Real time virtual histology video of squamous and basaloid neoplasms using a portable confocal microscope

Summary: Abstract Body: The time from biopsy to histological diagnosis can take days to weeks due to traditional histology preparation times. Reflectance Confocal Microscopy (RCM) can offer real time cellular-level resolution of skin lesions, but output images do not resemble traditional histologically stained slides as images are in grayscale. Herein, we applied our machine learning-based “virtual histology” algorithm to RCM images of the skin to determine if transformations can occur in near-real time for rapid assistance of diagnosis. We hypothesize that combining portable RCM with near instant videos of VH could aid in rapid, accurate skin diagnosis. The study was conducted as a prospective, consecutive trial at a VA Medical Center. In-vivo confocal RCM videos were taken with the portable RCM from the epidermis to dermis and run through the generative adversarial network-based algorithm to create real time H&E pseudo-stained VH videos. Whether VH combined with standard or portable RCM, we were able to use virtual histology to convert captured RCM images to virtually stained H&E-like images with a few second delay. The near instant VH videos directly correlated with pathology seen in conventional histology slides. The combination of portable RCM with real time VH video represents a promising clinical tool to consolidate visits, improve workflow, decrease time to management, and improve access to care. Rachel Wahhab^{1, 2}, Nelly Kokikian^{1, 2}, Yijie Zhang³, Yuzhu Li³, Stephanie Martin^{1, 2}, David Beynet^{1, 2}, Aydogan Ozcan³, Philip Scumpia^{1, 2} 1. Division of Dermatology, University of California Los Angeles, Los Angeles, CA, United States. 2. Dermatology, VA Greater Los Angeles Healthcare System, Los Angeles, CA, United States. 3. Electrical and Computer Engineering, University of California Los Angeles, Los Angeles, CA, United States. Translational Studies: Preclinical