Algorithm Of Deep Learning Is Established For Skin Cancer
Stanford researchers have trained algorithm hoped to get better medical care and diagnosis of skin cancer. Artificial intelligent diagnosis would help to access universal health care. The algorithm is fed up by 130,000 skin diseases images and could visually diagnose potential cancer. The device performed with inspiring accuracy in the initial tests.
An adjunct Professor from Stanford Artificial Intelligence Laboratory Sebastian Thrun said it is feasible and work well just like a human dermatologist. This would have long-term effects as to serve humanity. The product is being tested by 21 board-certified dermatologists in ability to detect skin lesions.
Algorithm for skin cancer?
5.4 million skin cancer cases estimated in US. The survival rate is 97% if detected in early stage which is further drop to 14% if detected in later stages. The diagnosis is usually a visual examination. A dermatologist examines the skin lesion with naked eye then with a low level magnification dermatoscope. These methods provide the inconclusive results to believe the lesion is cancerous, further biopsy is required.
Artificial intelligence or deep learning is just model as neural networks in the brain. The algorithm will provide the examination process follow with computing. The computing will relate the visual process with deep learning. Deep learning explored decade earlier in computer science but recently has applied to visual processing tasks.
The researchers successfully developed the powerful learning algorithm that will itself figure out rather just computer codes. The algorithm is feed with the images and associated disease label. However, it would take time for processing or sorting images. So the trained algorithm sort out from a huge data-set of skin cancer.
Health care by smartphone
The team is struggling to make the algorithm compatible to smartphone. It will surely provide a reliable diagnosis based on our fingertips. The ubiquity of smartphone would enable detection with number of sensors and camera in it. Scientists believed despite of challenges, deep learning would contribute in visual diagnosis in many medical fields someday. Techmasair