Mole Gazer: Proof-of-concept study to improve early detection of melanoma using time-series analyses of evolution of naevi

ended

Early detection of melanoma improves survival. Individuals with multiple naevi (moles) are at an increased risk of developing melanoma, but sequential monitoring by dermatologists is time-consuming and inefficient. Artificial intelligence (AI) methods can potentially diagnose melanoma from single time-point lesion images, but a more clinically-relevant question is whether melanoma can be detected early, based on automated detection of changes in naevi using total body photography (TBP; high-resolution standardised images of body-parts). Cutting edge astronomical surveys use sequential images to monitor the night sky. With state-of-the-art AI techniques, these surveys identify and track subtle changes, robustly classifying the nature and prognosis of each event from just three images. Both astronomy and dermatology therefore face similar challenges: robustly predicting the evolution of sources from sparsely-sampled images. With this motivation we propose an innovative solution: adapting AI algorithms, developed in astronomy and the ERC SPCND project, for use in the automated detection, characterisation and monitoring of skin lesions. With a wealth of experience tackling this problem in astronomy, this proof-of-concept project will characterise the properties and evolutionary path of naevi in preparation for the next stage: the early detection of melanoma.