This week’s list of data news highlights covers July 1 – 7, 2017, and includes articles about how a Chinese initiative to develop open source self-driving cars and an artist using AI to make art.
Researchers at Stanford University have developed a machine learning system that can identify heart arrhythmias in data from an electrocardiogram (ECG) more accurately than human experts. Heart arrhythmias can cause patients to have irregular heartbeats, however identifying when irregular heartbeats are indicative of an arrhythmia and require medical intervention or are just benign. The researchers partnered with wearable ECG device manufacturer iRhythm to train their system on 30,000 30-second heartbeat recordings from arrhythmia patients. In a test on 300 undiagnosed recordings, the system proved more effective at identifying arrhythmias than cardiologists.
Baidu has organized a partnership of over 50 ridesharing companies, manufacturers, technology firms, and Chinese automakers to support its Apollo project, an initiative to accelerate the development of open-source autonomous vehicle technology. Apollo is a platform consisting of self-driving software, cloud services, and hardware such as cameras and LIDAR sensors. Baidu plans on testing Apollo vehicles this month and aims on having fully autonomous Apollo cars driving on public roads in 2020.
Fertility experts at Sao Paulo State University in Brazil have partnered with Boston Place Clinic in London and the British Fertility Society to develop an AI system that can predict which embryos are more likely to result in successful in vitro fertilization (IVF). The AI system can analyze 24 different characteristics in images of embryos, such as texture and cell homogeneity, that are difficult or impossible for human eyes to distinguish, and determine how likely IVF of an embryo will result in a healthy baby. In a test evaluating cattle embryos, the AI system proved to be more accurate than human embryologists.
Seven European research institutions are working on a EU-funded project to develop BADGER, a “roBot for Autonomous unDerGround trenchless opERations, mapping and navigation,” that can autonomously steer clear of obstructions as it bores holes underground. Underground excavations to install pipes and cables or dig tunnels can be time consuming and expensive due to underground hazards, such as large rocks and existing infrastructure. BADGER will use ground penetrating radar and sensors to detect and avoid such obstacles as it digs, allowing developers to install new pipes and cables without having to excavate straight lines and worry about disturbing existing infrastructure.
Researchers at City, University of London have launched a project to develop an application called DMINR that uses machine learning to help journalists analyze data and fact-check information. The researchers will test DMINR with the help of 30 European newsrooms, including the Guardian, Telegraph, and Ireland’s national broadcaster RTÉ. The goal of DMINR is to help journalists conduct investigative journalism more effectively with fewer resources and more quickly identify and dispel fake news.
Berkeley, California startup AnimalBiome is offering microbiome sequencing services for cats and dogs that allows customers to learn about their pet’s health and potentially identify digestive disorders. AnimalBiome sequences DNA of bacteria found in pet’s biological samples and uses analytics techniques to “barcode” each kind of bacteria based on particular regions of their DNA. Customers can view the composition of their pet’s microbiome in a dashboard and compare this data to other similar breeds.
Furniture retailer West Elm has launched an AI tool called the West Elm Pinterest Style Finder that uses artificial neural networks to analyze customers’ Pinterest profiles to identify their sense of style and recommend furniture. Pinterest allows users to build galleries of images based on certain topics or designs. Style Finder analyzes these galleries to identify desired styles, which it can do with less than 10 images, and provides recommendations from West Elm’s inventory with similar designs or patterns.
The U.S. Food and Drug Administration (FDA) has developed a plan to use computer modeling to make the regulatory approval process for new drugs and medical devices faster and less expensive. FDA will expand on existing efforts to use computer modeling, such as how it simulations to help predict the performance of treatments for Parkinson’s and Alzheimer’s disease and relies on virtual patients for device testing. FDA also plans to expand its capacity for high performance computing to ensure it can quickly analyze the large amounts of data involved in the drug approval process.
The Welsh government has launched an initiative called the Genomics for Precision Medicine Strategy to accelerate the development of personalized treatments for genetic diseases. The government will provide £6.8 million ($8.6 million) in funding to implement the strategy, which includes developing a single genomics services for the country, increasing its number of genomics and data analytics expertise, and improving its data sharing infrastructure.
Artist Mario Klingemann is making art with artificial neural networks that he calls “neurographs” by training his neural networks on photos, videos, and drawings to have them produce abstract imagery. Klingemann uses a variety of different machine learning techniques to produce his neurographs, such as using adversarial networks to generate realistic footage based on the styles of 21st century selfies and 19th century portraits.