Data-driven diagnoses

Doctors typically use medical guidelines to assess a patient’s risk of heart attack. Those guidelines—based on years of experience and clinical practice—track risk factors for heart disease such as age, cholesterol level, and blood pressure.

But scientists have demonstrated that self-taught artificial intelligence algorithms perform better than the standard medical guidelines, not only significantly increasing heart attack prediction rates, but also identifying unanticipated risk factors such as ethnicity, arthritis, and kidney disease.

In a study released in April, Stephen Weng, an epidemiologist at the University of Nottingham in the United Kingdom, and his colleagues had four different machine-learning algorithms crunch a data set of 378,256 electronic medical records of patients in the United Kingdom. They used about 78 percent of the data to identify patterns—identifying important predictors associated with cardiovascular events. They tested the algorithms’ performance on the remaining 22 percent of the data.

Each of the four AI methods performed significantly better than guidelines published by the American College of Cardiology/American Heart Association (ACC/AHA). In a test sample of 83,000 records, the best performing algorithm—a neural network—predicted 7.6 percent more events than the ACC/AHA guidelines, with 1.6 percent fewer false positives, amounting to 355 more patients whose lives might have been saved through preventative measures.

Some of the strongest predictors not included in the ACC/AHA guidelines were severe mental illness and the use of oral corticosteroids. But none of the algorithms included diabetes, which is on the ACC/AHA list, as a Top 10 risk factor.

“I can’t stress enough how important it is,” Elsie Ross, a vascular surgeon at Stanford University in Palo Alto, Calif., who was not involved with the work, told Science magazine. “I really hope that doctors start to embrace the use of artificial intelligence to assist us in care of patients.”

Skilled smugglers

Prisons have always had to deal with attempts to smuggle contraband inside their walls. But criminals using increasingly sophisticated drone technology to smuggle drugs and mobile phones to prisoners has become so widespread in the United Kingdom, the country’s government has established a special task force to deal with the problem.

Drone photo illustration Rachel Beatty

The team, made up of both police and prison officials, will share intelligence on drones recovered in smuggling attempts and pass the information and leads to local police and organized crime officers, according to a U.K. government announcement.

The formation of the task force came after high-profile drone smuggling convictions in which two men were sentenced for a combined total of 10 years for smuggling contraband worth around $60,000 into prisons across the United Kingdom. —M.C.

Drying by sound

We spend a lot of energy getting our clothes dry. Clothes dryers annually consume 4 percent of household electricity use, according to the U.S. Energy Information Administration. The technology for drying clothes hasn’t changed for decades: heating the air around the tumbling clothes to evaporate the water.

(Krieg Barrie)

But a new approach to clothes-drying technology that uses high-frequency sound waves, rather than heat, may speed up drying and use 70 percent less energy than conventional dryers.

Scientists at the Oak Ridge National Laboratory have developed a tumble dryer whose drum is lined with panels that convert electrical energy into ultrasonic sound waves that vibrate the water droplets out of fabrics. The droplets then form a cool mist that drains into a collection tank.

The research was conducted in partnership with General Electric Appliances, which plans to use the ultrasonic technology in future tumble dryer models, according to MIT Technology Review. —M.C.