This week’s list of data news highlights covers June 4-10, 2016 and includes articles about a Moneyball-style approach to basketball and a new system for diagnosing concussions.
Vice President Biden announced the launch of the Genomic Data Commons (GDC), an open-access database of genomic and clinical data run by the National Cancer Institute (NCI) to support cancer research and accelerate the development of more personalized treatments. The GDC contains data from 12,000 patients, including the genetic makeup of their cancers, the type of treatment they received, and how their cancers responded to these treatments, and NCI will regularly contribute more data from future clinical trials.
Google is developing algorithms for its self-driving car that will enable it to honk in the style of a “patient, seasoned driver,” so that the horn sounds encourage safer driving without confusing or distracting other drivers. The algorithms will honk differently based on its analysis of the situation, playing two short, quieter beeps when it only needs to get a driver’s attention, and a longer, louder beep in more urgent situations. This approach should help self-driving cars communicate with and operate on the same roads as human drivers.
The Golden State Warriors, Cleveland Cavaliers, and 14 other basketball teams are using analytics software developed by sports data firm Second Spectrum to measure and improve a wide variety of coaching, playing, and management decisions. For example, Second Spectrum uses computer vision and machine learning algorithms to analyze footage about a player’s movements, make accurate predictions about the success of potential plays, and help teams make more informed decisions about signing free agents and trading players. Baseball teams have used similar approaches for years, but it has only caught on with basketball teams recently as advances in computer vision technology make it possible to analyze nuanced player movements, which are more relevant in sports like basketball, that traditional statistics do not capture.
Researchers at Microsoft have developed a data mining technique that could help warn a search engine user he or she may have pancreatic cancer before they are officially diagnosed. The researchers identified Bing search engine users who searched for phrases indicating they were officially diagnosed with pancreatic cancer, such as “why did I get cancer in pancreas,” and then analyzed previous searches to reveal searches related to pancreatic cancer symptoms. By applying this technique in reverse, the researchers could identify up to 15 percent of pancreatic cancer patients before their diagnoses, which could serve as a valuable early warning method for the disease and lead to early detection and more effective treatment.
Researchers at the Massachusetts Institute of Technology and Harvard University have developed an algorithm called the Continuous High-Resolution Image Reconstruction (CHIRP) that pieces together data from arrays of radio telescopes around the world, which could allow astronomers to capture the first real image of a black hole. Radio telescopes can observe objects much farther away than traditional telescopes, but would have to be impossibly large in diameter to capture high-resolution images. CHIRP works around this obstacle by coordinating and stitching together the measurements of observatories participating in the Event Horizon Telescope project, an international collaboration to study black holes, effectively simulating a much larger radio telescope. With this approach, astronomers could potentially directly observe black holes and their immediate environments, which thus far has been impossible due to their relatively small size and extreme distance away from Earth.
Startup BrainScope has developed a technique for improving concussion diagnoses, for which no objective test currently exists, using head-mounted sensors and a smartphone app. Doctors diagnose concussions based on analysis of a patient’s symptoms, which vary and can be hard to detect, and a computerized tomography (CT) scan, which only reveals if a patient’s brain has been seriously injured, but not if he or she suffered a concussion. BrainScope instead analyzes abnormalities in the brain’s electrical activity that can directly reveal whether or not a patient has suffered a concussion. Currently, BrainScope only uses its technology to help doctors decide whether or not to administer a potentially risky CT scan, but it is working on producing a standalone diagnostic version of its test.
University of Minnesota computer scientists funded by the Federal Highway Administration are deploying a pilot system that uses cameras and computer vision algorithms to monitor parking lots and rest stops to inform long-haul truckers of available spaces. Long-haul truckers sleep in their trucks en route, but 75 percent report having trouble finding safe places to park, causing many to resort to parking in dangerous locations or continuing to drive while tired. The system, which first launched as a small pilot in 2013, can inform truckers of available parking spaces on their routes in real time with 95 percent accuracy.
Researchers at the University of Wisconsin have installed underwater wave pressure sensors at several beaches along Lake Superior that can provide officials with real-time data about small but potentially dangerous strong underwater currents, known as riptides, which can drown swimmers. The National Weather Service and local forecasters will combine data from these sensors with data about wind conditions to issue accurate riptide warnings and help researchers create accurate models of underwater currents.
South Africa’s Council for Scientific and Industrial Research (CSIR) has unveiled its new one-petaflop supercomputer, named Lengau, the most powerful high performance computing system in Africa and among the top 200 most powerful in the world. CSIR developed Lengau to support data-intensive projects such as climate modeling, genomics, and astronomy.
Doctorate students at the Lausanne Federal institute of Technology (EPFL) in Switzerland have launched kickoff.ai, a website that uses machine learning algorithms to predict the outcomes of the Union of European Football Associations (UEFA) Euro 2016 tournament, which began June 10. The website models individual players’ performances, unlike traditional methods which focus on the performance of a whole team, and uses a statistical technique called Bayesian inference to predict the likeliness of different results, which is useful for estimating the outcomes of matches with teams that have never played each other before. The students tested their model on the Euro 2012 tournament and found it produced similar odds as professional bookmakers.
Image: Alain r.