High-fidelity super-resolution microscopy datasets spanning multispectral to hyperspectral domains via diffractive...

Area of Science:

• Optical Microscopy
• Bioimaging
• Data Science

Background:

• Paired datasets are crucial for data-driven microscopy but are limited for spectral imaging.
• Existing synthetic datasets lack the fidelity of real biological structures.

Purpose of the Study:

• To present a comprehensive super-resolution dataset for spectral imaging.
• To provide a benchmark for developing image restoration and deep learning algorithms.

Main Methods:

• Diffractive optics-based structured illumination microscopy.
• Pattern-illuminated Fourier ptychography and analytical phase-shifting.
• Acquisition using a compact lattice SIM system.

Main Results:

• A dataset comprising multispectral (4-channel, 3-channel) and hyperspectral (503-689 nm) images.
• Paired diffraction-limited widefield (WF) and super-resolution (SR) images of cellular structures.
• Includes system point spread functions for accurate analysis.

Conclusions:

• The dataset enables rigorous benchmarking of image restoration, spectral unmixing, and cross-modality deep learning algorithms.
• Facilitates the extraction of nanoscale insights from standard optical microscopy setups.