Hearts And Sciences
What is it? Researchers at Geisinger Health System trained neural networks to examine electrocardiograms to identify patients at risk of developing an irregular heartbeat or dying within a year. It’s one of the first times an artificial intelligence system has used ECGs to predict, rather than diagnose, health problems. The system’s powers of prediction were so sharp that it could detect a risk of death even in ECGs that had been deemed normal by cardiologists.
Why does it matter? “This is exciting and provides more evidence that we are on the verge of a revolution in medicine where computers will be working alongside physicians to improve patient care,” said Brian Fornwalt, a Geisinger professor and senior author of two studies just released at the American Heart Association’s Scientific Sessions 2019 in Philadelphia. It’s not just heart scans that AI might revolutionize: It can also help doctors sort through chest X-rays and quickly separate out urgent cases.
How does it work? As with other neural networks, the AI developed by Geisinger’s doctors was trained using a mountain of data: In this case, 2 million ECG results culled from more than three decades from the health system’s archives. In one study, the AI was trained specifically to look for indications of atrial fibrillation, which is associated with risk of heart attack or stroke; in the other, the system analyzed ECG signals to identify patients at risk of dying within the following year from any cause. The latter is where the AI outshone the human doctors in ECG readings, uncovering subtle patterns that indicated heightened risk. Fornwalt said, “This is the most important finding of this study. This could completely alter the way we interpret ECGs in the future.”
See Me, Feel Me
What is it? At Northwestern University, researchers developed an “epidermal virtual reality” system that could offer the possibility of long-distance physical contact. Essentially, it’s a soft, thin, flexible patch that transmits the feeling of touch.
Why does it matter? Tech like this could add another dimension to video chats between family members, for instance — “Imagine holding hands with a loved one on the other side of the world,” as Northwestern puts it in a press release — or to online gaming and entertainment. It could also lead to more sensitive prosthetic limbs. Northwestern’s Yonggang Huang, who co-led the research, said, “We are expanding the boundaries and capabilities of virtual and augmented reality. By comparison to the eyes and the ears, the skin is a relatively underexplored sensory interface that could significantly enhance experiences.”
How does it work? The prototype patch, 15 centimeters by 15 centimeters, consists of millimeter-scale actuators that resonate at 200 cycles per second — “where the skin exhibits maximum sensitivity,” per Northwestern. They’re embedded in a soft silicon polymer that could be sewn into wearable garments like VR shirts, gloves and headsets. The patch connects wirelessly to a touchscreen interface, which another user can manipulate to create the sensation of touch on the patch wearer. The tech is described further in a new paper in Nature.
The Critical Ingredient Is Time
What is it? A team of doctors at the University of Maryland School of Medicine has begun a trial that involves placing patients in suspended animation while treating traumatic injuries like bullet or stab wounds, New Scientist reported. It’s the first time such a procedure, which has been successfully tested on animals, has been tried on human patients.
Why does it matter? Referred to officially as emergency preservation resuscitation, the technique is a kind of last-ditch effort to save the lives of critical patients — such an emergency measure, in fact, that the FDA exempted doctors from needing to obtain patient consent to try the procedure, the magazine said. Patients’ hearts will have stopped beating, they’ll have lost more than half their blood, and they’ll have less than a 5% chance of survival before the Baltimore doctors undertake EPR. The intervention cools the body to 10 to 15 degrees Celsius by replacing the blood with ice-cold saline, according to the story. The objective of the trial is to compare the results of 10 patients who received EPR with 10 eligible patients who didn’t receive the treatment because the correct medical team wasn’t in the hospital.
How does it work? When the heart stops beating, cells don’t get the oxygen they need to produce energy, and irreversible brain damage can occur in as little as five minutes. At super-low temps, cells need less oxygen, so rapidly cooling the body buys doctors time (about two hours) to repair the damage on critical patients before rewarming their bodies and restarting their hearts. The results of the trial in Baltimore, which would reveal how many patients have survived as a result of the procedure, aren’t public yet; doctors hope to release their full findings by the end of 2020, New Scientist said.
Photonic Hammer Time
What is it? 3D printing is transforming manufacturing in sectors from aviation to healthcare. Now researchers at the Laboratory of Thermomechanical Metallurgy at Switzerland’s Ecole Polytechnique Fédérale de Lausanne have come up with a new method that could increase the number of the technology’s applications for metal parts exposed to large temperature changes at high temperatures, which can lead to cracking.
Why does it matter? As EPFL put it in a press release, “the results speak for themselves” — the technique eliminated up to 95% of cracking associated with a nickel-based superalloy, and will now be applied to other crack-prone alloys. The patented technique “could be used to manufacture new power-generating turbine blades or key aircraft components,” the lab reports. It’s described in the journal Additive Manufacturing.
How does it work? Standard 3D printing uses a laser to fuse metal or alloy powders into a predetermined shape, melting the material layer by layer. The EPFL technique involves using a second laser treatment every few layers; called laser shock peening, the second application “acts as a sort of high-energy photonic hammer” that “generates stresses in the component in targeted locations to eliminate cracks.”
This Computer’s A Comedy Critic
What is it? Researchers at Harvard Business School tested computer algorithms to see if they could predict which jokes would elicit a chuckle out of human subjects. AI had the last laugh, of course, coming out ahead against a team of people also tasked with predicting which jokes would land.
Why does it matter? The researchers are thinking about ways to bridge a trust gap between humans and AI in commercial settings; there are plenty of situations in which algorithms provide tailored recommendations for consumers (think Netflix), but surveys have shown that people don’t necessarily have a lot of faith in AI’s abilities in this sphere. The researchers — who published their findings in an article entitled “Making Sense of Recommendations” — went on to probe the fact that, even though computers were better at predicting what sorts of jokes people would like, people still preferred to get those sorts of recommendations from fellow humans. “It is not enough for recommender systems to be accurate,” they concluded, “they must also be understood.”
How does it work? The researchers selected 75 pairs of people, some well-known to one another — including close friends and spouses. While partners predicted each other’s joke responses on a scale of “not funny at all” to “extremely funny,” a computer algorithm did the same, basing their predictions on participants’ previous joke preferences. In the end, the algorithms got it right 61% of the time, compared to 57% for the human predictors. Michael H. Yeomans, who led the study, said, “Humans would seem to have many advantages over computers, but that didn’t matter. I was especially surprised that the recommender system outperformed people who had known each other for decades. I was really rooting for spouses to have an edge!”