After reviewing more than 20,000 articles on the effectiveness of artificial intelligence in detecting disease, researchers have found that the methodology used in these studies in not sufficiently solid for them to draw reliable conclusions.
Is artificial intelligence truly able to make an effective medical diagnosis from images taken by a scanner, IRM or other device? The answer to that question remains uncertain according to a study published in The Lancet Digital Health in late September and which points to a lack of reliable studies on the subject. “Artificial intelligence (AI) seems able to detect disease from medical imagery just as accurately as human healthcare experts,” concluded the study’s authors, who also found, however, that the true potential of AI remains uncertain due to the lack of quality analysis available.
“We reviewed over 20,500 articles, but less than one percent of these were sufficiently robust in their design and reporting that independent reviewers had high confidence in their claims,” said Professor Alastair Denniston from University Hospitals Birmingham, who led the research.
The academic added: “What’s more, only 25 studies validated the AI models externally, using medical images from a different population, and just 14 studies actually compared the performance of AI and health professionals using the same test sample.”
When applied to disease diagnosis, AI uses a technique known as “deep learning”, which allows machines to perform the complex tasks they have been programmed to perform, such as voice or visual recognition.
“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,” commented Denniston.
No conclusive proof
“Perhaps the better conclusion is that in the narrow public body of work comparing AI to human physicians, AI is no worse than humans, but the data are sparse and it may be too soon to tell,” said Tessa Cook, an assistant professor of radiology at the University of Pennsylvania (USA), making an independent observation on the findings.
She added that comparing the two was risky and said that human physicians work in the real world, where data are “messy, elusive, and imperfect”.
Contact Allianz Partners
May 7, 2019
Inspired by the ability of dogs to sniff out certain illnesses in their initial stages, a number of teams of researchers are currently working on perfecting bioelectronic medical noses. [...]
Feb 14, 2019
One of the many potential applications of artificial intelligence in the field of medicine, AI could prove particularly useful in the early diagnosis of Alzheimer’s disease. Thanks to [...]
Feb 7, 2019
Researchers in Boston (USA) have developed DeepGestalt, artificial intelligence capable of diagnosing around 100 rare genetic diseases based on patients’ faces, working on the basis t [...]