The day of computers assisting doctors in diagnosing diseases could be here soon.
Researchers from Beth Israel Deaconess Medical Center (BIDMC) and Harvard Medical School (HMS) collaborated in ‘training’ artificial intelligence to read and decipher pathology images. Using a ‘method based on deep learning’, the researchers used the same approach typically employed to condition AI to familiarize itself with speech, images, and objects.
The team got to show off just how effective their technique is at the annual International Symposium of Biomedical Imaging. In the competition, the AI was tasked to identify breast cancer in images of lymph nodes.
Training started with the machine being fed hundreds of slides, marked with indicators that distinguished cancerous cells and normal cells. They identified which slides the machine struggled with, then added eve more difficult samples. This method yielded an improvement to a 92% accuracy rate, winning two categories in the contest.
The AI is getting close to human efficiency; pathologists cite an accuracy score of 96%.
Science is the real winner here, as the AI and pathologists won’t work against each other. Beck says what’s truly exciting is the two forces combine yield a 99.5% accuracy rate.
“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.”
The team published a paper detailing their AI training experiences.