A CANDLE for a deeper in vivo insight

Area of Science:

• Biomedical imaging
• Microscopy
• Image processing

Background:

• Laser scanning multiphoton microscopy is crucial for deep tissue imaging but suffers from low signal-to-noise ratio (SNR) in scattering biological specimens.
• Existing denoising methods struggle with the noise inhomogeneity inherent in fluorescence microscopy images.

Purpose of the Study:

• To introduce a novel denoising algorithm, CANDLE (Collaborative Approach for eNhanced Denoising under Low-light Excitation), for 3D multiphoton microscopy images.
• To enhance the robustness of image processing under low SNR conditions.
• To enable deeper imaging capabilities in scattering biological tissues.

Main Methods:

• Developed CANDLE, an optimized non-local means filter that adaptively smooths image patches.
• Compared CANDLE against a state-of-the-art method and a locally adaptive non-local means filter.
• Validated the filter using synthetic data, microsphere images, and in vivo images of Xenopus tadpole brains.

Main Results:

• CANDLE demonstrated competitive denoising performance, particularly under low SNR conditions (PSNR < 8dB).
• The algorithm effectively handled noise inhomogeneity in laser scanning fluorescence microscopy images.
• Enabled successful imaging of neurons and fine axonal processes deep within scattering Xenopus tadpole brains.

Conclusions:

• CANDLE provides a robust solution for denoising multiphoton microscopy images acquired under low-light conditions.
• The method significantly improves image quality and enables deeper penetration into scattering biological tissues.
• CANDLE offers a valuable tool for advancing in vivo neuroscience research and other deep tissue imaging applications.