Interdisciplinary Initiatives Program Round 11 - 2022

Project Investigators:

Adam de la Zerda, Structural Biology
Kavita Sarin, Dermatology
 

Additional Collaborators:

  • Yonatan Winetraub, Structural Biology
  • Sumaira Aasi, Dermatologic Surgery
  • Kerri Rieger, Dermatopathology

Abstract:

There are over 5 million skin cancers annually in the United States with basal cell cancer (BCC) being the most common. The gold standard for diagnosis of skin cancer involves biopsy of the lesion and histologic examination under a microscope. However, biopsies are not an optimal approach for individuals with high numbers of skin cancers as they may suffer from excessive procedures and scarring. In addition, some skin cancers such as small BCCs may be able to be monitored over time, preventing unnecessary excisions and resulting surgical morbidity. Optical Coherence Tomography (OCT) is a promising non-invasive diagnostic technology which can image the skin from the surface. OCT has the potential to enable the diagnosis and detection of skin cancer in real-time and without requiring invasive biopsies. However, the translation of OCT into clinical use for skin cancer detection has been limited by difficulties in interpretation of images.

We recently demonstrated that we can generate virtual microscopic images of normal skin from OCT images using a novel alignment method and a trained neural network we call OCT2Hist. This has the potential to enable a clinician to see a “virtual biopsy” by putting an OCT scanner on the skin without actually cutting the skin. In this proposal, we plan to extend OCT2Hist for BCC detection as our first proof-of-principle study for clinical utility. In Aim 1, we will train OCT2Hist with an additional 3000 aligned OCT-H&E image pairs derived from actual samples as well as augmented sections of 50 BCC and 100 clinically similar samples. We will optimize OCT2Hist for BCC detection. In Aim 2, we will test the accuracy of OCT2Hist in an independent dataset of 30 BCC and 70 clinically similar lesions to confirm the validity of our approach. Our ultimate goal is to develop a real-time non-invasive “virtual skin biopsy” for the detection of BCC. If successful, our proposal will pave the way for real-time non-invasive longitudinal monitoring, detection, and margin control for BCC and support additional development for clinical translation and expansion to other skin lesions.