AI Can Spot Disease As Accurately As Medical Professionals, Study Says

Prince William, Duke of Cambridge (R) looks through a dual console of the da Vinci XI robot before a highly complex robotic cancer operation to remove a tumour of the oesophagus at the junction between the heart, lungs and aorta, by lead surgeon Asif Chaudry and colleague surgeon Myles Smith with patient Charles Ludlow, 63, during his visit to The Royal Marsden Hospital on January 10, 2018 in London, England.
Daniel Leal-Olivas - WPA Pool / Getty Images

A new study in The Lancet Digital Health reveals that artificial intelligence (AI) might be able to detect diseases from medical imaging as accurately as humans, Newsweek reports.

According to the study, which reviewed 14 existing studies that compared AI and human performance between January 2012 and June 2019, deep learning algorithms were able to correctly detect disease in 87 percent of cases compared to 86 percent accuracy by humans. Additionally, AI was able to accurately exclude patients that didn’t have disease 93 percent of the time compared to healthcare experts, who were accurate 91 percent of the time.

While these results might sound scary, the study has its limitations. The researchers claim that there is currently a lack of quality analysis in studies that compare AI and human performance. Professor Alastair Denniston from University Hospitals Birmingham, one of the lead researchers of the study, says that less than 1 percent of the 20,5000 articles the research team reviewed had a “sufficiently robust” design that gave reviewers confidence in their claims. Ultimately, only 14 articles sufficiently compared AI and health professional performance with the same test sample, hence the small number of studies that were examined for the analysis.

“Within that handful of high-quality studies, we found that deep learning could indeed detect diseases ranging from cancers to eye diseases as accurately as health professionals. But it is important to note that AI did not substantially outperform human diagnosis.”

It’s also important to note that the 14 studies were conducted in isolation without factoring in additional clinical information that medical professionals consider when making a patient’s diagnosis. Regardless, the team believes that using AI to diagnose medical disease holds “enormous potential.”

“From this exploratory meta-analysis, we cautiously state that the accuracy of deep learning algorithms is equivalent to healthcare professionals, while acknowledging that more studies considering the integration of such algorithms in real-world settings are needed,” the study reads.

Tesla CEO Elon Musk — who wants to implant computers inside human brains — claims that AI is outperforming our ability to understand it. During a recent debate with Alibaba co-founder Jack Ma, Musk suggested that humans “underestimate the capability of AI.”

Loading...

“It’ll be much smarter than the smartest human,” he said, per Wired.

Musk claims that many AI researchers cannot imagine “something smarter than themselves,” adding that AI will almost definitely become “vastly smarter” than humans at some point.

The tech entrepreneur also pointed to the rate of advancement of computers, which he called “insane.” He pointed to video games 40 to 50 years ago and compared games like Pong to the current photo realistic games that act as real-time simulations.