Calum Cunningham

AI Applications

By Calum Cunningham
Calum Cunningham

Artificial intelligence has numerous practical applications in diagnostic imaging; the key to making them work for clinicians and patients lies in developing and embracing integrated workflow networks.

Of all the healthcare data available, medical imaging is among the richest. In terms of clinical indicators, imaging data contains a tremendous amount of valuable information that’s easily identifiable, as well as so much that’s not visible to the human eye. And because most patient’s healthcare journeys begin with a diagnostic image of some sort, it’s no wonder that U.S. hospitals and health systems spend about $65 billion every year on imaging. These images are often captured at the beginning of the journey, when providers and patients are making many of the most important and consequential decisions.

How can healthcare organizations harness this wealth of data to create better outcomes? The answer increasingly lies within artificial intelligence (AI) and the ways it can be applied to create value for the entire healthcare ecosystem.

The power of AI in diagnostic imaging is frankly underutilized. There are algorithms available to support diagnosis, make image interpretation more efficient, augment clinical decision making, inform procedural interventions and therapies, and even support utilization management and authorizations. The reason these tools are underutilized is because many of these AI models work in siloes; they’re not integrated into the radiology workflow in ways that make them usable or useful.

But that is changing. Today, integrated workflow networks are transforming how radiologists, healthcare providers and other imaging stakeholders can use AI to improve clinical and financial outcomes as well as the radiology experience. What’s particularly valuable about an integrated workflow network is the end-to-end patient care solutions that are emerging.

For example, lung cancer screening programs are now exceedingly common in the U.S., and for good reason. Chronic lung conditions and lung cancers are underdiagnosed. Yet, early diagnosis of these conditions is essential to saving lives, as early intervention through lung health programs contributes to better outcomes and improved quality of life for patients, not to mention better financial returns for healthcare organizations.

These programs can, however, exacerbate radiologists’ already burgeoning workloads, which makes lung programs a perfect application for an AI-powered integrated workflow network. Consider how, until recently, people with late stage/severe emphysema have had relatively few treatment options. Now, endobronchial valves have opened up a new and much less invasive option for this population of patients. AI models are not only extraordinarily helpful in identifying and qualifying patients for the procedure, but they’re also an excellent tool for assessing which region of the lung should be targeted for optimal recovery. Likewise, once the procedure has been performed, patients can return for follow-up imaging and the same AI model can be applied to analyze the patient’s progress and recovery.

We are seeing more and more use cases like this one emerge, including practical applications in neurology, stroke and breast cancer. The challenge is to not treat each use case within its own silo; IT departments simply don’t have the bandwidth to support individual point solutions for lung health, stroke centers, breast density, etc. Even more importantly, no radiologist wants to be—or has time to be—traveling between platforms throughout the day.

Therefore, it is essential to establish a central, scalable, cloud-based AI infrastructure. This infrastructure empowers radiologists, referring providers, payers, medical technology manufacturers and other life science companies to meaningfully collaborate for real solutions that lead to improved clinical and financial outcomes for both provider and patient.

With that architecture in place, healthcare organizations can then adapt these tools to their most pressing challenges—whether that means automating repetitive tasks and alleviating some of the documentation burden from radiologists, sharing images rapidly between providers at disparate sites, layering in AI models that help screen for and detect cancers earlier or, one day, all of the above.

About The Author

Calum Cunningham