The healthcare industry is already using artificial intelligence (AI) applications to automate many tasks such as simplifying and cutting down on administrative errors or automating claims processing. However, many new AI technologies are focused on clinical applications. Radiologists can find some of the most current AI initiatives in areas such as population health screening and analytics, post-acute triage for urgent conditions and traditional peer review and quality assurance.
The availability and adoption of these types of AI applications has the potential to improve patient care and significantly impact patient health outcomes. Because of this potential, many different entities have entered the market to pursue clinical AI applications, including start-up companies, healthcare industry leaders, and some healthcare systems, making it extremely complicated for providers to find, select, and evaluate the AI applications that will best fit a healthcare facility’s needs. A recent webinar hosted by GE Healthcare presented some of the challenges that exist within the current landscape of AI application offerings, and how providers might change their approach to finding what they need to incorporate into their radiology workflow.
Building AI applications in radiology
The webinar, “Artificial Intelligence in Radiology, from Hype to Adoption—Intelligently Efficient”, featured Karley Yoder, Vice President and General Manager of AI at GE Healthcare, who shared that 80 percent of healthcare leaders see growth or savings from using AI and that in 20191, there were more than 350 start-up companies working on AI applications in healthcare.2 While Yoder stressed that the potential in all of these AI applications is exciting, the question healthcare providers need to be asking is how to find and deploy the right AI solution from among the tidal wave of available options without crushing their teams.
AI developers are making strides in areas such as triage, decision support, assisted reading and structured reporting. One example highlighted in the webinar was ClearReadTM CT, powered by Riverain Technologies. This AI solution is a concurrent-read AI product that automatically detects and measures properties of solid, part-solid, and ground-glass nodules by leveraging patented vessel suppression technology. Typically, low-dose CT scans can be noisy and can obscure subtle, relevant findings. Overlapping tissue and vessels can hide nodules that would otherwise be visible. ClearRead CT has offered radiology promising results, such as improvements in reading time, improvements in nodule detection and automating lung nodule measurements.
Other examples include INSIGHT CXR, powered by Lunit, which can automatically detect nine major radiologic findings in x-ray with a high degree of accuracy. It has also proven effective for patient monitoring in COVID-19 settings using enhanced sensitivity for lung consolidation/ground glass opacity, suggestive of COVID-19 infected pneumonia. Accipio Ix, powered by MaxQ AI, is a triage application that provides automatic, rapid, and highly accurate detection of suspected intracerebral hemorrhage.
Revising the current AI acquisition approach
Yoder explained the long and often exhaustive acquisition process that providers typically experience when they are in search of new AI solutions for their institutions.
“Despite the incredibly powerful use cases and applications coming to market and their potential to drive real outcomes in the healthcare space,” Yoder explained, “today’s acquisition approach—with all the vendors and start-ups—simply will not scale.”
Shifting to an AI ecosystem mindset
Yoder went on to explain the need for the healthcare industry to shift to an ecosystem mindset and process, where providers can easily find the applications they need, easily try those applications, and easily buy those applications.
“It needs to be the same type of e-commerce experience that we’re used to in our lives everywhere else outside of healthcare. But we know there’s a huge infrastructure of work behind those three steps, and outside of those three steps that supports it so that providers or institutions can really execute a find, try, buy approach to leveraging this ecosystem of AI that is really coming to life.”
GE Healthcare’s Edison™ Ecosystem was built on the find, try, buy premise. It is a new way to acquire healthcare applications that can scale and evolve with a provider’s needs. Edison was created to develop, house, test and deploy AI applications. Embedded within existing workflows, Edison applications can integrate and assimilate data from disparate sources, and apply analytics or advanced algorithms to generate clinical, operational, and financial insights for providers and institutions.
Partnership program to benefit all
As an industry leader, GE Healthcare has a proven track record in generating scalable solutions and efficiencies, so when it came to Edison Ecosystem, it took the concept a giant step further. One of the most unique aspects about Edison Ecosystem is the Edison Developer Program to support, enable and deploy rapid innovation.
The Edison Developer Program was designed to connect the best innovations in AI with healthcare providers in an intelligently efficient, and scalable way. Developers who are interested in bringing their innovations onto the Edison Ecosystem go through a five-step process that starts with identification of the opportunity and leads the developer through to product distribution on the Edison Marketplace.
Yoder explained how the program works. Each developer works with the GE Healthcare AI team to conduct an analysis and clinical evidence review to determine the specifics of the product and how it would fit into current healthcare workflows. The team works together to define the best approach for integrating the product across the Edison platform and for meeting the regulatory, and quality requirements needed for all GE Healthcare offerings. There would be a beta testing at select locations to ensure the product is ready for a full-scale launch, and finally, the application would be launched across the entire Edison ecosystem with full support from GE Healthcare’s sales and marketing teams.
The Edison Marketplace is currently providing more than 100 AI applications, developed by GE Healthcare as well as many third-party developers.
Seamless AI integration for intelligent efficiency
Overall, AI is an exciting area for radiology, and it is recognized as a tool that can help radiologists perform their jobs more efficiently, more accurately, and with time savings, but most importantly, without slowing them down. The true test of an AI application is that it is seamlessly and invisibly integrated into the user workflow with little additional training or workflow time added to the radiologists already demanding schedule. Providers considering any new AI application, across any process, should also consider what additional training will be needed to incorporate or adopt the AI into their current workflows.
With the evolution and continued adoption of AI technology and clinical applications in radiology, providers can be confident they will be able to improve care delivery as well as impact patient care by improving or assisting in the accuracy of diagnosis.
Learn More:
Not all products or features are available in all geographies. Check with your local GE Healthcare representative for availability in your country.
REFERENCES
- MIT Tech Review Insights – October 2019
- Healthcare Dive – January 2020
