DiLFM: an artifact-suppressed and noise-robust light-field microscopy through dictionary learning

Area of Science:

• Biomedical Imaging
• Microscopy Techniques
• Computational Imaging

Background:

• Light Field Microscopy (LFM) enables 3D biological dynamics recording at camera frame rates.
• Standard LFM reconstruction methods, like Richardson-Lucy (RL) deconvolution, introduce artifacts.
• LFM performance degrades in low-light conditions due to a lack of sample priors.

Purpose of the Study:

• To analyze LFM artifacts and develop an improved reconstruction technique.
• To enhance noise robustness and artifact suppression in LFM.
• To demonstrate the effectiveness of the new technique in biological samples.

Main Methods:

• Developed Dictionary Light Field Microscopy (DiLFM) using an over-complete dictionary.
• Analyzed various LFM reconstruction artifacts.
• Applied DiLFM to scattering biological samples and dynamic imaging scenarios.

Main Results:

• DiLFM substantially suppresses reconstruction artifacts compared to traditional methods.
• Improved noise robustness allows for reliable imaging in low-light conditions.
• Demonstrated artifact-suppressed reconstructions in Drosophila embryos and brains.
• Achieved robust blood cell counting at 100 Hz and revealed more neurons in zebrafish whole-brain recordings.

Conclusions:

• DiLFM offers superior artifact suppression and noise robustness for LFM.
• The technique significantly enhances the quality and utility of 3D biological dynamic imaging.
• DiLFM is effective for diverse applications, including cell counting and in vivo neural activity recording.