This week’s list of data news highlights covers February 9-15, 2019, and includes articles about an AI system that can lipread and facial recognition catching an imposter at the JFK airport.
Chinese and American researchers have created an AI system that accurately diagnoses common childhood conditions. The system diagnosed asthma with 90 percent accuracy and gastrointestinal disease with 87 percent accuracy. The researchers trained the system on the electronic health records of 600,000 Chinese patients, and the system uses data on a patient’s symptoms, medical history, lab results, and other clinical data to make diagnoses.
The U.S. Department of Health and Human Services (HHS) has proposed new rules to allow patients to have easier access to their health data. One of the new regulations would require that healthcare providers make all information in a healthcare provider’s electronic record, including doctors’ notes, freely available to patients. The rule would also require that providers adopt standardized application programming interfaces, which would allow patients to access their health data on mobile devices such as their phone.
Researchers from Chinese and U.S. universities have developed LipPass, an authentication system that uses deep learning to accurately identify mobile phone users by their mouth movements. LipPass authenticates users by analyzing their lip protrusion, tongue movement, and jaw angles when they speak. The system achieved 90.2 percent accuracy for authenticating legitimate users and identified spoofers with 93.1 percent accuracy.
A group of researchers, including individuals from the University of Georgia and the U.S. Food and Drug Administration, have created a machine learning model that accurately detects the animal sources of salmonella. The researchers trained the model on 1,300 salmonella genomes that had known animal sources. The model predicts the source of salmonella with up to 92 percent accuracy depending on its level of certainty.
Researchers from universities in the United States, Russia, and Singapore have used AI to accurately predict how altering a material’s shape will affect its properties, such as how it conducts electricity, transmits light, or conducts heat. Changing the structure of a material can improve the speed that it allows electrons to move through it, which can enable devices using semiconductors to perform faster.
Researchers from the University of Cambridge and from Pfizer have developed a machine learning model that has identified four molecules linked to Alzheimer’s disease and schizophrenia. The model separated relevant and irrelevant chemical patterns and helped the researchers determine the four relevant molecules out of a set of six million molecules.
Otsu, a city in Japan, is developing an AI model to predict how suspected cases of bullying might escalate. The city is training its model on data from 9,000 suspected bullying cases between 2012 and 2018. The training data includes the ages, genders, absenteeism records, and academic achievements of the students involved in the alleged cases.
Researchers from Michigan State University and the University of Michigan have developed a machine learning model that predicts which plant genes make metabolites, which are substances that attract and repel other organisms such as pests. The researchers incorporated over 10,000 features of genes, such as the gene family size, in the model. Other researchers could use this model to identify the necessary genes for drugs, disease-resistant crops, and artificial flavors.
French researchers have created AntBot, an autonomous six-legged robot that can navigate without using GPS. Instead of GPS, the robot uses sensors to gather UV light data, which allows the robot to navigate using a “celestial compass.” The robot also counts its steps to know how far it has traveled.
U.S. Customs and Border Patrol used facial recognition technology to catch a 26-year-old woman using a passport that was not hers at the John. F. Kennedy airport in New York. Since the introduction of facial recognition technology at land ports of entry in the fall of 2018, the technology has helped U.S. Customs and Border Patrol identify 55 people attempting to enter the United States with another individual’s travel documents.