As explained by Professor Haenssle, the AI resembles a tightly-knit net of artificial neurons that work together based on what they can 'see.' Furthermore, the deep learning convolutional neural network or CNN for short learns after each encounter and becomes even better at diagnosing skin cancer. Once trained it was tested against 58 dermatologists from across 17 countries, who were shown benign moles and malignant melanomas.
The CNN has the ability to learn fast by seeing images and teaching itself in order to improve its performance considering what it has learned (a process is known as machine learning).
For level two, the dermatologists improved by diagnosing 88.9 percent of malignant melanomas and 75.7 percent benign when given additional clinical information and close-up images, but again were outperformed by the CNN.
On average, the dermatologists correctly detected around 86.6 percent of melanomas while the CNN identified 95 percent of them. The American Cancer Society estimates 9,320 people will die from melanoma in 2018 while 91,270 new cases will be diagnosed.
It can be cured if detected early, but many cases are only diagnosed when the cancer is more advanced and harder to treat. This would suggest that human dermatologists tend to over-diagnose malignant melanomas, playing it safe rather than risk passing of a unsafe mole as benign. A Google algorithm devised to recognise unusual moles not only picked up more cancers but also ruled out more benign lesions.
'Only dermoscopic images were used - that is lesions that were imaged at a 10-fold magnification.
Health officials say the incidence of both non-melanoma and melanoma skin cancers has been increasing in recent years.
The researchers said they did not foresee artificial intelligence to replace dermatologists, but would help them make better diagnoses in the future. The CNN also had some limitations of its own, such as poor performance with images of melanomas on certain sites such as the fingers, toes, and scalp.
They conclude: "Currently, there is no substitute for a thorough clinical examination".
A type of machine learning called a convolutional neural network was trained...