What We Learned This Year (Medical Imaging Edition)

As this year wraps up, I wanted to share a few quick lessons from the projects we worked on, especially around building web DICOM viewers and integrating AI into imaging workflows.

Nothing too long. Just simple, practical things we kept seeing over and over.

• Zoom-anywhere matters more than we expected

Clinicians don’t want to fight the UI. They want to zoom wherever they’re looking: axial, coronal, sagittal, or 3D without switching modes.

When zoom works everywhere, people move faster. When it doesn’t, everything feels heavier.

• MPR should “feel obvious”

If MPR feels like a feature, it’s wrong. If it feels natural like moving through slices without thinking then it’s right.

Simple crosshair behavior and synced panes go a long way.

• Keep tools modular

Distance, angle, overlays, 3D meshes, every feature should be its own module.
It keeps the viewer stable and makes adding new tools easier.

Clients request custom features, so modularity wins every time.

• AI belongs in the CRM, not the viewer

Running inference inside the viewer slows everything down.
Running inference in the CRM keeps the model out of the UI and makes updates safer.

The viewer should simply show the output, not process it.

• Performance is a feature

Fast viewers lead to faster decisions.
Caching, preloading, lighter meshes, and efficient MPR layouts make a noticeable difference.

Small optimizations add up in clinical workflows.

• Small tools create big impact

Distance measurement, angle measurement, linked zoom, simple overlays.

These aren’t “flashy” features, but teams rely on them every single day.

When they work smoothly, the whole platform feels stronger.

Thanks for being here

That’s it for this year’s final note.

Thanks for following PYCAD, for reading these updates, and for caring about medical imaging technology.
More practical tools, ideas, and breakdowns coming next year.

See you in 2026.