Identifying skin cancer with computer vision

IBM researchers are fine-tuning a computer system that early research shows is more effective at identifying a form of skin cancer than expert dermatologists. The IBM system combines recent developments in deep learning with established machine learning approaches, creating ensembles of methods that are capable of segmenting skin lesions, as well as analysing the detected area and surrounding tissue for melanoma detection. The system was evaluated recently using the largest publicly available benchmark dataset of dermoscopic images, containing 900 training and 379 testing images.

Compared to the average of eight expert dermatologists on a subset of 100 test images, the proposed system produced a higher accuracy (76% vs 70.5%), and specificity (62% vs 59%) evaluated at an equivalent sensitivity (82%). “Our vision is that taking pictures to diagnose melanoma might one day be as routine as drawing blood to detect other diseases,” said Noel Codella, an IBM Research scientist and author of the research.

The paper, which is available online, will also be published in a 2017 issue of the IBM Journal of Research and Development.