Noninvasive basal cell carcinoma detection algorithm with OCT

Basal Cell Carcinoma (BCC) is one of the most common non-melanoma skin cancers. Each year, over three million cases cost hundreds of millions of dollars in care. The standard for BCC diagnosis is biopsy which carries risks from anesthesia, bleeding, scarring, and infection. Imaging techniques such as dermoscopy, reflectance confocal microscopy and optical coherence tomography (OCT) provide opportunities for noninvasive disease evaluation, but are impeded by a steep learning curve. Although several algorithms exist to expedite clinician training, many studies that evaluate these algorithms utilize relatively homogenous sample groups. In this study, a novel OCT image classification algorithm was tested on a diverse dataset including 47 subjects and 19 anatomical sites, representing varying ages and FitzPatrick skin types. The net BCC classification accuracy was 85%, with significant variation across demographics and anatomical sites.