Label- and slide-free tissue histology using 3D epi-mode quantitative phase imaging and virtual hematoxylin and eosin...

Area of Science:

• Biomedical optics
• Digital pathology
• Artificial intelligence in medicine

Background:

• Hematoxylin and eosin (H&E) staining is the gold standard for tissue diagnosis but requires laborious formalin-fixation, paraffin-embedding (FFPE).
• The FFPE process is time-consuming, hindering applications like real-time surgical margin assessment.

Purpose of the Study:

• To develop a rapid, label- and slide-free method for generating H&E-like images from fresh tissues.
• To leverage quantitative oblique back illumination microscopy (qOBM) and AI for virtual histopathology staining.

Main Methods:

• Integration of 3D quantitative phase imaging (qOBM) with an unsupervised generative adversarial network (GAN) pipeline.
• Mapping of qOBM phase images from unaltered thick tissues to virtually stained H&E-like (vH&E) images.
• Validation using neural network classification and a user study with neuropathologists.

Main Results:

• High-fidelity conversion of qOBM images to vH&E with subcellular detail in mouse liver, rat gliosarcoma, and human glioma tissues.
• Demonstration of H&E-like contrast for volumetric imaging.
• Successful validation of vH&E image quality and fidelity by AI classifiers and expert pathologists.

Conclusions:

• Deep-learning-enabled qOBM provides a rapid, cost-effective alternative to traditional H&E staining.
• This approach facilitates new histopathology workflows, potentially improving cancer screening, detection, and treatment guidance.
• The technology offers real-time feedback capabilities for in vivo applications.