Making Radiology More Understandable With AI
In a study of 2,000 U.S. adults reviewing sample radiology reports, comprehension accuracy increased from 58% with traditional reports to 68% with AI-generated plain language versions. As patients gain more immediate access to imaging results, providing information is no longer enough. Healthcare organizations also need to consider how patients will understand that information, what questions it may raise, and how technology can help connect complex findings to meaningful next steps.
In this interview, Dr. Nina Kottler, Chief Medical AI Officer at Mosaic Clinical Technologies, shares her perspective on how AI is transforming radiology and patient knowledge. She explores how healthcare organizations can use AI to support radiologists and help patients better understand their imaging results while maintaining clinical oversight and trust.
1. Patients now often see their imaging reports in the portal before their doctor has walked them through the results, and those reports were written for clinicians. What does that shift mean for the role of patient education in healthcare, and who should ultimately be responsible for making complex medical information understandable?
Technology is enabling patients to become greater owners of their own health journey. Wearables, for example, give people immediate feedback and allow them to see how changes in sleep, exercise, diet, or other behaviors may affect their health data. Immediate access to imaging reports is part of that same shift toward greater transparency and patient participation.
But access to information is not the same as understanding it. Radiology reports are written for clinicians and often contain specialized terminology, nuanced uncertainty, and incidental findings whose significance depends on the patient’s broader clinical picture. Some patients who view their reports may turn to a general-purpose AI chatbot for a translation. While understandable, these tools may lack the clinical context and safeguards needed to explain the findings accurately. Consumer-facing applications may also lack appropriate protections for health information, creating privacy concerns when patients enter their reports or other personal medical information.
To address this gap, some healthcare systems provide secure, reliable, patient-friendly explanations alongside the clinical report. The goal is to help patients understand what the findings may mean, what remains uncertain, and what should happen next, while preserving important clinical nuance rather than oversimplifying the information.
Responsibility for making complex medical information understandable is shared. Radiologists are responsible for communicating imaging findings clearly and accurately, treating clinicians provide the broader clinical context, and healthcare organizations and technology developers should create trustworthy ways for patients to understand and act on their information. As we give patients greater access to medical information, we should also give them the tools to meaningfully use that access.
2. You’ve talked about AI making radiologists “better, not busier.” If AI can give radiologists more time, where do you think that time can have the greatest impact on the patient experience, and what role should communication play in that?
Right now, many radiologists are working through growing worklists under tremendous time pressure. If AI can create additional capacity, the first benefit is practical: helping us keep up with demand, get answers to patients and their physicians sooner, and reduce the relentless workload that contributes to burnout.
But I hope we do more than use AI to work more efficiently. I think we can evolve from #EfficiencyAI to #OpportunityAI. Efficiency AI helps us perform the work we already do more effectively. Opportunity AI could expand what imaging allows us to do for each patient.
By incorporating more clinical information into AI models, we can begin moving from population-based care toward precision medicine, including predicting and potentially preventing disease and determining which incidental findings require attention for each individual patient. More broadly, #OpportunityAI could enable less invasive approaches to diagnosis and treatment and extend subspecialty-level care to patients regardless of where they live. These opportunities are not simply aspirational. We are already seeing early evidence of several of them, although much more clinical validation and thoughtful implementation will be needed before they become routine care.
Communication will remain essential, but that does not necessarily mean more conversations. Technology should enable more timely and effective ways to get the right information to the right clinician and patient, with the appropriate context and a clear next step.
3. Imaging increasingly uncovers incidental findings that may be harmless but can still create significant anxiety for patients. What does this reveal about the difference between giving patients more information and actually giving them better health education?
In our work with AI models, we are seeing enhanced detection of findings, including some that are incidental. But “incidental” means a finding was discovered on an exam performed for another reason. It does not mean the finding is unimportant. Some are harmless, while others can provide an opportunity to identify significant disease earlier.
As our ability to detect findings improves, our responsibility to explain their significance grows. For a patient, an unfamiliar term in a report can feel like a diagnosis. Better health education helps answer the questions behind that anxiety: What is this? How concerned should I be? Does it explain my symptoms? Do I need to do anything, and how soon?
That requires translating medical terminology into plain language and placing findings in the context of the individual patient, explaining what is known and uncertain, and providing a clear next step. When no follow-up is needed, explaining why can be just as valuable as explaining why another test is recommended. When follow-up is needed, patients should understand its purpose, timing, and who will coordinate it.
The goal is to help patients respond appropriately, without unnecessary alarm or false reassurance. Enhanced detection creates an opportunity to improve care, but realizing that opportunity requires clinical context, understandable communication, and a clear path from finding to appropriate action.
4. As AI becomes more involved in the care journey, from triaging worklists to drafting findings, patients will have to navigate a healthcare experience shaped by technology they may not fully understand. What does it take to build trust when patients don’t necessarily know how a technology was involved in their care?
Patients do not need to understand the technical details of every tool involved in their care, but they should be able to trust that someone with the appropriate expertise is responsible for how it is used. That trust must be earned through clinical evidence, thoughtful implementation, and ongoing oversight, including taking action when concerns arise.
For me, this commitment means evaluating AI before deployment, preparing clinicians to use it appropriately, and monitoring both its performance and how it affects clinical decisions after deployment. A model may perform well in testing, but we still need to know whether it is helping clinicians provide better care in practice. Monitoring alone, however, is not enough. We need to act on what we learn, addressing concerns and making changes when needed. Action is what turns monitoring into governance.
An essential part of this oversight is having the AI results interpreted by a qualified end user. The user needs the clinical expertise and training to evaluate the AI’s outputs, understand its limitations, recognize when it may be wrong, and determine the appropriate next step.
Transparency is needed by the person consuming the AI results. In radiology, that is usually the radiologist. How results are presented, including the information available to assess them, can influence whether the radiologist appropriately accepts or challenges an output. Transparency should support informed clinical judgment and help mitigate human-AI bias, including over- and under-reliance on AI. When AI-generated explanations are provided to patients, they should preserve clinical accuracy and uncertainty, undergo appropriate clinical review, and clearly identify when patients should seek guidance from their treating clinician.
Ultimately, I want patients to know that we are using technology thoughtfully to improve their care, that qualified clinicians remain engaged, and that we have a process to identify and address problems.
5. We often think about healthcare content as something patients consume after a diagnosis or procedure. What would change if we instead treated education and content as an integral part of the care journey, from preparing for an exam, to understanding results, to knowing what to do next?
Education is an essential part of care, and in radiology that includes both patients and the clinicians caring for them. Radiologists serve as the “doctor’s doctor,” helping referring physicians select the appropriate imaging exam, understand what the findings mean, and determine how those findings should inform care.
For patients, education begins before the exam or procedure: Why is it being performed? How should I prepare? What should I expect? Afterward, it means explaining the results in context, including what remains uncertain. When a next step is recommended, patients should understand why it matters, when it should happen, and who will coordinate it.
For referring clinicians, our role goes beyond delivering a report. We help connect imaging findings to the clinical question, clarify the significance of unexpected findings, and guide appropriate next steps. As imaging and AI provide more information, that consultative role becomes even more valuable.
AI makes education even more essential while creating new ways to deliver it. Appropriate use requires an understanding of AI in general and the capabilities, limitations, and intended use of the specific tool. AI can also support education by translating complex clinical information into explanations tailored to patients and referring clinicians, from patient-friendly imaging reports to video explanations based on physician content. With appropriate review, these tools can help bridge the gap between receiving medical information and understanding what it means.
With appropriate clinical oversight, AI can help patients better understand their imaging results and how those results relate to their care. Clear explanations can help patients ask informed questions, participate in decisions with their treating clinicians, and understand recommended next steps.
About Nina Kottler
Dr. Nina Kottler is the Chief Medical AI Officer at Mosaic Clinical Technologies, where she leads the vision and deployment of artificial intelligence at scale across Radiology Partners, the largest radiology practice in the world. A practicing radiologist for more than 20 years with graduate training in applied mathematics and optimization theory, she has focused her career on harnessing innovation to improve quality and value in medical imaging. As Radiology Partners’ first radiologist, Dr. Kottler has held multiple leadership roles spanning clinical operations, remote imaging, AI governance, and organizational culture. Externally, she is an Associate Fellow at the Stanford AIMI Center and serves on committees for the ACR, RSNA, and SIIM. She is also a frequent international speaker and industry consultant. Dr. Kottler is passionate about advancing diversity and fostering a culture of belonging. She serves as Chair of RadEqual and as a member of SIIM’s Belonging, Engagement, and Outreach (BEO) Committee. In 2025, she received the Dr. Ruth Dayhoff Award for advancing women in medical imaging informatics, and in both 2025 and 2026 was recognized as a leading voice in radiology AI.
About CARAVAN
CARAVAN Wellness is the #1 global leader in expert content solutions for health, wellness, workplace, financial care, and education. CARAVAN impacts millions of people by helping healthcare organizations, health plans, employers, and digital platforms educate, engage, and empower people with trusted expert content. At CARAVAN, we envision a world where everyone can access trusted education and resources, regardless of their age, income, or location. Learn more at caravanwellness.com
