People write about their aha moments with AI. The story where it clicked. The first time they thought, “This is going to change everything.”
This is mine.
I’m a cancer survivor. Stage IV lung cancer. Treatment worked. The tumors disappeared. All metastases resolved.
Now I do maintenance. Regular scans.
I had a CT scan done recently.
In Brazil, you get a written summary about four business days after a scan. Those reports are written for doctors, not patients, but you can get the gist by running them through ChatGPT or Claude.
I don’t have the patience to wait four days.
What I do like about Brazil: the raw images hit your portal the same day.
I set out to read those images myself, using Claude Code.
Medical imaging uses a standard format called DICOM. Think of it like a zip file containing hundreds of raw images and a pile of metadata. A Python library exists to work with these files.
I gave Claude Code my previous scans and the new one, and asked it a simple question:
What changed?
It worked through every prior observation and mapped individual slices to findings. I had metastasis in a couple of vertebrae. It wrote a program to extract the slice ranges for each vertebra in my spine. It went through every slice in my abdomen.
Twenty minutes later, I had a structured comparison.
No new spread. Existing sites smaller or gone. Stable!
Consistent with the findings I received four days later. Consistent with what my oncologist told me.