A machine had more proficiently detected skin cancer than human dermatologists, a recent study – aimed at better and faster diagnosis – has found.
A team of researchers from Germany, France and the United States fed an artificial intelligence system – a deep learning convolutional neural network (CNN) – with more than 1,00,000 images of malignant melanomas and benign moles to differentiate skin lesions from the benign ones.
The machine was pitted against 58 experts from 17 countries around the world. The dermatologists were divided into three groups – 50 per cent with over five years of experience, 19 per cent with two to five years and 29 percent with less than two years’ experience.
According to research papers published in the journal Annals of Oncology, the CNN correctly detected 95 per cent skin cancer from the photographs, while the experts spotted 86 per cent.
The CNN’s detection of more melanomas cancer – that begins in the melanocytes of the skin – suggested the machine had higher sensitivity than the experts. Also, since the machine “misdiagnosed fewer benign moles as malignant melanoma” than humans, it indicates that deploying machines would achieve greater success. Benign moles is a kind of pigmented lesion on the skin.
This had opened a whole new vista to treat skin cancer with the help of artificial intelligence system in future. But the system may not be used in surgery or treatment entirely. Rather, it would be used to assist human doctors to ensure better and faster treatment.