New rules and requirements add to the complexity of healthcare – something that artificial intelligence can address, according to Vignesh Shetty, senior vice president and general manager of Edison AI and Platform for GE Healthcare (GEHC).
Healthcare IT News asked Shetty about the progress that has been made on the 360-patient view since he last spoke with us about the exciting advances in data-driven insights, as well as how algorithms achieve challenging data exchanges.
Q. In terms of the 360-patient view, what progress has been made toward improving health outcomes?
A. From MRI scanners used by doctors to detect tumors to mobile x-ray units in the ER or ICU used to image the lungs of COVID patients, doctors and patients are benefiting from artificial intelligence embedded in medical devices.
The goal is to have AI tools ready when and where they’re needed to contribute to faster diagnoses and, ideally, better patient outcomes.
One example of the tangible impact of AI in healthcare is GEHC’s Critical Care Suite, a collection of industry-first AI algorithms embedded in a mobile X-ray device. Built in collaboration with the University of California San Francisco (UCSF) using GEHC’s Edison platform, the AI algorithms help radiologists prioritize and flag a suspected pneumothorax, a common, but sometimes life-threatening, type of collapsed lung, to help reduce the turn-around time.
A recent review at University Hospital Cleveland showed that while using Critical Care Suite, there was a 78% decrease in reporting time – from 3 hours and 22 minutes down to 44 minutes – for urgent exams positive for pneumothorax.
Another innovation among GEHC’s AI-enabled devices is AIRTM Recon DL technology, a pioneering, deep-learning-based image reconstruction algorithm for MRI exams that can decrease scan time by up to 50% without sacrificing image quality.
Chicago area-based Duly Health and Care, which uses the technology, reduced wait times allowing them to see more patients sooner.
One of the doctors said the reduction in time on the table they experienced – under 10 minutes as opposed to up to 45 minutes – is a significant improvement for patients, especially those who suffer back pain.
Q. Interoperability has lagged because data standards have not been strictly defined. How is this changing?
A. We believe that the Cures Act adoption, which enforces FHIR representation of healthcare data as a model and interoperability supported through loose coupling by way of API representation, will help drive interoperability across traditional EHR systems and new-age innovative workloads that will run on top of traditional platforms.
These supplement international and national initiatives with similar objectives. For example, the United States Office of the National Coordinator for Health Information Technology (ONC) has released several interoperability standards and guidelines for the healthcare industry, which should significantly help alleviate current challenges.
Q. With no standard for genomics data, how can AI help tackle data exchange beyond a 1:1 negotiation for these rich data sets?
A. AI algorithms can be used to process and analyze genomics data and extract relevant information, such as genetic variants or mutations, which can be organized and formatted in a standard way.
AI accelerates the goal of precision medicine by building predictive models or other analytical tools that can help to make sense of genomics data and extract insights.
Apart from traditional standards that have been prevalent in the genomics world, HL7 working groups have put together a standardized representation of genomic data through five resources for FHIR version 4.
FHIR V5, which is being actively developed, has an even better representation of genomic resources, so we expect this trend to accelerate.
Q. With info blocking rules in effect, how can AI help accomplish increased data sharing and interoperability with demands for data privacy?
A. With the enforcement of the info blocking rule, we notice that the saliency of insights becomes the main source of value. Investments that would appear unnecessary in the context of data scarcity become far-sighted in the context of abundance.
For example, as more data flows through the automation of APIs, it becomes humanly impossible to identify leakages across data pipelines.
In the case of structured data, machine learning algorithms can identify protected health information features to stay compliant from a privacy standpoint. For unstructured data, we can use a combination of deep learning algorithms and natural language processing to structure them and address interoperability.
Andrea Fox is senior editor of Healthcare IT News.
Email: [email protected]
Healthcare IT News is a HIMSS publication.
Source: Read Full Article