Home BlogWeekly News 10 Bits: the Data News Hotlist

10 Bits: the Data News Hotlist

by Michael McLaughlin
by
MRIs and Machine Learning

This week’s list of data news highlights covers August 18-24, 2018, and includes articles about AI detecting mood disorders and an effort to find unique identifiers in human brains.

1. Using Machine Learning to Make MRIs Faster
Facebook and New York University (NYU) are working together to drastically reduce the time it takes to complete a magnetic resonance imaging (MRI) scan. While other researchers are already using AI to read medical images, this partnership hopes to reduce the time a scan takes by 90 percent. It can take up to an hour to complete a full MRI scan—sometimes while a patient is in significant pain—because the MRI combines multiple 2D images to create a 3D one. Facebook and NYU plan to use machine learning to allow MRI machines to take fewer photos because AI can fill in the gaps between images.

2. Helping Autonomous Vehicles Drive in the Snow
WaveSense, a startup that creates tools for autonomous vehicles, is using ground-penetrating radar (GPR) to help autonomous vehicles drive better in the snow. While MIT Lincoln Laboratory, a U.S. Department of Defense research center, first developed the technology to help military vehicles detect improvised explosive devices underground, WaveSense is combining GPR with an algorithm that uses the GPR data to create a five-dimensional picture of the roadway for autonomous vehicles. Until now, autonomous vehicles have usually relied on LIDAR, a spinning laser, to map the space around the vehicle. LIDAR can confuse snowflakes with solid objects and relies on seeing lane markers, but WaveSense’s technology can detect the snow and provide redundancy when poor conditions confuse LIDAR.

3. Using Smart Roads to Detect Accidents
Denver is installing smart pavement in one of its intersections as a pilot test. The pavement, created by the smart-roads startup Integrated Roadways, uses accelerometers, fiber optic cables, and a magnetometer to detect the size, speed, and position of vehicles on the road. This data is particularly useful to detect when vehicles drive off the road. Colorado is hoping to install smart pavement along the mountainous Highway 285 in early 2019. The highway is too narrow to widen or add guardrails, so the pavement could be crucial in alerting the authorities when a vehicle has careened off the road.

4. Getting Better Wind Data for More Accurate Forecasts
The European Space Agency has launched Aeolus, a wind-mapping satellite that could, within a year, start improving the accuracy of weather forecasts. The satellite measures wind patterns by sending an ultraviolet laser to the ground of the earth and measuring the return signal as air molecules and particles scatter the light. The launch is significant because current wind-mapping technology only detects wind at certain heights and locations, but Aeolus will provide meteorologists with data from all over the world to compare with their models.

5. Leveraging LinkedIn Data for Economic Insight
LinkedIn has announced that it is allowing academic researchers access to its data. Though the company has collaborated with researchers on a limited basis in the past to examine areas, such as the gender gap, LinkedIn is now openly accepting proposals. While the company is accepting project ideas in artificial intelligence, analytics, and economics, all research must take place in its sandbox environment. That means researchers will not be able to download or retain the anonymized data. Nonetheless, LinkedIn is hoping greater access to its data will lead to more insights into the labor market and economy.

6. Fighting Fake Photos Online
Two University of California, Berkeley undergraduates have developed SurfSafe, an Internet browser extension that can help users identify if an image is fake when they hover their mouse over it. While some organizations, such as Facebook, are using AI to remove content that violates their policies, SurfSafe stores the digital footprints of photos on more than 100 news and fact-checking sites. It then shows users the earliest known instance of the image they are seeing online. This system then warns users if the image has appeared on fact checking websites. The tool not only allows users to flag an altered photo, it also stores signatures of every image users see to help improve the tool.

7. AI Reduces Prevalence of Sleeping Sickness
Columbia University researchers are applying machine learning to reduce the prevalence of sleeping sickness, a disease caused by tsetse fly bites that results in fevers, pain, and potentially death, in Senegal. Female tsetse flies only mate once a lifetime. The researchers trained a machine-learning algorithm on pictures of the flies as pupae, a stage in life where the flies are inactive and immature, to identify which flies are males. These males are then sterilized and the researchers introduce them to areas where there are infestations of females. This research has led to both a 98 percent reduction in the tsetse fly population and a reduced prevalence of sleeping sickness in Senegal.

8. Using AI to Detect Mood Disorders
Researchers from the Lawson Health Research Institute, the Mind Research Network, and the Brainnetome Center have developed an AI system that can detect the characteristics of bipolar disorder and major depressive disorder (MDD) in brain scans with over 92 percent accuracy. The researchers trained an algorithm on brain scans of 78 young adults, 66 of which had one of the disorders, as well as 33 individuals who had no mental illnesses. The scans showed subtle differences between patients who suffered from bipolar disorder and MDD, and the researchers are working to create a system that can detect a wider range of disorders in the brain.

9. Detecting Lung Cancer with AI
University of Central Florida researchers have developed an AI system powered by computer vision that detects small specks of lung cancer in computed tomography (CT) scans with roughly 95 percent accuracy. The system vastly outperforms human radiologists, which can only detect the same specks with 65 percent accuracy. The researchers developed this system by analyzing over 1,000 CT scans provided by the National Institutes of Health.

10. Developing a Unique Brain Identifier
Researchers at Oregon Health and Science University have used machine learning to uniquely identify individuals and families from strangers by analyzing the connections between different regions of the brain. The researchers found that while there is little difference between individuals in the areas of the brain that perform basic tasks, such as moving the hand, the neural connections between distinct brain regions are roughly 30 percent unique to the individual. This research could help predict how a brain will develop over time and identify which characteristics of neural connections mean that individuals are at a higher risk for developing neurological conditions.

Image: GeorgeWilliams21

You may also like

Show Buttons
Hide Buttons