NIR-VIS multispectral fused imaging with extended depth-of-field for high-throughput virtual staining.

Area of Science:

• Histopathology
• Medical Imaging
• Computational Biology

Background:

• Accurate histological analysis is crucial for disease diagnosis.
• Virtual staining offers an alternative to traditional histochemical staining but often uses low-throughput methods.
• Limitations in current virtual staining hinder accurate tissue architecture reconstruction.

Purpose of the Study:

• To propose a high-throughput virtual staining framework for robust tissue characterization.
• To integrate visible (VIS) and near-infrared (NIR) spectral information for extended depth-of-field analysis.
• To enable accurate mapping from multispectral transmittances to stained RGB images.

Main Methods:

• Development of a multispectral microscopic imaging system for data acquisition.
• Establishment of HSSNet for efficient high-dimensional spectral feature extraction.
• Application of the framework to 10-µm-thick tissue sections, including liver cancer samples.

Main Results:

• The proposed method achieves high color fidelity in virtual staining.
• It reliably reconstructs differentiation-dependent tissue structural features in liver cancer.
• The framework demonstrates robust characterization of tissue architecture with extended depth-of-field.

Conclusions:

• The high-throughput virtual staining framework provides a promising approach for histopathological assessment.
• It effectively addresses limitations of low-throughput methods in virtual staining.
• This technology supports the analysis of heterogeneous tumors by reconstructing key structural features.