This week’s list of data news highlights covers February 8-14, 2020, and includes articles about an AI model that helps robots pick up clear objects and a system that is helping diagnosis the coronavirus.
Researchers led by an individual from Stanford University have developed a machine learning algorithm that can predict if patients with depression would benefit from antidepressant medication. The researchers tested the algorithm, which analyzes electrical activity data in the brain to make its predictions, on data from more than 200 individuals. The algorithm could help doctors decide the proper treatment for patients.
The U.S. Food and Drug Administration (FDA) has approved software that uses AI to help medical professionals without specialized experience take echocardiograms—ultrasound images of the heart. The system, which San Francisco-based AI startup Caption Health created, offers real-time guidance on how to maneuver the ultrasound probe to achieve images that meet diagnostic standards. The startup trained the system to identify high-quality ultrasound images and automatically record the best videos from a particular angle.
Researchers from Google, Columbia University, and Synthesis AI, a San Francisco-based startup, have developed an AI model that helps robots pick up transparent objects. Algorithms that analyze data from cameras and lidar struggle to estimate the depth of transparent objects, such as glass containers, because they assume objects reflect light evenly in all directions—transparent objects both reflect and refract light. The researchers trained their model on simulated and real images that contained transparent objects and their corresponding depths, finding that the model helped a robotic gripper improve its success rate of gripping transparent objects from 12 percent to 74 percent.
Researchers from Omdena AI, an organization that connects AI experts to solve challenges in two months, have developed a system that uses AI to predict the safest route to take after an earthquake in Istanbul. The system analyzes satellite imagery to determine the density of buildings and the width of roads. The researchers then integrated this data with Open Street Map data to chart the safest and shortest paths between locations.
Researchers from the University of Science and Technology of China have demonstrated quantum entanglement—a state in which two connected atoms influence each other’s state—over 30 miles of fiber optic cables. This entanglement allows physicists to measure one atom and instantly know the state of the other atom. Quantum entanglement can lead to a quantum internet that would allow for secure communications because an interception of the communication by a hacker would compromise the entanglement, revealing the presence of the hacker.
Researchers from Indiana University have found that investment recommendations by robo-analysts—machines that automate research analysis—outperform the recommendations made by human analysts. The researchers analyzed 76,000 reports from seven robo-analyst firms between 2003 and 2018; finding that the automated services, which analyze documents such as income statements, balance sheets, and company reports, also produced a more balanced distribution of buy, hold, and sell recommendations than humans. This finding suggests that the robo-analysts exhibit fewer biases than humans.
Researchers from MIT and Harvard have shown how analyzing virus genomes from patients can help map the path of an outbreak and inform efforts to control the spread of viruses in the future. The researchers analyzed genomic data of the mumps virus during outbreaks that occurred in 2016 and 2017 in Massachusetts. Their analysis revealed that cases of the virus at Harvard and in East Boston, which public officials believed were unrelated because they lacked an obvious connection, were genetically similar. This revelation helped the researchers identify transmission links between the two different communities.
Infervision, a startup based in Beijing, has developed an AI system that has helped diagnose more than 2,000 cases of the coronavirus. The system analyzes computed tomography (CT) images of lungs to detect ground-glass opacities, partial fillings of air spaces in the lungs by masses of fluids or cells, that may develop as a result of the virus. The system is helping to shorten the time it takes to perform a diagnosis.
Oil and gas firm Aker BP will use Spot, a four-legged robot from Boston Dynamics, to autonomously patrol an oil rig in the Norwegian Sea. Spot can help the firm take mechanical readings, inspect ship equipment, and capture data on gas leaks. The firm will also test the robot’s ability to investigate areas it deems too dangerous for human engineers.
Researchers from multiple European and U.S. universities have developed a deep learning algorithm that can accurately analyze computed tomography (CT) scans to detect calcified plaque in coronary arteries, which can indicate an individual’s risk for heart disease. The researchers trained and tested the algorithm using more than 10,000 CT scans from European and U.S. patients. Early diagnosis of an individual’s risk for heart disease can help doctors suggest appropriate lifestyle changes.