[ad_1]
Experts praised the results of an examination according to which artificial intelligence (AI) could be as effective as medical professionals to diagnose diseases from medical imaging.
However, it was agreed that a lack of high quality research meant that the true value of AI remained uncertain.
The authors of the study, published in The Lancet Digital Health , called for more stringent research standards on the specific challenges posed by "deep learning" in order to improve future assessments.
Some experts have stated that a combination of AI with judgments made by medical professionals could be the preferred future of medical diagnosis.
The first study revealed a lack of good quality data
The systematic review and meta-analysis – the first to be undertaken – have revealed only a few studies of sufficient quality since 2012.
Professor Alastair Denniston of Birmingham's University Hospitals, responsible for research, said in a press release: "We reviewed over 20,500 items, but less than 1% of them were robust enough in their design and reported that independent reviewers had a great deal of confidence.Moreover, only 25 studies validated AI models externally and only 14 compared the performance of AI and health professionals. health using the same test sample. "
However, despite the reservations raised by high quality data, he said: "Among these few high quality studies, we found that in-depth learning could indeed detect diseases ranging from cancer to eye diseases with as much precision as health professionals did not surpass the human diagnosis. "
This close-up analysis was confirmed by an analysis of data from 14 studies that showed that AI could detect disease in 87% of patients, compared to 86% of health professionals.
With regard to the exclusion of the non-affected, the AI reached 93% against 91% for the professionals of the health.
Among the main drawbacks identified by the authors of the study, there was the fact that diagnoses of AI were often done in isolation, so as not to reflect clinical practice. For example, only four studies provided health professionals with additional clinical information that they would normally have used to make a diagnosis in clinical practice.
Co-author Dr. Livia Faes of the Moorfields Eye Hospital in London said: "Comparisons with alternative diagnostic tests in randomized controlled trials should demonstrate how the diagnostic algorithms can be used. IA will modify the results of patients.I use an AI algorithm to see what happens to results that really matter to patients, such as fast treatment, time out of the hospital or even the survival rate. "
The experts' point of view
A number of experts shared their views on the study with the Science Media Center.
David Curtis, an honorary professor at the Genetics Institute of University College London, commented, "I think the most striking aspect is that over 20,000 applications studies using the same technology," he says. IA for medical imaging published in scientific journals, only 14 were good enough to use – one in a thousand.
"Almost all studies published on AI for medical imaging did not use the proper methods and could be safely ignored.
"Among the very few truly valid studies, the results show that AI can interpret imaging as well as health professionals, but in many cases, professionals have been denied access information that would have been available to them in a real clinical scenario. "
Richard Mitchell, professor of cybernetics at the University of Reading, commented: "There are some examples where a combination of human and artificial intelligence gives an even better result, and that is maybe the better way to go. "
David Spiegelhalter, president of the Winton Center for Risk Communication and Evidence at the University of Cambridge, said: "In-depth learning can be a powerful and impressive technique, but clinicians and commissioners should to ask the crucial question: what does it really add to clinical practice? "
Dr. Nils Hammerla, Director of Machine Learning and Natural Language Processing in Babylon, said: "Machine learning can have a huge impact on health problems, whether they're big or small. small, but unless you convince clinicians and the public of their safety and abilities will be of no use to anyone. "
Dr. Nils Hammerla: He is a shareholder in Babylon and is a shareholder. It uses artificial intelligence and machine learning to provide health care tools to patients and clinicians.
The Lancet Digital Health, a systematic analysis and meta-analysis comparing the performance of in-depth learning compared to health professionals in the detection of diseases with the aid of medical imaging. Paper .
[ad_2]
Source link