[ad_1]
The researchers conducted the first systematic review of existing research on AI in the health sector and published their findings in The Lancet Digital Health.
It's focused on an artificial intelligence technique called deep learning, which uses algorithms, big data, and computational power to mimic human intelligence.
This allows computers to identify disease patterns by examining thousands of images before applying what they have learned to new individual cases to establish a diagnosis. Technology is arousing a growing craze, and the US Food and Drug Administration has already approved a number of AI algorithms for use in health care.
Artificial intelligence has been hailed as a way of reducing the workload of overwhelmed health professionals and revolutionizing health care, but so far scientific research has not been up to the task.
Of the 20,500 items reviewed, less than 1% were found to be robust enough, said Professor Alastair Denniston of Birmingham University Hospitals, NHS Foundation Trust, UK, in a statement.
"Among these few high quality studies, we found that deep learning could indeed detect diseases ranging from cancer to eye diseases as accurately as health professionals," said Denniston.
"But it's important to note that AI has not far surpassed human diagnoses."
Using data from 14 studies, researchers found that deep learning algorithms correctly detected disease in 87% of cases, compared to 86% for health professionals.
AI was also able to correctly identify patients who were free of the disease in 93% of cases, compared to 91% for health professionals.
While these results are promising, researchers say more research and more in-depth reporting is needed to improve our knowledge of the true power of deep learning in health care settings.
This will involve better design of the study, including testing the AI in situations identical to those in which health professionals work.
"Evidence on how AI algorithms will alter patient outcomes should come from comparisons with other diagnostic tests in randomized controlled trials," said Livia Faes, of Moorfields Eye Hospital, London, in a statement.
"Until now, there is virtually no such testing where the diagnostic decisions made by an AI algorithm are exploited to see what will happen." then results that really matter to patients, such as fast treatment, the time needed to get out of the hospital or even survival rates. "
The experts welcomed the review while stressing the need for further research.
"In my opinion, the big caveat is that history does not say" AI can be as valuable as health professionals ", but the" general criterion for assessing the performance of the IA is mediocre, "said Franz Kiraly of University College London.
Nils Hammerla of Babylon Healthcare, a company that claims to use artificial intelligence technology to improve the affordability and accessibility of health care, believes that there is still much work to be done so that the AI can reach its full potential.
"Machine learning can have a huge impact on health problems, big and small, but unless clinicians and the public are convinced of their safety and abilities, it will not help nobody, "he said.
The global AI market in the health sector is booming and is expected to grow from $ 1.3 billion in 2019 to $ 10 billion in 2024, according to Morgan Stanley Investment Bank.
Hospitals around the world are already using this technology, including the Moorfields Eye Hospital in London.
Doctors can use an algorithm developed by DeepMind, a Google-based UK-based AI research center, to return a detailed diagnosis in about 30 seconds using coherence tomography scanners. optical (OCT).
The AI technology, called DeepGestalt, has outperformed clinicians by identifying a range of syndromes in three trials and could bring significant added value to personalized care.
[ad_2]
Source link