Lead study author Dr. Luke Oakden-Rayner, of the School of Public Health at the University of Adelaide in Australia, and colleagues believe that their findings could advance the field of precision medicine.
The National Institutes of Health (NIH) define precision medicine as “an emerging approach for disease treatment and prevention that takes into account individual variability in genes, environment, and lifestyle for each person.”
As the study authors note, precision medicine relies on the discovery of biomarkers that are accurate indicators of disease risk, response to treatment, or disease prognosis. They believe that radiology has an important role to play in this field.
“[…] we propose that images derived from routine radiological testing have been largely ignored in the context of precision medicine, and motivate the use of powerful new machine-learning techniques applied to radiological images as the basis for novel and useful biomarker discovery.”
“Recent advances in the field of medical image analysis have shown that machine-detectable image features can approximate the descriptive power of biopsy, microscopy, and even DNA analysis for a number of pathologies,” they add.
Patient mortality predicted with 69 percent accuracy
For their study, Dr. Oakden-Rayner and colleagues set out to investigate whether they could teach a computer to “learn” information in computed tomography (CT) scans, in order to predict a patient’s 5-year mortality.
Firstly, the team gathered more than 15,000 CT images of seven different tissues – including heart and lung tissue – from patients aged 60 and older. Using logistic regression techniques, the researchers identified a number of image features that were linked to 5-year mortality.
The team then combined the data with a “deep learning” technique. Dr. Oakden-Rayner explains that this is a method whereby computers can “learn how to understand and analyze images.”
“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,” he adds.
Next, the researchers used the computer to analyze CT chest images of 48 patients aged 60 and older. They found that it was able to predict their 5-year mortality with 69 percent accuracy, compared with mortality predictions made by healthcare professionals.
“Although for this study only a small sample of patients was used, our research suggests that the computer has learnt to recognize the complex imaging appearances of diseases, something that requires extensive training for human experts,” says Dr. Oakden-Rayner.
The next step for the team is to use the computer technique to analyze the CT images of tens of thousands of patients.
In the meantime, the researchers say that their study offers proof of concept that CT images and computer learning could lead to significant advances in precision medicine.
“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.”
Dr. Luke Oakden-Rayner