A workflow for melanocytic lesion evaluation

In today’s current practices, detecting melanoma in its early stages is a challenging task for dermatologists. Early stage melanoma is nearly identical to benign melanocytic lesions, requiring doctors to detect subtle changes in shape or size of lesions that are easily missed, but are indicators of development of the disease. Dermatoscopy, a discipline that involves analyzing suspicious patterns in magnified pictures of skin lesions, helps dermatologists keep track of these changes. However, this approach has its shortcomings. Typically, the process of identifying melanoma is not consistent between doctors, or even within each doctor. This results in two problems: (1) it creates data that is not comparable between doctors, such as when one doctor captures photos while another does not, and (2) it produces data that is not consistent within a doctor’s own records of a patient from visit to visit, such as when they switch between capturing photos of a mole or simply taking notes. In the second case, doctors may have gaps in photos available, causing difficulties in tracking the progress of moles. Further, their existing methods are time intensive, putting pressure on them to complete their evaluation quickly, and require them to perform actions that could be done by someone without a medical degree, such as taking pictures of a person’s moles. In our work, we introduce MoleMapper, an Android-based application that addresses the issues above by enforcing a consistent workflow. The key insight in our work is that, by having a standardized, repeatable workflow, we can create tooled support that both introduces consistency, and also automates parts of the process that were previously tedious. MoleMapper does this by: (1) introducing a structured workflow which guides the user in taking full body pictures for the doctor to review, (2) making use of a subdivision of the body that thoroughly captures all potential areas that may contain moles, and (3) automatically creating a mapping between the pictures taken by a ...