By Nicholas Galante, MD, and Rish Seth, MD, CIIP
In the mid-2000s, radiology was technologically ahead of the rest of health care, and not just incrementally but structurally, in terms of how the work got done. Entering radiology at that time meant stepping into a specialty that had already solved problems the rest of medicine hadn’t yet acknowledged. The transition from film to digital imaging wasn’t simply an upgrade in tools. It was a fundamental shift in how radiology operated.
When the Baltimore VA Medical Center became the first filmless facility in 1993, it demonstrated that an entire specialty could reorganize around a new technological foundation. By the early 2010s, PACS had been deployed in nearly 90% of hospitals, effectively closing the chapter on physical film.
That transformation required solving interoperability. The adoption of DICOM created a common language for imaging data that allowed systems from different vendors to communicate reliably, years before the rest of health care. Around the same time, radiology embraced voice recognition more completely than any other area of medicine, with most practices migrating from transcription to speech-to-text by the mid-to-late 2000s. Together, these developments made radiology the first digitally native specialty in medicine.
While radiology continued to build on what it had already built. Each new capability, whether advanced visualization, AI-assisted triage, enterprise imaging, or structured reporting, was introduced into an environment designed for an era with lower imaging volumes, less data complexity, and fewer interoperability demands.
The systems weren’t designed wrong. They just weren’t designed for what came next. This is what we think of as digital rot, or the gradual degradation that occurs when new capabilities are layered on an outdated infrastructure, creating a system that becomes more fragmented and more difficult to maintain.
More Capability, More Complexity
It can be difficult to reconcile how a specialty that has adopted so much technology can feel strained. PACS was transformative. It allowed radiologists to read more efficiently and flexibly, removed logistical friction, and fundamentally changed what a radiology department could accomplish. The expectation was that subsequent technologies would deliver similar gains. In theory, the combination of AI, structured reporting, enterprise imaging, and tighter EHR integration should make information more accessible and work more streamlined.
In practice, it’s more complicated. The issue isn’t that individual tools fail in isolation. It’s that each new tool has been introduced into an environment where integration was never fully resolved, making the daily experience of practicing radiology more demanding.
Older systems created inefficiency through slowness and limited capability. Newer systems introduce inefficiency through fragmentation and interruption. The number of applications involved in a typical read has grown, along with the number of decisions required outside of image interpretation. The cognitive overhead required to navigate between systems, reconcile information, and manage a workflow that was assembled rather than designed has grown, too.
Inside a radiology department, this complexity shows in constant toggling between imaging and reporting, repetitive navigation through templates, copying and pasting of information that should move automatically, and redundancy built into systems that don’t communicate effectively. Over the course of a workday, these tasks add up to a substantial and largely invisible burden, with burnout rates between 45% and 60%.
Radiologists are also often not the primary decision-makers when selecting these tools. Decisions may be driven by administrative priorities or financial considerations that aren’t aligned with how radiologists actually work. When tools are introduced this way, they tend to persist even when they don’t deliver their intended value. Institutional knowledge starts becoming concentrated in a small number of individuals, creating bottlenecks and risk.
Meanwhile, the clinical reality continues to shift. Imaging volumes are rising. Data complexity is increasing. Turnaround time expectations have tightened. The workforce isn’t growing at a rate proportionate to demand, which makes efficiency a structural requirement.
At the same time, radiology is generating more data than ever before, much of it from imaging, yet that data remains fragmented across systems and institutions. It is often locked within legacy infrastructure, difficult to access, and disconnected from the workflows where it could be most useful. The issue is not a lack of information. It is that the information is not organized or integrated in a way that allows radiologists to fully use it. It’s artificial scarcity in a system defined by abundance.
Develop Organizational Support
Core systems are being updated, replaced, or sunset by vendors operating on their own timelines. This introduces new workflows, requires restraint, and creates periods of instability in environments that rely on consistency. In other industries, this disruption is manageable. In radiology, it affects diagnostic throughput and patient care.
At the same time, AI is proliferating rapidly across the specialty but without a coherent integration framework. Many tools perform well on the tasks they were designed to address, but they’re being deployed into environments that weren’t built to absorb them cleanly. As a result, AI isn’t consistently reducing workload. In many cases, it’s adding steps and decisions, which increase cognitive load rather than relieve it.
The trajectory is clear. The gap between demand and capacity will continue to widen. Cognitive load will continue to increase. Burnout will remain persistent. Delays in interpretation will affect downstream care in ways that are already visible.
Radiology functions as the core diagnostic infrastructure for modern medicine. When that system slows or fragments, the impact extends well beyond the radiology department. Radiologists, clinicians, and patients are feeling this now more than ever.
Radiology has historically been a leading indicator for how technology enters clinical practice. That remains true, but the lesson is no longer about the adoption of technology, but rather adoption of what comes after.
No specialty in medicine has absorbed more technological change, more rapidly, and continued to deliver. Filmless reading, voice recognition, AI-assisted triage, and remote interpretation are transitions that radiology did not resist, it led them. The specialty that digitized first, standardized first, and adopted AI first is also the specialty best positioned to show the rest of medicine what getting this right actually looks like.
The next phase requires a different focus. The question is no longer whether new tools can be added. It’s whether the environment can be organized into something that supports how radiologists actually work.
The workflow friction, the fragmentation, the cognitive overhead are not signs of a specialty in decline. They are the predictable growing pains of an industry that moved fast and is now ready to mature. The foundation was right. The standardization was right. What is needed now is the same clarity of purpose applied to the integration of workflows, AI tools, and clinical systems into a coherent experience for the radiologist at the center of it.
— Nicholas Galante, MD, is the medical director for informatics at Radiology Associates of North Texas.
— Rish Seth, MD, CIIP, is the chief medical innovation officer for Rad AI.