Stain transformation using Mueller matrix guided generative adversarial networks.

Area of Science:

• Digital Pathology
• Biomedical Optics
• Computational Imaging

Background:

• Traditional histopathology relies on chemical staining, which is time-consuming and resource-intensive.
• Virtual staining techniques offer a promising alternative to traditional methods.
• Mueller matrix microscopy provides label-free structural information beyond bright-field imaging.

Purpose of the Study:

• To develop a novel virtual staining method using Mueller matrix microscopy.
• To enable the transformation of Hematoxylin and Eosin (H&E) stained images into Masson trichrome (MT) stained images.
• To leverage polarization information for enhanced image generation in digital pathology.

Main Methods:

• Proposed Mueller matrix guided generative adversarial networks (MMG-GAN).
• Integrated polarization information from Mueller matrix microscopy into a generative adversarial network framework.
• Trained the MMG-GAN to translate H&E images to MT images.

Main Results:

• MMG-GAN successfully generated accurate MT-stained images from H&E inputs.
• Demonstrated the effectiveness of incorporating polarization information for stain transformation.
• Validated the potential of the proposed method for virtual staining applications.

Conclusions:

• Mueller matrix guided generative adversarial networks (MMG-GAN) offer an accurate virtual staining solution.
• Polarimetry-assisted digital pathology holds significant potential for future advancements.
• This technique can reduce the need for chemical staining, saving time and resources in histopathology.