Multi-modal Image Fusion for Multispectral Super-resolution in Microscopy.

Area of Science:

• Biochemistry
• Microscopy
• Image Analysis

Background:

• Spectral imaging is vital in biochemistry but suffers from lower spatial resolution than structural microscopy.
• Partial voluming and low light exposure in spectral images hinder interpretation and analysis.
• Upsampling spectral images using high-resolution structural data is needed for fused representations with high spatial and spectral specificity.

Purpose of the Study:

• To develop a framework for fusing co-registered structural and spectral microscopy images.
• To create super-resolved spectral image representations by leveraging high-resolution spatial information.
• To validate the proposed framework for maintaining spectral integrity and super-resolution applicability.

Main Methods:

• Developed a framework for fusing co-registered structural and spectral microscopy images.
• Applied the framework to super-resolve spectral images of retinal tissue using structured illumination microscopy data.
• Applied the framework to super-resolve mass spectroscopic images of mouse brain tissue using histology data.

Main Results:

• Successfully generated super-resolved spectral images by integrating high-resolution spatial information.
• Validated the model's assumptions, ensuring the preservation of original spectral characteristics.
• Quantified spectral prediction accuracy using functional R² values and spatial quality using normalized mutual information.

Conclusions:

• The proposed framework effectively fuses structural and spectral microscopy images to achieve super-resolution.
• This approach enhances the interpretability and analytical capabilities of spectral imaging in biological tissues.
• The validation confirms the method's reliability for biochemical and biomedical applications.