Addressing Radiology’s Workforce Crisis: Can a Shift from Being Passive AI Users to Active AI Creators Have an Impact?
Amid the burgeoning imaging volumes and ongoing radiologist shortage, this author suggests that more active involvement by radiologists in the creation and development of AI models could go a long way.
Health care is experiencing a workforce shortage at all levels of experience that touches nearly every therapeutic area, a shortage patients experience whether they are simply trying to find a new primary care doctor or waiting to hear back on potentially life-altering test results. However, while the shortage is felt everywhere, it is arguably most acute in the diagnostic departments that power clinical decision-making.
In radiology, where I have spent the majority of my career, there is a 10-year deficit of more than 20,000 clinicians, driven by a combination of increased demand for imaging and subsequently high rates of burnout in the field.1 As imaging technology evolves, it also expands the volume of images that scanners produce, meaning radiologists must review more than 10 times the data per case than before.2 With more work to do and a shortage of people to do it, radiologists face pressure to work harder and longer, increasing the effects of burnout.
It’s no wonder then that AI developers have promised that this new technology can solve radiology’s burnout issues and shortages. Unfortunately, the narrative often sounds like AI is trying to replace clinicians, something that neither patients nor clinicians want.3 Where AI can actually make meaningful improvements to clinician workflows is in relieving administrative burden, taking care of the repetitive, tedious, and non-diagnostic aspects of the role.4 For radiologists, reclaiming the time lost to these mundane tasks is the single most effective way to stem the tide of burnout and return cognitive energy to the complex diagnoses that AI cannot replicate.
How Report Automation Allows More Time for Image Interpretation
Discerning which imaging AI tools are helpful is no easy task. Part of the problem is that there are so many point solutions on the market today with each one designed to help radiologists detect a certain kind of nodule or abnormality. While they may perform well in isolation, using too many of these tools simultaneously may create frustration for radiologists, who are already under pressure to move quickly through their cases.5 As a profession, we dread AI platforms that use widgets, create pop-ups, and require separate logins because they slow us down and clutter our workspace.
For radiology practices seeking AI tools that relieve administrative burden and reduce burnout, the key is report automation. These tools can take patient measurements, push approved AI-generated insights straight to the report, and turn dictated notes into a cohesive narrative, eliminating the tedious parts of a radiology workflows and giving them more time to focus on the case itself.
Where the AI Triage Coordinator Has an Impact
While automating the report itself saves minutes per case, AI can also provide value before a radiologist even opens a file by intelligently organizing the virtual case stack.6 However, prioritization tools are not a one-size-fits-all solution. Their utility depends entirely on the clinical setting.
In a high-volume outpatient practice or during a busy overnight shift where a single radiologist may be covering multiple facilities, AI-assisted prioritization acts as a critical safety net. By scanning images in the background the moment they are acquired, AI can flag urgent findings and move those cases to the top of the worklist. For the late-night worker, this ensures that life-threatening cases don't sit behind routine scans, triaging the workload when human resources are at their thinnest.
To be truly effective, prioritization tools must be deeply integrated with a practice’s radiology information system and electronic health record. The goal is for AI to function as a silent triage coordinator running in the background, syncing with existing hospital systems, and only intervening when the data suggests a change in priority is medically necessary. When implemented this way, AI actually protects the clinician’s focus.
Empowering Local Innovation with AI
While point solutions and prioritization tools address existing workflows, the next frontier in reducing burnout lies in giving radiology practices the power to build their own tools within a secure environment. Historically, developing a custom AI model required years of coding, massive compute power, and the painstaking accumulation of local data, a luxury few busy practices could afford. This often left clinicians dependent on third-party solutions that don’t always align with their specific patient demographics or clinical preferences.
The emergence of imaging AI foundries changes this dynamic entirely.7 These platforms provide a secure environment equipped with pre-validated foundation models already trained on millions of diverse medical images. With the heavy lifting of initial development already done, practices can bypass the years-long development cycle. Instead of starting from scratch, clinicians can fine-tune these existing models with a smaller, curated set of their own local data to ensure the tool reflects their specific diagnostic standards.
By utilizing an AI foundry, a practice can move beyond the limitations of off-the-shelf software and create bespoke tools, whether it’s a specific measurement automation for a unique surgical group or a specialized screening tool for a local population. This shift from being passive consumers of AI to active creators not only solves specific workflow bottlenecks but also restores a sense of agency to the profession. When radiologists can design the very tools they use, AI stops being a source of secondary frustration and becomes a tailored instrument for professional longevity.
Dr. Siddiqui is the co-founder, chairman, and CEO of HOPPR.
References
- Chen JY, Chen KC, Buch K, Hagiwara M, Lexa FJ. In plain sight, a radiology workforce crisis in the making — gap between job growth and the radiology training pipeline. AJNR Am J Neuroradiol. 2025;46(9):1769-1775.
2. McDonald RJ, Schwartz KM, Eckel LJ, et al. The effects of changes in utilization and technological advancements of cross-sectional imaging on radiologist workload. Acad Radiol. 2015;22(9):1191-1198.
3. O’Reilly KB. More than 80% of physicians use AI professionally: AMA survey. American Medical Association. Available at:
4. Henry TA. Physicians’ greatest use for AI? Cutting administrative burdens. American Medical Association. Available at:
5. Hasan MK. Digital multitasking and hyperactivity: unveiling the hidden costs to brain health. Ann Med Surg (Lond). 2024;86(11):6371-6373.
6. O’Neill TJ, Xi Y, Stehel E, et al. Active reprioritization of the reading worklist using artificial intelligence has a beneficial effect on the turnaround time for interpretation of head CT with intracranial hemorrhage. Radiol Artif Intell. 2020;3(2). Available at:
7. Slavkova KP, Traughber M, Chen O, et al. HOPPR medical-grade platform for medical imaging AI. 2024. ArXiv. Available at:
Related to this article
