Stanford University develops an AI app that translates a photo into a skin cancer diagnosis

June 18, 2017, 10:42 a.m. By: Pranjal Kumar

Skin diagnosis

People living in remote areas finds it difficult to travel hundreds of miles to visit a good doctor. They all have to bear the cost of traveling in the process.

But now, researchers from Stanford University have created a deep learning algorithm which helps in skin cancer diagnosis. It takes the image of skin lesion and classifies as the begin or malignant. The accuracy achieved by this method is similar to that by a dermatologist.

The images can be taken from the simple camera of your smartphone. This makes this technology cost effective.

According to a survey, more than 5 million skin cancer are diagnosed every year in the US. The most of these cases are from the rural areas.

Andre Esteva, one of the researchers said,” We have been trying to develop a technology for early stage diagnostics for a very long time.”

The researchers have demonstrated the technology on a desktop app. The team trained their models on NVIDIA TITAN GPUs. Their deep learning algorithm is powered by CUDA and cuDNN.

The team has trained the machine from the dataset of about 130,00 skin disease images. These images span over than 2,000 disease categories. They have used open source medical repositories and local hospitals for these records.

The recent developments clearly suggest that the AI might power its way in the medical field. As the smartphones are becoming more powerful, these AI software’s could be run on these smartphones only.

More information: News Stanford Edu