This week’s list of data news highlights covers April 29 – May 5, 2017, and includes articles about a new initiative to use supercomputers to advance veterans’ health and a report ranking national governments on open data.
Startup OtoSense is developing machine learning software for cars that can analyze sound to detect signs of danger, such as changes in the way a vehicle’s brakes sound or the siren of an oncoming emergency vehicle. The software can be trained to identify specific noises and automatically diagnose problems with the vehicle itself, and in tests, it could correctly identify undesirable changes in a car’s engine, wheels, and other components with 95 percent accuracy. OtoSense has partnered with French car manufacturer PSA Group to test their software in its vehicles.
The U.S. Department of Veterans Affairs (VA) and Department of Energy (DOE) have launched the VA-DOE Big Data Science Initiative to use DOE’s supercomputers to analyze veterans’ health data to advance research into suicide prevention, cancer, heart disease, and other issues affecting veterans. DOE’s National Laboratories will oversee the initiative and will analyze electronic health records and genomic data from over 500,000 veteran volunteers participating in the VA’s genomics-focused Million Veteran Program.
Chemical sensor manufacturer Owlstone has developed a prototype sensor that can analyze the molecular composition of a patient’s odor to help detect the presence of certain diseases. Some diseases, such as cancer or diabetes, can cause subtle but distinct changes in the smell of a patient’s break, skin, or bodily fluids, and the sensor works by ionizing a sample of this odor and running an electric current through it to identify the presence of certain chemicals produced by these changes. The British National Health Service is conducting a 3,000-person clinical trial to use Owlstone’s sensor to diagnose lung cancer, and Owlstone is exploring the potential for its sensors to determine the appropriate drugs for asthma patients based on analysis of their breath.
Researchers at the Illinois Institute of Technology in Chicago have created a machine learning system capable of predicting U.S. Supreme Court decisions with 70 percent accuracy, and predict how individual judges will vote with 72 percent accuracy. The researchers trained their system on the Washington University in St. Louis’s Supreme Court database, which contains information about every Supreme Court case from 1791 through 2015. In tests, the system was more accurate at predicting the outcomes of cases than legal experts.
The Public Health Agency of Canada, the U.S. Centers for Disease Control and Prevention, and the U.S. Department of Health and Human Services have launched the Healthy Behavior Data Challenge to encourage the development of innovative new methods for collecting health data. Both countries’ respective health agencies are offering funding to participants that propose the best method for using wearables, mobile apps, or social media to collect health data related to nutrition, physical activity, sleep, and sedentary behavior.
Nvidia has developed an architecture for neural networks used in its self-driving car technology that can help humans understand what a car’s algorithms are focusing on to make decisions by highlighting these areas in video. Nvidia initially trained an autonomous system to drive a car just by observing human behaviors as they drive. Then, despite Nvidia not teaching the autonomous system specifically what it should pay attention to as it drives down a road, the system nonetheless focuses on the edges of roads, lane markings and other cars to make decisions, just like a human driver would.
Scientists at the British Antarctic Survey have assessed the entire global population of the Northern Royal albatross by using high-resolution satellite imagery. Northern Royal albatross typically nest on inaccessible rocky islands, making traditional population survey methods involving people tallying the birds on-site very difficult. The scientists analyzed imagery from the U.S. government’s DigitalGlobe WorldView-3 satellite, which can identify features on the Earth as small as 30 centimeters long—a resolution high-enough to identify every individual bird in their nesting grounds.
Silicon Valley startup Nauto has developed an aftermarket device that uses sensors and deep learning to monitor the behavior of drivers and alert them when they become distracted. The device has inward- and outward-facing cameras that track a driver’s eye, chest, and head movements, as well as the road, and compares this against a control image of a non-distracted driver. If it detects a significant difference, such as drivers looking down at their phone, it will warn them to pay attention to the road.
Microsoft and the nonprofit Global Initiative for Inclusive Information and Communication Technologies have developed a free toolkit for smart city development detailing strategies, and technologies that can help cities prioritize accessibility as they deploy smart city technology. The toolkit includes communication plans, procurement guidelines, accessibility standards, a testing roadmap, and a database of accessibility-related smart city applications.
The nonprofit Open Knowledge International has published its fourth annual Global Open Data Index ranking 94 national governments on how well they publish 15 key datasets as open data, which can be indicative of a country’s open data policies and practices as a whole. Open Knowledge International and volunteers scored the the datasets, which include important topics such as government budgets, national statistics, weather forecasts, and national maps, based on whether they meet standards for openness, such as whether or not the data is openly licensed and timely. Australia and Taiwan tie for first place in this year’s rankings.
Image: JJ Harrison.