Machine learning can help predict whether a patient could develop diabetes in the future.

Researchers at the Kanazawa University Graduate School of Medical Sciences in Kanazawa, Japan, trained a computer model to flag patients likely to be afflicted by diabetes. The process, called machine learning, teaches an AI algorithm to recognize patterns from historical data. When running on live data, the algorithm can recognize the patterns it has learned. It can then issue an alert, for example, on the potential for contracting a disease. (Medical Xpress)

Diabetes, the silent killer

Diabetes is a chronic, incurable condition that can lead to serious health issues like coronary heart disease, cerebrovascular disease, retinopathy (disease of the eye), nephropathy (kidney disease), kidney failure and neuropathy (nerves disease).

Warning of the disease can help patients to change their lifestyles and perhaps prevent its occurrence. Unfortunately, no reliable methods exist to diagnose it earlier on.

However, a computer algorithm might be up to the task.

“Using machine learning, it could be possible to precisely identify high-risk groups of future diabetes patients better than using existing risk scores,” said lead author Akihiro Nomura, M.D., Ph.D. “In addition, the rate of visits to healthcare providers might be improved to prevent the future onset of diabetes.”

Machine learning

Researchers at Kanazawa University, led by Nomura trained such a computer algorithm on 509,153 nationwide annual health checkup records from 139,225 participants from 2008 to 2018 in the city of Kanazawa. The data also included 65,505 participants who did not suffer from diabetes.

The researchers trained the algorithm on patient data such as physical examinations, pathological tests, and participant questionnaires.

Accuracy of prediction

The researchers had identified a total of 4,696 new diabetes patients during the period under study. This constituted 7.2% of the total population.

When they ran the trained computer algorithm, it predicted future diabetes cases with an accuracy of 94.9%.

Future course

According to Nomura, he plans to put patients identified by the model as high risk for diabetes through clinical trials using statins as a treatment.

Related Story:     Artificial Intelligence: AI-based Study of Newly Infected Coronavirus Patients Predicts Severity of Disease