Becker's Hospital Review

110 CIO / HEALTH IT Microsoft rolls out genomics service on Azure cloud By Jessica Kim Cohen M icroso launched genomic pro- cessing services on Azure, the company's cloud computing service, in the U.S., Western Europe and Southeast Asia Feb. 28. e service, dubbed Microso Genomics, comprises a set of cloud-based processing tools for clinicians and scientists pursuing genomics research. e Microso Genom- ics service is part of Healthcare NExT, an initiative the company rolled out in early 2017 to integrate artificial intelligence and cloud computing advancements into the healthcare industry. e Microso Genomics service aims to accelerate genomics research by preparing clean and accurate data. e service utiliz- es processing tools to reduce "artifacts and noise that can cloud the [genomic] data," according to a Feb. 28 Microso blog post. Storing genomic data on the Azure cloud also proves useful for researchers, since a single human genome takes up roughly 100 gigabytes of storage space. "We know we need to have good data and if we can make it very, very easy for people to have good data, then we can bring the biological information to analytical tools in the cloud and, hopefully, make people much more productive and improve their discovery rate," said Bob Davidson, a prin- cipal soware architect in Microso's ge- nomics group. Microso developed Microso Genomics in partnership with Memphis, Tenn.-based St. Jude Children's Research Hospital and DNAnexus, a platform for managing ge- nomic data that runs on Azure. To date, collaborators from St. Jude have stored a half petabyte of genomic data — enough to fill 750,000 standard CD-ROM discs — on Azure for analysis. e Microso Genomics service is covered by Microso under the HIPAA Business Associate Agreement. n UC San Diego researchers apply AI to diagnose eye diseases: 4 things to know By Jessica Kim Cohen A n international team of researchers developed an artificial intelligence tool to screen patients for various eye diseases and facilitate earlier treat- ment interventions. Here are four things to know about the AI tool. 1. The researchers — led by scientists at Shiley Eye Institute at UC San Diego Health and UC San Diego School of Medicine — used a neural network to review 200,000- plus eye scans, in an effort to teach an AI tool to identify macular degeneration and diabetic macular edema, two common causes of irreversible blindness. 2. Following the use of the neural network, the researchers applied a machine learning technique called "transfer learning," which requires a smaller dataset than traditional AI methods to reach conclusions. "For example, an AI neural network optimized to recognize the discrete anatomi- cal structures of the eye, such as the retina, cornea or optic nerve, can more quick- ly and efficiently identify and evaluate them when examining images of a whole eye," according to a Feb. 22 UC San Diego Health statement. 3. Within 30 seconds after analyzing an eye scan, the AI tool would make a di- agnosis and generate a decision on whether or not a patient should be referred for treatment. The researchers found the AI tool's diagnoses matched diagnoses from five ophthalmologists with 95 percent accuracy. 4. The study authors suggested a simplified version of the AI tool would be particularly useful in parts of the world where medical resources, including specialists, are lacking. n Number of health records breached hits 4-year low: 5 things to know By Julie Spitzer T he total number of healthcare records breached in 2017 is at a new, four- year low clocking 4.7 million individuals affected, according to a report from Bitglass, a cloud security provider. The fourth annual Healthcare Breach Report aggregated data from HHS' Office for Civil Rights data breach portal to identify the most common causes of data breaches and how they've changed over the years. Here are five things to know. 1. The amount of healthcare records breached was down 72 percent between 2015 and 2017, and down 95 percent since 2016 — excluding the massive breach- es at Anthem and Premera. 2. The number of data breaches declined slightly, from 328 in 2016 to 294 in 2017. 3. Breaches in 2017 were primarily hacking and IT incidents (70.9 percent), fol- lowed by loss or theft (20 percent) and unauthorized disclosure (8.5 percent). 4. However, organizations are doing a better job mitigating a breach's damage, with each data breach, on average, compromising 16,060 records. 5. The cost per leaked record increased from $269 in 2016 to $380 in 2017. n

• Cover