Compressed fluorescence lifetime imaging via combined TV-based and deep priors

Area of Science:

• Optics and Photonics
• Biomedical Imaging
• Computational Imaging

Background:

• Compressed fluorescence lifetime imaging (Compressed-FLIM) enables single-shot widefield FLIM with high temporal resolution.
• Existing Compressed-FLIM reconstruction algorithms limit precision, especially for complex decay models.
• Need for improved reconstruction accuracy in large-scale Compressed-FLIM applications.

Purpose of the Study:

• To develop a more effective combined prior model for enhancing Compressed-FLIM reconstruction accuracy.
• To address limitations in current Compressed-FLIM precision, particularly for large-scale imaging problems.
• To validate the proposed method's performance against state-of-the-art techniques.

Main Methods:

• Development of a novel combined prior model, 3DTGp V_net, within the Plug and Play (PnP) framework.
• Extensive numerical simulations to evaluate reconstruction artifact reduction and accuracy improvement.
• Single-shot FLIM experiments using Rhodamine reagents for practical validation.

Main Results:

• The 3DTGp V_net model effectively eliminates reconstruction artifacts introduced by Deep denoiser networks.
• Reconstructed accuracy improved by approximately 4dB (peak signal-to-noise ratio; PSNR) compared to TV+FFDNet.
• Experimental validation demonstrated promising reconstruction performance with negligible lifetime bias.

Conclusions:

• The proposed 3DTGp V_net algorithm significantly enhances Compressed-FLIM reconstruction accuracy and reliability.
• This advancement offers improved precision for analyzing complex FLIM signals in biological and material science.
• The method shows practical utility for single-shot FLIM imaging, reducing bias and improving data quality.