Neurograsp

Neuroscience visualizations are stuck behind clunky desktop tooling. Researchers and students orbit cortical models with click-and-drag UIs built for CAD software in the 1990s — slow, unintuitive, and impossible to use while talking through anatomy in front of a class.

Inspired by MediaPipe's hand-landmark work and the FreeSurfer fsaverage5 template, we built a browser-native explorer: point a webcam at your hands, and natural gestures drive a real cortical surface mesh with cascade-wave activations rippling across it.

The tools neuroscientists use to show brains — Freeview, MRIcron, Slicer — render beautifully but interact terribly. Every demo I've watched ends with the presenter fumbling a trackpad while trying to point at the temporal lobe.

And the gesture-control work that does exist depends on Leap Motion or Kinect hardware nobody has on their desk.

Hand tracking runs entirely in the browser via Google's MediaPipe — 21 landmarks per hand at ~30 fps, no server round-trip, no data leaving the machine. A small gesture state machine maps pinch-drag to rotation, two-hand pinch distance to zoom, and a sustained two-hand spread to a hemisphere explode.

The brain itself is the fsaverage5 cortical surface from FreeSurfer (Harvard/MIT), rendered with Three.js + React Three Fiber. Neural activations are simulated cascade waves that propagate across the mesh in real time.

Working demo running locally in Chrome. Four gestures fully wired: pinch-rotate, two-hand zoom, hemisphere split, open-palm reset. Cortical surface renders smoothly with live activation waves.

Add region labeling so users can pinch-to-select and read out anatomy. Test with a neuroanatomy instructor for an actual classroom demo. Explore VR/AR hand tracking as a no-webcam path.