Watson is learning to help clinicians save more lives.
Although melanoma is the deadliest skin cancer, it can be treated if detected early. Watson is learning by analysing structured and unstructured data, including millions of medical images and notes, to help clinicians identify melanoma in its earliest stages.
But even cognitive technology needs human input. And the more data Watson has, the more easy and accurate melanoma detection can become.
How Watson is learning to identify melanoma
Watson is learning to use 6 key data points to analyse images and determine the probability of melanoma.
Watson analyses a skin mole and searches for the presence of any of the six suspicious colours, which range from tan to black. The more colours are identified, the higher the risk of melanoma.
Watson divides skin moles into 8 regions, and assesses the irregularity of their borders and assigns them a score. A total score is then calculated for the entire mole. The higher the score, the higher the likelihood of melanoma.
Watson analyses the major and minor axes of a skin mole to determine whether it is fully symmetrical, half symmetrical or fully asymmetrical. The more asymetrical the mole, the higher the risk of melanoma.