With the help of artificial intelligence, doctors can potentially improve their accuracy in diagnosing cancer and other diseases, researchers from the Beth Israel Deaconess Medical Center (BIDMC) and Harvard Medical School (HMS) have shown.

The AI method is based on deep learning that uses algorithms for applications including speech recognition and image recognition, said Dr. Andrew Beck, Ph.D., director of bioinformatics at the Cancer Research Institute at BIDMC and an associate professor at Harvard, in a statement.

“This approach teaches machines to interpret the complex patterns and structures observed in real-life data by building multi-layer artificial neural networks, in a process which is thought to show similarities with the learning process that occurs in layers of neurons in the brain’s neocortex, the region where thinking occurs,” he said.

At the International Symposium of Biomedical Imaging (ISBI), the team’s artificial intelligence solution competed against private companies and academic research facilities around the world, where it examined images of lymph nodes to decide whether or not they contained breast cancer.

The team, including post-doctoral fellows Dayong Wang, Ph.D. and Humayun Irshad, Ph.D., and student Rishab Gargya, together with Aditya Khosla of the MIT Computer Science and Artificial Intelligence Laboratory, used the artificial intelligence system, whereupon it was able to diagnose the lymph nodes at a 92 percent accuracy.

A human pathologist has a success rate of 96 percent, and Beck said the truly exciting news is the 99.5 percent diagnosis accuracy of their artificial intelligence method when it was combined with a pathologist’s analysis.

“Peering into a microscope to sift through millions of normal cells to identify just a few malignant cells can prove extremely laborious using convention methods,” said Beck. “We thought this was a task that the computer could be quite good at – and that proved to be the case.”

To build the artificial intelligence computer, the researchers used hundreds of training slides where a pathologist had labeled regions that were cancerous and regions that were not, said Wang. They then extracted millions of the training examples and used deep learning to build a computational model that would classify them as cancerous or normal.

They identified where the computer kept making mistakes and retrained it by using more difficult examples. By doing so, according to the announcement, the computer’s performance continued to improve.

“Our results in the ISBI competition show that what the computer is doing is genuinely intelligent and that the combination of human and computer interpretations will result in more precise and more clinically valuable diagnoses to guide treatment decisions,” said Beck.

Jeroen van der Laak, Ph.D., who was an organizer of the competition and also leads a digital pathology research group at Radboud University Medical Center in the Netherlands, said in a statement that the fact that computers had almost comparable performance to humans was beyond what he had expected. He also said that it is a clear indication that artificial intelligence is going to shape the way they “deal” with histopathological images in the coming years.