This week’s list of data news highlights covers August 3-9, 2019, and includes articles about an AI system that can detect the signs a volcano will erupt and researchers who have mapped the brain of the fruit fly.
Researchers from MIT have developed a machine learning tool that can predict the cognition scores of patients at risk of developing Alzheimer disease up to two years in advance. The researchers trained the tool using 1,700 Alzheimer and healthy patients’ cognitive tests scores and biometric data. The tool could help identify at-risk patients for clinical trials, which can lead to the development of better drugs.
Researchers from multiple German and Italian organizations have developed an AI system that can analyze satellite images to detect signs that indicate a volcano could erupt. The system analyzes images for signs of large deformations in volcanoes, which indicate that magma is moving underground. The system has already spotted signs of recent eruptions in Hawaii and Italy and sends automated emails to users who have signed up for updates.
Researchers from McGill University in Canada and Amirkabir University of Technology in Iran have shown that machine learning algorithms can accurately classify the level of expertise of neurosurgeons. The researchers fed algorithms data from 50 individuals, ranging from neurosurgeons to medical students, who performed 250 complex tumor removals in a virtual reality simulator. The data included outcomes such as the amount of the tumor removed, blood loss, and the amount of force an individual applied. The best performing algorithm, which assessed the level of expertise of each participant with 90 percent accuracy, could help determine the ability of neurosurgeons.
Hartford Financial Services Group, a U.S. investment and insurance company, is using AI software to analyze workers’ compensation claims to find people at risk of opioid addiction. The software’s algorithms scan medical records for signs of an at-risk individual, such as anxiety, insomnia, a recent divorce, or a death in the family. Between 2014 and 2017, the software has helped reduce the number of opioid prescriptions for its workers’ comp policyholders by 45 percent.
Researchers from Google, the Howard Hughes Medical Institute (HHMI) Janelia Research Campus in Virginia, and Cambridge University have used AI to help map the brain of the fruit fly. The researchers used an electron microscope to take over 40 trillion pixels of brain imagery and used a generative adversarial network to fill in missing data. The research can help improve the understanding of the nervous system’s role in health and disease.
FedEx is using an autonomous delivery robot to make same-day deliveries in Manchester, New Hampshire. The 4-foot, 2-inch robot can travel up to 10 miles per hour on sidewalks and uses LIDAR, cameras, and machine-learning algorithms to plot its path, avoid objects, and follow safety rules. The robot can carry packages up to 100 pounds and signals when it turns or stops.
Researchers from MIT have developed a deep-learning model that can accurately predict sites or objects that trigger smokers to smoke cigarettes, such as a pool table. The researchers developed the model using 5,000 pictures of sites in which smokers did or did not frequently smoke. The model, which identified areas associated with smoking with 77 percent accuracy, may be able to help researchers optimize a smoker’s environment during an attempt to quit.
Researchers from the Netherlands have shown that machine learning can accurately predict if an inpatient at a psychiatric facility will display violent behavior. The researchers used clinical notes from electronic health records to train their model to assess the risk a patient would be violent within the first four weeks of admission. The researchers found that the model could correctly classify patients with a nearly 80 percent probability.
Researchers from LinkedIn and Indiana University have used LinkedIn data to create a global map of labor flow. The researchers analyzed 130 million job transitions from more than 500 million people between 1990 and 2015, finding strong connections between an influx of educated workers and productivity growth. In addition, the researchers found that industries such as retail are experiencing significant declines in college-educated workers.
A group of researchers led by an individual from Swinburne University of Technology in Australia have developed an AI system that can detect fast radio bursts, which are powerful flashes of radio waves in space that only last milliseconds. The researchers trained the system to distinguish fast radio bursts from millions of other radio waves, such as waves from mobile phones, lightning storms, and the Sun. Fast radio bursts, which originate billions of light-years away from the Earth, can aid in the study of matter between galaxies that is difficult to see.