Skin cancer is the most common kind of tumour in people, with 100,000 new cases a year in the UK . Patients usually notice a strange looking mole, or mark on their skin, which is then examined by a doctor. If the doctor agrees that the mark looks suspicious, they may chose to look more closely using a dermoscope, and biopsies may then be taken to be examined in the lab. This process can take some time, and patients may not go to their doctor straight away when they notice something isn't quite right on their skin. Melanoma is by far the most dangerous skin cancer. Although they only make up for around 5% of skin cancers, they are responsible for around 75% of all of the skin cancer related deaths. If melanoma is treated in its earliest stages the 5 year survival rate is extremely good, around 99%. If it is left, and picked up in the late stages of disease this falls to just 14%.
A team of engineers and doctors from Stanford university have developed an innovative solution to the problem of early detection using the power of artificial intelligence. Researchers have trained machines capable of forming artificial neural networks to recognise skin cancers based on just a photograph. The AI was repurposed from software developed by Google, which was developed as part of an experiment to determine whether a machine could distinguish between cats and dogs.
The machine was shown 129,450 images of different skin lesions, and was told exactly what type of skin condition it was looking at in each one. It was then shown pictures of other potential cancers, and the diagnosis it gave was compared with the diagnosis from a panel of 21 expert dermatologists. Dr Andre Esteva told the BBC: "We find, in general, that we are on par with board-certified dermatologists." The system now needs to be tested in the clinic, alongside traditional methods.
Smart phones and other mobile devices have the capacity to run the same programs developed by the team from Stamford. This may allow people to proactively monitor any skin marks they are concerned about, and seek help from professionals if any marks are flagged up. With 6.3bn smart phone subscriptions projected by the year 2021, it is hoped that this technology may provide low-cost, universal access to vital diagnostic care. This could improve the early detection rate for skin cancers across the world, and save hundreds of thousands of lives annually.
Dr Esteva went on to say "The application of AI to healthcare is, we believe, an incredibly exciting area of research that can be leveraged to achieve a great deal of societal good."