Breast cancer is one of the most common forms of cancer in women. In 2018 it’s estimated that there will be over 200-thousand new diagnoses in the U.S. alone. In this episode of AI News, we discover how existing deep learning technologies can be utilized to train AI to detect breast cancer in unlabeled histology images.
Subscribe to the Intel Software YouTube Channel
Hi, I'm David Shaw, and welcome to AI News. In this episode, we discover how existing deep learning technologies can be utilized to train AI to detect breast cancer in unlabelled histology images.
Breast cancer is one of the most common forms of cancer in women. In 2018, it's estimated that there will be over 200,000 new diagnoses in the US alone. So how can AI help? It can reduce the need for medical staff to manually examine mammography slides, saving time and money, and even lives.
Intel technology is being used to create a deep learning neural network that is able to detect breast cancer. It's called the IDC classifier. IDC is short for invasive ductal carcinoma, the medical term for breast cancer.
To create this classifier, Intel software innovator, Adam Milton Barker, used an Intel AI DevCloud to train the neural network, an Intel Movidius product for carrying out inference on the edge, an Up Squared device to serve the trained model, making it accessible via an API, and an IoT connected alarm system built using a Raspberry Pi device.
This demonstrates the potential of using AI, and the device via the IoT jump-way, to create intelligent automated medical systems. The article "Machine Learning and Mammography" provides a technical walk through of creating your own computer vision program for classifying positive and negative breast cancer cells. It covers everything from prepping your training data to the set up of the server to test a positive and negative prediction.
Check the links provided for the article. Don't forget to like this video and subscribe. And we'll see you next week for more AI News.