Currently, the detection of metastatic cancer cells in tissues and lymph nodes are all done by the work of highly skilled pathologists by analyzing them under a microscope to look for irregularities within them. This can be a very time-consuming task for the doctors, and so researchers from Beth Israel Deaconess Medical Center (BIDMC) and Harvard Medical School collaborated to create an artificial intelligence form that could do this work them.
The AI that the team has developed uses a type of deep learning in order to replicate the work of a human. Their main aim in creating the AI was to teach it to be able to recognize patterns and structures within the cell. And, by achieving a success rate of 92 percent, the team are well on their way to perfection (or at least as good as a human, which is 97 percent). However, the team does feel that by combining the work of the AI and the human, we could achieve a success rate of 99.5 percent.
At the International Symposium on Biomedical Imaging (ISBI), the team won first place in two categories of the Camelyon Grand Challenge 2016 with their AI display. This just goes to show that other people are also recognizing the potential that this type of AI can bring. Not only can it free up time for pathologists to carry out more important tasks that are unsuitable for robots, but it also improves the detection rate if combined with human activity. Further research is due to continue to enable the system to detect other types of cancers too, and hopefully, we will soon see this being used in hospitals and clinics providing a more accurate and cost-effective diagnosis system.
Harvard Medical School’s technical report PDF – Deep Learning for Identifying Metastatic Breast Cancer
More News To Read
- Scientists Discover Brain Vessel Disease Links To Alzheimer’s
- New Eco-Friendly Airport System Could Soon Be Everywhere
- Researchers Used Graphene for Converting Electricity Into Light
- 3D Printing Goes to the Next Level With Map Printing
- The Researchers Created a New Enzyme That Can Produce Transcriptions Up To Ten Times More Accurate