Radiologist Perspective on AI: Adapting to Change and Staying Afloat

Kelly David Ludema, DO 3/10/2025

Oral and maxillofacial surgeons (OMS) work closely with radiologists nearly every day. How artificial intelligence is used within this medical specialty will affect many others, including OMS. A radiologist working at large health systems in Michigan shares with ACOMS Review readers his daily experience with AI in his practice, as well as his predictions for the future, based on what he has learned so far. 

ACOMS: How do you currently use AI in your daily radiology practice?  

Kelly Ludema, DO (KL): We use several AI-based tools in our practice, primarily leveraging machine learning to aid in report generation and workflow optimization. One widely used program analyzes the radiologists’ prior reports and the current reports’ dictated findings and automatically generates impressions. This reduces mental workload and improves efficiency, increasing productivity by an internally estimated 10-15%. 

Many of us also use large language models (LLMs) such as ChatGPT to assist in generating differential diagnoses and follow-up recommendations, especially for rare cases. I like to call these LLMs “thought partners.” When cross-checked with trusted references, these models can enhance decision-making and broaden a radiologist’s knowledge base. 

Our radiology group is closely involved with Michigan State University and Corewell Health in teaching medical students, residents, and fellows. We are responsible for providing medical education to 24 radiology residents. LLMs are used by some of us to rapidly sort through medical literature for relevant information, create summaries, and even create quizzes for the residents based on their current needs and expectations. This expands the knowledge base of the teaching radiologist and allows rapid assimilation of radiology information that couldn’t be done previously and is so important in the current, ever-changing medical environment. 

AI also progressively contributes to image acquisition and processing, reducing scan times and minimizing radiation exposure — which is especially crucial in pediatric imaging. In acquisition, AI optimizes imaging parameters, reduces noise, and enhances image quality while minimizing radiation dose. It also assists in real-time positioning and automatic segmentation of anatomical structures. In post-processing, AI improves image reconstruction and enhances contrast and resolution. 

Our work-list prioritization system classifies and routes cases based on urgency and subspecialty expertise, ensuring that the right radiologist reads the right case at the right time. 

AI-assisted computer-aided detection (CAD) is also widely used, particularly in mammography, lung nodule detection, virtual CT colonoscopy, and coronary artery calcification detection and risk stratification.  

ACOMS: Can you share an example of how using AI has improved your practice or that of your colleagues? 

KL: The most impactful AI tool in our practice is the machine learning-driven “impression generator.” Traditionally, radiologists dictated reports from scratch, summarizing demographics, study indications, prior reports, findings, impressions, and follow-up recommendations. Now, AI analyzes past reports of the specific radiologist and generates context-aware impressions based on the radiologist’s findings that closely mimic the individual radiologist’s language and style. This not only reduces cognitive load but also improves efficiency and report consistency. 

Additionally, our voice recognition system integrates with AI to: 

   •   Reduce spelling and syntax errors 

   •   Auto-suggest evidence-based recommendations 

   •   Rapidly search medical databases for relevant literature 

This system significantly reduces fatigue, allowing radiologists to focus on interpretation rather than documentation. 

ACOMS: What are the biggest challenges regarding AI programs in radiology? How are these addressed?   

KL: The biggest challenges with AI in radiology include: 

• Cost and reimbursement — AI tools require expensive licensing and ongoing subscription fees. Mostly, insurance does not reimburse AI use directly. The financial benefit comes only if AI increases productivity. Often, hospitals help subsidize costs, but financial sustainability remains an issue.
• Integration issues — There is no universal PACS (Picture Archiving and Communication System), making AI integration difficult across different systems. Custom programming and IT infrastructure are required, often leading to downtime and delays. 
• User adoption — Many radiologists and hospitals hesitate to implement new AI systems due to learning curves and potential workflow disruptions. Overcoming resistance requires dedicated AI working groups, physician training, and clear demonstrations of AI’s benefits. 

As AI continues to evolve, seamless integration with PACS, EMRs, and hospital workflows will be key to widespread adoption. 

ACOMS: What advancements in AI will have the biggest impact on radiology in the near future? 

KL: We are amid an AI revolution in radiology, similar to the industrial and digital revolutions of the past. Over the next five years, we will likely see: 

• Greater use of deep learning for report generation, making AI-generated impressions indistinguishable from human-written ones. 
• Integration with patient EMRs, allowing AI to consider a patient’s entire medical history when interpreting scans, leading to more personalized and precise diagnoses. 
• More advanced large language models trained specifically for radiology, reducing errors, and improving context-aware recommendations.   

The future will bring faster, more accurate, and fully integrated AI-driven radiology systems. 

ACOMS: How do you see the role of AI evolving in radiology over the next 10-15 years? 

KL: Predicting the future is always tricky — many past predictions overestimate short-term progress but underestimate long-term breakthroughs. 

In the next 10-15 years, AI may progress toward artificial general intelligence (AGI), capable of mimicking human cognition but millions of times faster. Some believe this is imminent, while others doubt its feasibility. 

One major technological hurdle is quantum computing. If large-scale quantum systems become viable, it could process radiology images in ways no current computer can, potentially allowing AI to generate entire radiology reports autonomously. However, current quantum computing is far from that level. 

For now, AI will remain an assistive tool rather than a full replacement for radiologists. The human brain’s flexibility, pattern recognition, and ability to interpret complex variations remain unmatched. While AI will continue to evolve and enhance our work, radiologists will still play a central role in medicine. 

ACOMS: What legal challenges do you foresee with AI in radiology? 

KL: There are both ethical and legal challenges surrounding AI in radiology. 

• Patient privacy and data security — AI models require vast amounts of imaging data, but patients do not currently provide explicit consent for their medical images to be used in AI training. Clear regulations and ethical frameworks are needed to balance innovation with privacy rights. 
• Liability and malpractice concerns — If an AI-driven system misdiagnoses a patient, who is legally responsible? The AI developer? The radiologist? The hospital? Current laws do not clearly address AI liability, making this a complex issue for medical-legal teams. 

Until regulations catch up with AI, hospitals and physicians will need clear guidelines on responsibility and risk management. 

ACOMS: Do you have any other thoughts to share? 

KL: We are living through an extraordinary period of technological transformation, particularly in artificial intelligence. AI presents unparalleled opportunities to advance medicine and our world in general but also brings challenges and ethical dilemmas. In the most hopeful scenario, we are witnessing changes that will forever alter the course of humanity and the world for the better. While dystopian outcomes are always a possibility, I remain optimistic about the future. 

As I tell my radiology residents and medical students: 

“The only sustained competitive advantage is the ability and willingness to change.” 

Or as Bob Dylan put it: 

“You better start swimming, or you’ll sink like a stone.” 

Dr. Ludema encourages ACOMS Review readers to reach out with any questions.