AI can detect breast cancer almost as efficiently as humans: Report

Artificial intelligence (AI) is almost as good as humans at diagnosing metastatic breast cancer. Researchers from Beth Israel Deaconess Medical Centre and Harvard Medical School are developing a way of training AI using deep learning methods to read as well as interpret pathology images.

In fact, they proved how effective the technique is in a competition at the annual International Symposium of Biomedical Imaging. At the competition, the AI was given the task of looking for breast cancer in images of lymph nodes.

The AI was trained by feeding it multiple images of parts with cancerous cells and also others with normal cells. The AI initially had trouble identifying the images, but as with all deep learning techniques, it learned from its mistakes. Gradually, the AI improved to about 92 percent accuracy. It’s still a long way to go when compared to human accuracy, which is at about 96 percent, but that’s close enough to prove that the system can work.

The researchers add that while a single pathologist is better than the AI, AI working in combination with a pathologist is even more accurate than either alone.

“Here, we present a deep learning-based approach for the identification of cancer metastases from whole slide images of breast sentinel lymph nodes. Our approach uses millions of training patches to train a deep convolutional neural network to make patch-level predictions to discriminate tumor-patches from normal-patches. We then aggregate the patch-level predictions to create tumor probability heatmaps and perform post-processing over these heatmaps to make predictions for the slide-based classification task and the tumor-localization task,” the researchers write.

Researchers have published a paper on their research.