Algorithms have proven useful for reading medical images, but a broader application of enterprise AI with medical imaging can improve patient outcomes.
By Kevin Landwehr
AI’s power to help radiologists read medical images has been a central health tech narrative in recent years, but effective adoption has been lagging. Broader deployment of enterprise AI in PACS could unleash its full potential to improve quality of care through medical imaging. AI and machine learning can bridge the gaps between different archive systems, granting radiologists and health systems more control over the flow of medical images, insights, and diagnoses associated with the images. It can also update multiple systems with fewer mishaps that detract from care.
Additionally, the radiology department holds valuable medical information that can inform patient care across the medical spectrum. Yet radiologists often don’t have the time or capacity to review images for all possible indicators of disease or irregularities. Enterprise AI that has already brought efficiency, accuracy, speed, and insights to workflows in other industries can break down the walls between departments, making sure the right images end up in the right hands and presenting them in a way that has immediate utility.
If the first phase of AI adoption in radiology was about functionality, the next phase will focus on connecting those back-end tools with intuitive user interfaces working across various PACS that are configured to the needs of different departments. AI and machine learning, where systems learn from experience to perfect workflows, can then deliver exponential value that will help radiologists detect diseases, including edge cases, sooner, thereby saving lives.
Complex Systems, Simple Tools
Over the past five years, PACS has evolved rapidly within health systems, as radiologists have seen the power of intelligent imaging to vastly improve patient outcomes, helping clinicians see symptoms earlier, monitor patients more closely, and notice unrelated health signals before they become serious.
In the next five years, health image management will become integral to quality of care at the patient level and efforts to create a more efficient health system. The wider PACS and RIS market is expected to be worth $4.17 billion within five years, which would represent around 6.5% annual growth.
Radiology’s expanded role within hospitals has occurred as user interfaces for AI tools and radiological images have improved significantly. This concurrence has created opportunities for adoption beyond medical research centers and high-end hospitals that can afford large investments in software development and personnel. Community hospitals and smaller practices can now also afford these technologies to quickly adapt and build on their capabilities.
Surveys of health professionals show ease of use and utility as the key factors when considering technology upgrades. Yet too many medical “innovations” around AI place an added burden on clinicians and staff, who are already spending less time with patients and more time with computers.
In our work with clinicians, we find an inverse relationship between how much time technology consumes and how likely clinicians are to use it. That doesn’t mean they are resistant to new technology; in fact, the opposite is true. Clinicians are eager to adopt meaningful and impactful solutions that have an obvious utility in diagnosing and treating patients while working to solve wider enterprise problems.
Smart Tech Keeps Learning
Achieving broader AI adoption in hospitals will require designing tools that deliver maximum support with minimal disruption. Additionally, solutions must be ready to use, flexible enough to fit clinicians’ real-world workflow, and immediately beneficial in an already overcrowded tech environment.
The burnout that radiologists are feeling is a result of data overload; the pace of incoming information outpaces a unit’s ability to process and disseminate data. By deploying enterprise AI across the value chain—analyzing images, contextualizing them with other patient records, and routing those insights to the appropriate people—health systems can delegate more work to the technology. This gives radiologists more time and insight into patients’ health and drives meaningful clinical decision making for better outcomes. Even at large hospitals that already have vendor-neutral archives, AI has tremendous potential to route between their systems and rank or triage studies in order of importance, based on certain classifiers or metatagging.
Whatever the size of your institution, enterprise AI must have a broad lens and a fine filter; it must be capable of viewing a health care system’s entire ocean of data but selecting only the information that clinicians need at a specific point in time. That capability doesn’t simply require powerful technology. It requires technology that can learn from its users, predict next-best steps, and adapt to individual preferences—without disrupting care in clinical spaces.
— Kevin Landwehr is a product manager at TeraRecon.