Automated analysis of dermoscopic imagery for recognition and detection of skin cancer

Conducted by Dr Prow T (UQ DRC) and PhD Candidate Sam Hames in collaboration with Professor H. Peter Soyer (UQ DRC)

This study focuses on automated detection of skin cancer in a variety of image types including clinical photography, dermoscopy and reflectance confocal microscopy.

In the first phase of this project, methods for numerically describing the appearance of lesions will be developed using computer based techniques such as the distribution of variation or patterns in brightness, the distribution of colours or the presence or absence of certain textures. The second phase of this project will turn the tools for describing and recognising lesions to detecting potentially malignant lesions.  The final stage of this project will investigate how the appearances of lesions change over time and how this is indicative of the development of skin cancer. Using the appearance of the lesion we will quantify how lesions change over time and examine patterns to determine how changing lesion appearance matches with the dermatologist’s assessment of a lesion.

Raphael, A. P., Kelf, T. A., Wurm, E. M. T., Zvyagin, A. V., Soyer, H. P., & Prow, T. W. (2013). Computational characterization of reflectance confocal microscopy features reveals potential for automated photoageing assessment. Experimental Dermatology, 22(7), 458-463. doi: 10.1111/exd.12176