About

Built by one person who wanted this to exist when they started.

The author

I'm Pearl Franz. B.S. Neuroscience and M.S. Biomedical Engineering, Carnegie Mellon University. Research in fiber tractography in Dr. Yeh’s lab at the University of Pittsburgh. I process diffusion MRI data for research groups, and the questions I get asked most often are not about the final statistics but about what happened on the way there. This site is the answer I wished I could link to.

How the demos work

Every image on the explainer pages is generated from a single synthetic phantom: a 2D "brain slice" with CSF, gray matter, white matter, two crossing fiber bundles, and realistic relaxation and diffusion parameters. Diffusion signals, k-space, tensor fits, tractography, and group statistics are all computed live in your browser from that phantom, with the same equations real pipelines use, simplified only where the simplification is stated. No real patient data appears anywhere on this site.

The viewer page uses NiiVue to render NIfTI files, and small Web Workers to run a linear tensor fit and a k-means segmentation. Files you load are read by your browser and never uploaded.

What this is not

It is not a processing pipeline you should publish with, and it is not medical advice. The in-browser fits skip the correction steps the explainer spends so much time on, precisely so you can see what they are for. For real work use the tools named at the end of each step.

Corrections and questions

If something is wrong, unclear, or missing, write to pfranz@alumni.cmu.edu. Corrections from people who know a method better than I do are the most useful mail I get.