This week’s list of top data news highlights covers April 30, 2022 to May 6, 2022 and includes articles on building urban heat maps with crowdsourced data and using augmented reality to better perform combat medicine.
Researchers at Beijing Tongren Hospital and Beijing Airdoc Technology, a health technology company based in China, have created an AI system that can detect retinal abnormalities in real time. The team trained the system to detect signs of retinal diseases like retinal vein occlusions with over 120,000 images of the inner lining of eyes. In tests, the system was better or equal to retinal specialists at detecting 77 out of 100 evaluated retinal diseases.
2. Inventing Artificial Touch Receptors
Researchers at Chemnitz University of Technology in Germany have created an artificial skin that can detect human touch. The skin contains artificial hairs that are attached to internal sensors. When placed in contact with an external object, the hairs transmit the movement to the sensors, which can determine the original direction of force.
3. Predicting Brain Injury Outcomes
Researchers from the University of Pittsburgh School of Medicine and University of Pittsburgh Medical Center have created an AI system that can predict six-month mortality or other unfavorable outcomes for patients with traumatic brain injuries. The team trained the system with brain scans and medical data from over 700 patients. In tests, the system outperformed the predictions of neurosurgeons.
The National Oceanic and Atmospheric Administration (NOAA) has partnered with Capa Strategies, a U.S.-based environmental data analytics company, and the National Integrated Heat Health Information System to create maps of urban heat islands, or areas characterized by a weather phenomenon in which pavement and other urban structures retain more heat than rural or suburban areas. The team and citizen scientists will traverse neighborhoods in 14 cities and counties across the United States with sensors that collect data on temperature, humidity, the time, and volunteers’ locations to build the map.
Researchers at Dartmouth University have created an AI system that can write reviews of wine. The system uses data on past reviews, the wine’s winery or company, its alcohol content, and the cost of a bottle to write the review. The team trained the system on 125,000 reviews published in Wine Enthusiast, an online wine magazine.
6. Strengthening Quantum Research Efforts
President Biden has signed an executive order to move the National Quantum Initiative Advisory Committee from the Department of Energy to the White House. The transition will improve inter-agency coordination on quantum information science.
7. Providing Access to AI Systems
Meta has opened access to a language model with 175 billion parameters to academic researchers, government employees, civil society and academic organizations, and industry research laboratories. Meta released the model under a noncommercial license to encourage and improve AI research efforts.
Wyoming officials have launched a data dashboard to provide residents and businesses with resources on workforce opportunities. The dashboard will provide data on employment opportunities by industry, compensation trends, and customizable career and education recommendations.
Researchers at Raytheon and Valkyries Austere Medical Solutions, a U.S.-based emergency response education company, have partnered to create an AI system that can guide combat medics in performing procedures. The team will use the system in conjunction with augmented reality goggles that can observe the user’s actions in real time and display prompts for next steps or appropriate measurements. The team will train the tool on medical datasets consisting of 2,500 3D videos and 50 million images of procedures.
10. Determining Adverse Drug Interactions
Researchers at the Gwangju Institute of Science and Technology in South Korea have created an AI system that can predict adverse reactions from the ingestion of multiple prescribed drugs. Current predictive approaches rely on the assumption that similar drugs will result in similar interactions and outcomes. This approach improves upon older methods by first determining the effect of a particular drug on gene expression, allowing for a better understanding of how similar drugs differ in certain properties.