By Khan Siddiqui

From the outside, my wife’s radiology job looks ideal: short commute, great institution, some flexibility to work from home. The reality? She starts at 6:30 a.m., gets home for dinner, then heads back to the reading room until midnight. Her experience is unfortunately the reality of the profession for thousands in the field right now.

Radiologists like my wife are suffering from severe burnout due to staffing shortages.
Each year, roughly 1,100 new radiology residents enter training. This is a number that has barely budged in a decade, even as imaging volume grows about 7% annually. Demand keeps climbing, but the training pipeline hasn’t expanded to match it. The result is delayed reads, mounting backlogs, and longer turnaround times for essential scans. At some health systems, chest X-rays can take six days or more to be read.

The shortage is compounded by an aging population, longer lifespans, and changing practice patterns, all of which contribute to increased demand for imaging. Workflows across emergency departments and primary care depend on imaging, and many new therapeutics require serial scans for safety monitoring. Radiologists entering the workforce have not been trained to use AI tools that could significantly lighten their load.

We can fix this. In fact, radiology is sitting on one of the biggest opportunities in healthcare: use AI to return valuable time to clinicians and deliver faster results.

The problem isn’t identifying disease, it’s writing about it
Radiologists are extraordinarily fast at perception. Spotting a 1-cm lung nodule can take a quarter of a second, the same amount of time it takes you to recognize a familiar face. The real time sink is converting what we see into a clean, standardized report. That cognitive load, when done hundreds of times a day, is the burnout engine.

Demand keeps rising due to an aging population, imaging-reliant workflows across ED and primary care, and new therapeutics that require serial scans, compounding demand. We won’t hire or train our way out of this. The only scalable lever is efficiency, measured in minutes saved per study and days shaved off turnaround times.
That sounds like a crisis. I see an opportunity: if we cut documentation time from minutes to seconds and auto-handle normal studies, we change the math and give radiologists back the time and focus that burnout has taken away.

The fix: Platform-based, site-tunable AI
Radiology doesn’t need another point solution; it needs a new foundation. The path forward is a platform-based model that radiology teams can train and tune to their own data without starting from scratch. It connects directly into existing PACS or RIS systems, fine-tunes on local cases, and continuously learns from real-world performance.

Artificial Intelligence Homepage