A deep learning-enabled imaging technology, developed by UCLA professor Aydogan Ozcan and colleagues, provides a noninvasive way to rapidly diagnose skin tumors, allowing earlier diagnosis of skin cancer. The technology bypasses reliance on skin biopsies, which are invasive, cumbersome, and time-consuming. It can take days to receive the results of a biopsy. The deep learning-based framework for the technology uses a convolutional neural network (CNN) to convert in vivo images of unstained skin, obtained using reflectance confocal microscopy (RCM), into virtually stained, 3D images with microscopic resolution. The CNN is trained to perform a virtual histology of the RCM images of the skin. Although RCM is a valuable diagnostic tool, its use requires specialized training. The CNN used by Ozcan and the other researchers changes the RCM images into images that look like the hematoxylin and eosin (H&E) images familiar to dermatologists. “I was surprised to see how easy it is for this virtual staining technology to transform the images into ones that I typically see of skin biopsies that are processed using traditional chemical fixation and tissue staining under a microscope,” said Philip Scumpia, assistant professor of dermatology and dermatopathology at the David Geffen School of Medicine at UCLA and the West Los Angeles Veterans Affairs Hospital. The researchers trained the network under an adversarial learning scheme, using RCM images of excised label-free tissue as inputs and using the microscopic images of the same tissue labeled with acetic acid nuclear contrast staining as ground truth. The researchers showed that the trained neural network was able to rapidly transform in vivo, label-free RCM images into virtually stained, 3D microscopic images of normal skin, basal cell carcinoma, and pigmented melanocytic nevi with H&E-like color contrast. “In our studies, the virtually stained images showed similar color contrast and spatial features found in traditionally stained microscopic images of biopsied tissue,” Ozcan said. “This approach may allow diagnosticians to see the overall histological features of intact skin without invasive skin biopsies or the time-consuming work of chemical processing and labeling of tissue.” The conversion of images obtained by noninvasive skin imaging to an H&E-like format could improve the clinician’s ability to diagnose pathological skin conditions. “The only tool currently used in clinics to help a dermatologist are dermatoscopes, which magnify skin and polarize light,” said Gennady Rubinstein, a dermatologist at the Dermatology & Laser Centre in Los Angeles. “At best, they can help a dermatologist pick up patterns.” A UCLA team achieved biopsy-free virtual histology of skin using deep learning and RCM. The conversion of images obtained by noninvasive skin imaging to an hematoxylin and eosin (H&E)-like format could improve the clinician’s ability to diagnose pathological skin conditions. Courtesy of Aydogan Ozcan/UCLA. To the best of its knowledge, the UCLA team is the first to apply virtual histology to intact, unbiopsied tissue. “This process bypasses several standard steps typically used for diagnosis — including skin biopsy, tissue fixation, processing, sectioning, and histochemical staining,” Ozcan said. “Images appear like biopsied, histochemically stained skin sections imaged on microscope slides.” The virtual “optical biopsy” with cellular resolution, in an easy-to-interpret visualization format, could enable more rapid diagnoses of malignant skin conditions and reduce invasive skin biopsies. Once the neural network is more fully trained using graphics processing units, it will be able to run on a computer or network, enabling a standard image to be rapidly transformed to a virtual histology image. The researchers’ aim to provide virtual histology technology that can be built into any device — large, small, or combined with other optical imaging systems. In future studies, the team will evaluate the usefulness of its approach across multiple types of skin neoplasms, and it will determine if this digital, biopsy-free approach can interface with whole-body imaging and electronic medical records. Additionally, the research team will determine if its AI platform can work with other AI technologies to search for patterns and further aid in clinical diagnoses. The research was published in Light: Science & Applications (www.doi.org/10.1038/s41377-021-00674-8).