|
Leeds. Early morning.
The imaging suite is quiet at 6 a.m. It always is now.
I used to arrive at this hour to get ahead of the queue. Sixty, seventy scans waiting. Chest X-rays. CT abdomens. The odd brain MRI that made your stomach drop before you even loaded it.
That was the old job. My hands on the mouse. My eyes on the screen. Twenty-five years of training compressed into a split-second feeling… something’s wrong here.
Now I walk in and the screens are already done.
MedSight — that’s what the trust calls the system — reads every scan that comes through Leeds General overnight. By the time I sit down, it’s sorted them. Green means clear. Amber means check. Red means a consultant needs to look. Right now.
I handle the ambers.
Most mornings, there are about thirty. The system is nearly certain they’re fine… but not certain enough. So I open each one. I look. I compare. I confirm or override.
It takes me ninety minutes. The old queue took eleven hours.
Here’s the thing. I should feel relieved. And in some ways I do. The reds get caught faster. The turnaround on a cancer flag used to be seventy-two hours. Now it’s under four. Patients are diagnosed sooner. They start treatment sooner. Some of them are alive because of that gap.
But I’d be lying if I said I don’t miss the craft.
There was something in reading a scan from scratch. Noticing the shadow nobody else saw. The weight of it — you held someone’s future in your interpretation. That feeling is gone. MedSight sees the shadow first. Every time.
My daughter thinks I’m being sentimental. She’s probably right.
She’s a GenPharm coordinator… she manages the AI-designed drugs coming through our regional formulary. Half the medicines on the ward didn’t exist five years ago. They were built by systems like Pharma.AI, tested in simulated trials, approved through the accelerated EU pathway. She doesn’t find any of this strange. She grew up with it.
I didn’t.
I trained at King’s College in 2009. We learned anatomy from cadavers. We read films on lightboxes. My supervisor, Dr. Patil, told me once that a good radiologist reads with his gut before his brain. I’ve never forgotten that.
The machine doesn’t have a gut. But it has six million training images. And it doesn’t get tired at 4 p.m. on a Friday.
The staffing crisis was what tipped it. I watched it happen. After the big NHS restructuring of 2031 — the one that gave AI systems prescriptive authority under clinical governance — we lost about a third of the junior radiologists. They moved into AI oversight, clinical data science, patient navigation. Some left medicine entirely.
But we also gained something.
The GP shortage that had been choking the system for years… it eased. Not because we trained more doctors. Because AI handled the diagnostic bottleneck. Freed up time. Freed up people.
My colleague Sarah used to read mammograms. Hundreds a week. Now she runs the regional breast screening programme. She sees patients face-to-face. She talks to them. She says it’s the first time in twenty years she feels like a doctor instead of a machine.
I think about that a lot.
The old system was breaking. I saw it breaking. We were burning out, leaving early, and the pipeline couldn’t keep up. Something had to give.
I just didn’t expect the change to feel like this. Like watching your own skill become… optional.
Still. I go in every morning. I check the ambers. And some days — maybe one in fifty — I catch something MedSight missed. A hairline fracture behind a rib. A tiny lesion the algorithm marked green.
On those days, I still feel it. The old weight. Maybe that’s enough.
|
