Noninvasive virtual biopsy using micro-registered optical coherence tomography (OCT) in human subjects

Area of Science:

• Biomedical Imaging
• Computational Pathology
• Medical Diagnostics

Background:

• Histological hematoxylin and eosin (H&E) staining is the gold standard for cancer diagnosis but is invasive and time-consuming.
• Virtual staining of unstained slides shows promise, but true virtual biopsy requires imaging intact tissue.

Purpose of the Study:

• To develop a method for virtual H&E staining of intact tissue using noninvasive imaging.
• To enable high-fidelity conversion of optical coherence tomography (OCT) images to H&E-like images for pathology.

Main Methods:

• Developed a micron-accuracy coregistration technique to align 2D H&E slides with 3D OCT images of the same tissue.
• Trained a conditional generative adversarial network (cGAN) on paired H&E and OCT datasets.
• Applied the cGAN to convert noninvasive OCT images into virtually stained H&E slices in 2D and 3D.

Main Results:

• Achieved high-fidelity virtual H&E staining from noninvasive OCT images.
• Demonstrated successful 2D and 3D virtual staining of tissue sections.
• Validated the coregistration and virtual staining method on fresh tissue samples.

Conclusions:

• The developed method enables virtual biopsy by converting OCT images to H&E slices.
• This approach has the potential to reduce the number of invasive biopsy procedures.
• Facilitates the clinical integration of OCT imaging for disease diagnosis and margin assessment.