The AI was built by researchers at the University of Adelaide, who published their findings in the journal Scientific Reports. They showed their program medical images from the chests of 48 different patients. With impressive accuracy, the AI was able to predict whether or not they would die soon.
The AI found the most success in patients suffering from chronic diseases such as emphysema and heart failure.
Like most modern AI programs, the technology is a deep learning program. Also called "limited AI," deep learning programs are able to draw conclusions based on large repositories of data previously fed into them. Impressively, and somewhat unsettlingly, the Adelaide team weren't actually sure what the program was looking at to justify its predictions.
"Although for this study only a small sample of patients was used, our research suggests that the computer has learned to recognize the complex imaging appearances of diseases, something that requires extensive training for human experts," said lead author Dr. Luke Oakden-Rayner in a press release.
"Instead of focusing on diagnosing diseases, the automated systems can predict medical outcomes in a way that doctors are not trained to do, by incorporating large volumes of data and detecting subtle patterns."
Oakden-Rayner also said that this AI could have significant medical applications. Examining the health of an individual organ can be time-consuming, never mind looking at all of them at once. If an AI can determine the health (or lack thereof) of an organ system quickly, it could help doctors create detailed treatment plans.
"Our research opens new avenues for the application of artificial intelligence technology in medical image analysis, and could offer new hope for the early detection of serious illness, requiring specific medical interventions."
He said that the next stage of the experiment will be to have the AI analyze the data of tens of thousands of patients.
A similar AI was developed by researchers at the London Institute of Medical Services, who published in the journal Radiology in January. That AI examined hearts exclusively, and the researchers claimed that it could predict whether or not a patient would die within the next year with 80 percent accuracy (compared to 60 percent accuracy from an average human doctor).