This week’s list of data news highlights covers September 7-13, 2019, and includes articles about automating the analysis of brain scans and using facial recognition at the 2020 Olympics.
Apple has announced three new health studies that Apple Watch users can participate in by choosing to share their Apple Watch data. The studies are focusing on women’s menstruation and overall health, how exposure to sound over time affects hearing, and how a person’s heart rate and mobility relates to their hospitalizations and overall heart health. Apple is sharing the data with academic and medical institutions for analysis.
The Sydney Neuroimaging Analysis Centre in Australia has developed an AI system using toolkits from Nvidia to automate the analysis of brain scans. Before the development of the system, it would take the Centre nearly an hour to analyze a brain scan to measure brain volume, lesion number and volume, and changes to lesions from one scan to another. Using AI, the process now takes less than three minutes.
Austrian technology firm smaXtec has developed a sensor that can monitor the body temperature, movement, and stomach acidity of cows. Cows swallow the sensor, which remains in one of their four stomachs permanently. Machine-learning algorithms analyze the data, alerting farmers to the early signs of calving and illnesses. This early detection has enabled farmers to reduce antibiotic usage in their cows by 15 to 30 percent.
Researchers from the University of Surrey, the University of Warwick, and the University of Florence have developed a neural network that can identify congestive heart failure with 100 percent accuracy after analyzing just one electrocardiogram. The researchers trained the network on public datasets of electrocardiograms of people with and without the disease. The network improves upon existing methods for detecting the condition, which can require doctors to take measurements over multiple days.
Intel and NEC are developing facial recognition systems to make it easier to verify the identities of the 300,000 athletes, sponsors, journalists, and volunteers at the 2020 Olympics in Tokyo. NEC will deploy hundreds of facial recognition systems at security checkpoints throughout Olympic facilities, and people using the system will register with government identifications.
Researchers from Taipei Medical University have developed an AI system that can identify a person’s risk of developing non-melanoma skin cancer with an 89 percent probability. The researchers trained the system on medical records from the Taiwan National Health Insurance Research Database. The AI system found that conditions such as hypertension and chronic kidney disease increased the risk of developing skin cancer.
Impira, a U.S. AI start-up, has developed software that uses computer vision and natural language processing to automatically tag and categorize content in photos, videos, and PDFs. The software makes it easier for users to search for visuals in large content libraries and can replace the manual annotation that companies usually have to perform to categorize visuals.
Researchers from multiple German institutions have developed an AI system that can accurately identify the differences between primary lung cancer and cancer that has spread from the head or neck. These cancers can be difficult to differentiate and require different types of treatment. The researchers trained the AI system on tissue sample data of the various tumor types, teaching the system to identify the different types with over 96 percent accuracy.
The Australian Institute of Health and Welfare and the Australian Treasury have developed a housing data dashboard that uses 29 datasets to help researchers, planners, and policymakers better understand issues relating to housing and homelessness in the nation. The datasets combine to include 7 million data points, including data on the number of approvals for new residential dwellings and the number of homeless persons living in boarding houses, tents, and temporarily in a household.
Researchers from MIT have developed a machine learning model that can accurately estimate a person’s risk of dying from a cardiovascular issue within 30, 60, 90, or 365 days of admission to a hospital. The model analyzes electrocardiogram data, and the researchers trained and tested the model on electrocardiogram recordings from 5,000 patients. The model assigns a risk score for each patient, and patients in the highest quartile are seven times more likely to die from a cardiovascular issue than the patients in the lowest risk group.
Image: Asao Tokolo