Improving Rural Health Outcomes with AI: From Data to Diagnosis

Q: Can you tell us a bit about yourself and your role at Emory?
A: I’m an associate professor of radiology at Emory and director of the Healthcare AI Innovation, Translation, and Informatics Lab — what we call the HIT Lab. Our work focuses on four main areas: building high-quality datasets for AI, evaluating how AI performs in real-world settings, examining bias and fairness in AI models, and improving generalizability across populations. I'm also passionate about training the next generation of data scientists and clinicians to bridge healthcare and technology.
Q: Tell us about the HIT Lab’s work in tackling rural health disparities.
A: A big part of our work is opportunistic screening, which uses existing imaging like mammograms to detect other health risks, such as heart disease in women. Many women, especially in rural areas, get mammograms for breast cancer screening. But the same images can be used to assess cardiac risk, which is often underdiagnosed in women due to atypical symptoms. We’ve developed an algorithm that uses mammogram data to flag potential cardiac issues. Although the algorithm is not yet in clinical trials, it is showing strong results in external validation.
We’re also studying how well AI models perform in real-world clinical environments. At Emory and Grady Memorial Hospital, we’re using algorithms that triage strokes, brain bleeds and pulmonary embolisms. We’ve found that these algorithms don’t always perform as well for outpatients as they do for emergency or in-hospital patients — and that's important, because many rural patients are outpatients who travel far for care.
Another area we're exploring is making medical records easier for patients to understand using large language models, like ChatGPT. This approach could help non-English speakers or those unfamiliar with complex medical terminology engage more with their health data.
Q: Why is the narrative behind the data so important when applying AI in rural health?
A: Data isn’t just numbers — it’s people's lives. If AI is trained primarily on data from urban academic centers, it won’t generalize well to rural settings where disease presentation, equipment and care pathways are different. AI models don’t know how to say “I don’t know.”They just give you a prediction, which can be dangerous if the input data doesn't reflect the rural reality.
Also, storytelling helps surface why an algorithm might fail. Maybe it’s not working well because the rural hospital has older equipment or because a patient's symptoms don’t match what the AI was trained on. Understanding those contexts is crucial for designing systems that actually help — not harm — rural populations.
Q: How can AI better support rural healthcare providers and reduce diagnostic delays?
A: AI has the potential to upskill rural healthcare providers, especially in places where specialists are scarce. A reliable AI assistant can help nurse practitioners or general physicians make more confident, informed decisions — even when they’re working outside their typical comfort zones.
Another benefit is the indirect improvement of healthcare systems. When we created a breast cancer dataset, we discovered errors in existing cancer registries. Fixing those issues improved not just the AI model but the healthcare system as a whole.
Lastly, AI tools like summarization algorithms can help providers quickly understand why a patient was transferred, especially when dealing with complex or lengthy records. This improves communication and coordination between rural clinics and referral centers like Emory.
Q: From your perspective, what are the most pressing healthcare challenges facing rural Georgia today?
A: First, there's the challenge of distance. Many patients have to drive two or more hours to see a specialist. Second, we have a major workforce shortage. There’s been a shift toward more advanced practice providers (APPs) like nurse practitioners and physician assistants filling gaps, but there’s variability in training and scope. APPs are great at specific tasks and care coordination, but more complex diagnostic work can be harder without specialized training.
Q: And what do you see as the biggest opportunity for AI to create lasting change in rural health?
A: The biggest opportunity is in population health. AI can help large health systems like Emory optimize their resources by focusing on the patients who need them most — for example, ensuring we prioritize cancer patients over low-acuity cases. For rural clinics, AI can function as a triage tool to identify which patients should be referred to tertiary centers.
It also holds promise in augmenting care, helping rural providers deliver more accurate diagnoses and manage patients effectively, even in resource-limited settings. And if we can leverage AI to engage patients earlier and more meaningfully, we’ll be able to prevent more diseases before they require high-intensity care.