This week’s list of data news highlights covers February 13, 2021 – February 19, 2021 and includes articles about monitoring breeding grounds for malaria and analyzing the effectiveness of COVID-19 vaccines.
Fujitsu Laboratories, the research and development division of the Japanese IT equipment and services company, and the International Research Institute of Disaster Science have used Fugaku, the world’s fastest supercomputer, to develop an algorithm that forecasts coastal flooding conditions. The team trained the algorithm to predict conditions by analyzing the waveforms of tsunamis from 20,000 Fugaku-generated simulations. The algorithm then used this analysis to predict flooding conditions and the impact of localized waves on urban infrastructure, such as roads and buildings. Although researchers trained the algorithm on a supercomputer, local governments can use the algorithm on ordinary PCs to predict flooding that may occur in the next 30 years along the Nankai Trough faultline.
Syrian Archives, an initiative that collects visual archives relating to human rights violations, has partnered with VFRAME, an open-source project that creates object detection models for illegal munitions, to identify weapons of war used during the Syrian civil war. Since there are not enough images of detonated munitions to train algorithms, the team created synthetic data including 2-D images meant to replicate the environments in Syria, and 3-D models to recreate post-blast videos in various locations around Germany. The team hopes their work can help bring war criminals to justice.
ZzappMalaria, an Israeli start-up focused on using software to eradicate malaria, has developed a mobile app that locates bodies of water used as breeding grounds for Malaria-carrying mosquitoes. The app uses a neural network to extract locations of affected houses from satellite images and analyzes the topography and radar data from the images to create a heatmap of water bodies probable for transmission. The app then alerts field workers about which specific locations could benefit from spraying insecticide treatments. It is currently operating in a Ghanaian town of around 200,000 people.
Researchers at Cranfield University and the University of Leeds in England have analyzed human waste data from 48 cities in Africa, Asia, North America, and South America to better understand how nature safely treats waste in regions that do not have proper sewage systems. By looking at field measurements of waste management techniques not connected to sewers, such as pit latrines and septic tanks, the team estimated that soil filters 38 million tons of liquid waste each year, which is equivalent to 3.2 billion pounds-worth of commercial water treatment. Researchers are hopeful that their work can help develop natural sanitation techniques for the 4 billion people who do not currently have access to safe sanitation services.
Doctoral students at the Indian Institute of Science have developed a platform that can identify which camera feeds crime investigators should analyze for video footage of an object or person of interest. The platform’s algorithm selects which cameras to analyze based on factors of the crime, such as the local road network, camera location, and the last-seen position of the entity being tracked. The team tested the algorithm using images from 1,000 virtual cameras that cover a 7 square kilometer region in Bangalore, and found that the algorithm tracked the person of interest within 15 seconds.
Researchers at Mayo Clinic, an American nonprofit academic medical center, have collaborated with the biomedical software start-up nference, to study the effectiveness of COVID-19 vaccines. The company’s software scanned and analyzed electronic physician notes from the medical records of early vaccine recipients for underlying health conditions, pathology reports, and clinical notes and used AI to create matched control groups of other patients who were demographically and geographically identical. By looking for people who are identical, the software compares how vaccine and non-vaccine recipients will fare in terms of infection rates.
Scientists at the National Astronomical Observatory of Japan have used ATERUI II, the world’s fastest supercomputer for astronomy simulations, to verify theories about how universes expanded after the Big Bang. The team first used ATERUI II to simulate the evolution of 4,000 universes and then reverse engineered the simulation to see how well the supercomputer could reconstruct the starting state of the simulations. They found the technique to be very effective and hope to use it to better understand how universes grow by trillions in size over time.
J.B. Hunt Transport Services, an American transportation and logistics company, is working with Google to develop machine learning models that will better match shippers and carriers to create more efficient deliveries. Google’s developers will use data on market demand, carrier capacity, and the locations, destinations, and loads of trucks to create models that shippers and carriers can use to predict future supply and demand needs, fuel and transportation costs, delivery times, and pricing data.
Scientists at the University of Gothenburg have developed an algorithm to assess the severity of melanoma, a serious form of skin cancer that controls the pigment of skin. The scientists trained the algorithm to identify different skin lesions using 937 images collected from a dermatological microscope. When tested on 200 images, the algorithm diagnosed melanoma as accurately as seven independent dermatologists.
Virti, a healthcare education software company from England, has created a training program for doctors to simulate realistic patient interactions using speech recognition, computer-generated patients, and virtual reality (VR). Doctors wear VR goggles to engage in simulations and can ask virtual patients questions about their conditions. The system then uses speech recognition to analyze the tone, cadence, and quality of a doctor’s answer to assess what needs improvement, such as conveying information with less medical jargon.
Image credit: Jevgenji Voronov