Extended range and aberration-free autofocusing via remote focusing and sequence-dependent learning

Area of Science:

• Optical microscopy
• Machine learning in imaging

Background:

• Real-time autofocusing is crucial for tracking 3D sample variations and motion.
• Existing methods face challenges with long distances and aberration correction.

Purpose of the Study:

• To develop an autofocusing technique for fast, long-distance axial scanning with aberration correction.
• To integrate remote focusing with a bidirectional long short-term memory network for enhanced performance.

Main Methods:

• Utilized a deformable mirror and Shack-Hartmann wavefront sensor for remote focusing.
• Implemented a sequence-dependent learning approach using a bidirectional long short-term memory network.
• Tested the technique in a compact reflectance confocal microscope.

Main Results:

• Achieved a 120 µm autofocusing range in air and refractive-index-mismatched media.
• Demonstrated consistent performance with liquid layers up to 1 mm.
• Validated the technique on untrained samples and for tracking continuous axial motion.

Conclusions:

• The proposed technique enables real-time, aberration-free autofocusing over large axial ranges.
• Offers significant advantages for biomedical, holographic, and related imaging applications.
• Shows potential for robust sample tracking and dynamic imaging scenarios.