for chromatic aberration correction and focal spot shift manipulation.

Main Results:

• Demonstrated chromatic aberration correction up to 2200 mm and focal spot position shifts of 4000 mm.
• Reinforcement learning agents successfully controlled the tunable lens to correct aberrations.
• Improved imaging quality was achieved, as shown with biomedical samples, including human thyroid tissue.

Conclusions:

• Reinforcement learning-driven tunable achromatic lenses offer effective correction of chromatic aberrations.
• This approach significantly enhances spatial resolution and imaging quality in fluorescence microscopy.
• The method is applicable to both systematic and sample-induced aberrations in biological imaging.