On the front lines of AI: Improving detection and diagnoses

AI solutions are becoming a routine part of clinical care, with nearly two-thirds of physicians now using it for tasks like recording and summarizing patient visits. But the field is rapidly evolving, and it leaves many wondering how to get started and stay up to date.

WellSpan Health is a national leader in using the latest game-changing AI solutions, including technologies that improve the detection of breast cancer and critical conditions, such as stroke. Behind the effort is Heidi Beilis, MD, DABR, FACR, vice president and chief medical officer of the Diagnostic Service Line at WellSpan Health.

Heidi Beilis, MD, DABR, FACR,

In the Q&A below, Dr. Beilis shares how we evaluate new AI solutions, the current and future applications she’s most excited about and tips for starting with AI in your practice.

When did AI begin to make a real difference in your practice?

About five years ago. That’s when AI reached a point where it could help us review imaging data in real time and support clinical decision-making. WellSpan Health was among the early health systems to bring this kind of advanced clinical AI into everyday practice.

Before that, we had used basic automated tools in imaging equipment for many years, but they supported technical tasks rather than diagnosis.

How are AI solutions currently benefiting patients?

Today at WellSpan Health, AI is playing key roles in helping us diagnose critical conditions, detect early-stage breast cancer and communicate with patients.

Diagnosing critical conditions faster 

AI solutions rapidly evaluate CT images to help identify areas of radiographic concern in patients showing immediate clinical signs of stroke, pulmonary embolism, brain hemorrhage, aneurysm or cervical spine fracture. 

In 2024, AI enabled us to diagnose these conditions 900 hours faster across all WellSpan Health sites, which allowed our teams to quickly spring into action, providing time-sensitive treatment and ultimately life-altering care.

Detecting more breast cancers

We recently integrated an AI tool that helps detect 4% more breast cancers — most of which are early-stage and potentially curable.

The solution also accelerates care of the patient in our imaging facilities if a concerning finding is flagged. The radiologist is directed to review those images first, notify patients and rapidly schedule follow-up testing — often before patients leave the initial mammogram visit. 

Streamlining patient communication

WellSpan Health helped create Ana, an AI voice agent that engages patients in personalized conversations to streamline their experience and improve care. For example, Ana calls patients to encourage participation in at-home colorectal cancer screenings (Fecal Immunochemical Test). Surveys show patients appreciate these phone reminders, along with the typical letters and patient portal messages. We’re also using Ana to call patients with negative mammogram results to notify them of their results, reinforce the importance of routine mammograms, and help them schedule their next annual screening.

What are the most exciting AI applications on the horizon?

One of the most promising areas is reducing administrative burden and giving physicians more time to interact with patients. 

Although many of us already use AI to take visit notes, what’s coming next is even more exciting. Soon, AI solutions will be able to synthesize information from prior visits and test results, record the treatment plan and complete documents to order tests, giving us more time to focus on the patient.

Another rapidly developing area will enable improved cancer detection and aid in our ability to diagnose early issues and provide follow-up care. WellSpan Health will soon begin evaluating a tool to enhance lung cancer detection in low-dose CT screening and other CTs. We also plan to explore AI to help identify other findings like liver, pancreatic and kidney lesions.

How do you decide when new AI solutions are ready to use in clinical practice?

We follow a thorough and rigorous metric-driven process to ensure every AI tool we adopt is safe, secure and beneficial for patients and providers. Here’s our process: 

• Monitor FDA approvals: We consider AI tools that get approved by the U.S. Food and Drug Administration.
• Consider clinical benefit: We assess whether the tool could improve patient care or physician workflow.
• Assess data security: Our IT team ensures the tool meets stringent standards for protecting patient data.
• Optimize workflow integration: Our IT team works with the technology’s developers to ensure it fits into WellSpan Health workflows.
• Evaluate performance: We conduct a pilot study at one of our sites to assess real-world performance before systemwide integration.
• Gather feedback: We regularly conduct patient and physician surveys to confirm the tool is effective and well-received.

Many tools don’t make it through this evaluation — and that’s intentional. Failing fast lets us focus on the AI solutions that truly improve personalized and simplified care.

What are some tips for physicians who are interested in using  AI in their practice?

A good starting point is to think about one area that you’d like to improve — something that could streamline provider or staff workflow or make care easier for patients. Then ask: Is there an AI solution that might help? Talk with colleagues to get their thoughts on these solutions.

Try to attend informatics conferences to learn about new solutions and look around other medical conferences — many now have sessions on AI. 

Keep in mind that it can be overwhelming when you first start working with AI. The field is rapidly evolving. Don’t be discouraged or hesitate to ask for help. I frequently collaborate with physicians internally and in other healthcare systems who want to get started.