AI Takes Radiology from Diagnosis to Prevention

From the Applied Radiology booth at RSNA, one theme surfaced repeatedly across conversations: artificial intelligence is no longer just about reading images faster, it is reshaping how radiology contributes to patient care, access, and population health. That message came through clearly during a live discussion between Kieran Anderson, Group Publisher at Applied Radiology, and Suzie Bash, MD, neuroradiologist and Medical Director at RadNet.

Early in the conversation, Bash described how radiology’s role is evolving from a reactive diagnostic service to a central driver of earlier, biomarker-based disease detection. “We’re seeing a transition from purely clinically based diagnosis to imaging-based diagnoses,” she explained, pointing to Alzheimer’s disease as a prime example. Once diagnosed clinically, Alzheimer’s is now increasingly identified using imaging biomarkers, including amyloid and tau PET and quantitative MRI. “Amyloid is present at the very earliest stage, even before the patient is symptomatic,” Bash noted, underscoring how imaging enables earlier intervention.

AI, she emphasized, is accelerating this shift by enabling meaningful “stage shifting” across multiple disease states. Bash cited RadNet DeepHealth’s breast cancer screening tool, noting that it detects cancer “21% more than just a human alone, and in fact, one to two years earlier.” With RadNet performing roughly two million mammograms annually, she stressed that earlier detection at this scale can significantly impact morbidity and mortality. Similar gains are being seen in lung and prostate cancer, as well as cardiovascular imaging, where AI-assisted coronary artery calcium analysis can change patient management decisions about half the time.

When Anderson asked which technologies are having the greatest enterprise-wide impact, Bash pointed to a less-discussed but pressing issue: workforce constraints. “Our imaging volumes are going up at such a rate that we’re exceeding the technologist workforce right now,” she said, citing dramatic growth in amyloid PET, MRI brain, breast MRI, and PET/CT volumes. To address this imbalance, Bash highlighted RadNet’s “Tech Live” program, which allows expert technologists to remotely assist or operate scanners across modalities. “This type of technology has resulted in a 42% improvement in MR room closures,” she said, calling it a major step toward improving access to care.

AI-driven deep learning reconstruction is also helping practices cope with rising demand. Bash explained that faster scanning, up to 50% quicker with improved image quality, allows centers to image three to four additional patients per day without extending technologist hours. “It prevents having to find people that will work overtime,” she added.

Access and equity were another focus of the discussion. Bash described creative initiatives such as a mobile, AI-enabled mammography coach that brings screening directly into underserved communities. “A lot of these patients have never had a mammogram before,” she said, noting that thousands of studies have already been performed through the program. Locating mammography units in nontraditional settings, such as large retail locations, has also helped reach first-time screening patients.

As RSNA wrapped up, Bash reflected on the meeting’s theme, Imaging the Individual. “What I’m really seeing is much more personalized, patient-centric imaging initiatives,” she said, adding that this focus was evident across the exhibit floor. From the Applied Radiology booth, the takeaway was clear: AI is helping radiology extend its influence, earlier in disease, deeper into communities, and across the entire imaging enterprise.