MRI Head Explorer

Welcome inside my head 🧠

Hi, I’m Volodymyr. In October 2026 I had a routine brain MRI. The clinic only gave me a viewer with grey slices, so I pulled the raw scan data out, ran it through a bunch of open-source neuroscience tools, and built this. Everything you see is really me: skin, skull, every fold of my brain, the inner ears and even the veins. None of it is a textbook model.

Getting around

• Rotate: drag with the left mouse button (or one finger)
• Move: drag with the right mouse button (or two fingers)
• Zoom: mouse wheel or pinch
• Learn: hover over anything to see what it is and what it does; click to keep the note on screen

Not sure where to start? Take the tour

▶ Take the tour walks you through it, from the raw scan to the inner ear. Pause, skip, or leave with Esc whenever you like. The guided views below it jump straight to one scene.

Tools

• Volume rendering shows the raw scan as a glowing cloud. Move Threshold to peel off the outer layers.
• Cut slices the head open; ⇄ picks which half goes away.
• Ruler: click two points to measure in millimetres. Esc clears it.
• Split brain pulls the two hemispheres apart.

This is for curiosity, not a diagnosis. The algorithms are good but not perfect, so borders can be a millimetre or two off.

How I made this

I had a routine brain MRI (3 Tesla Siemens Skyra, 23 series). The clinic’s portal only shows pictures, but its viewer uses a hidden “download the whole study” function, so I called it directly and got the original DICOM files. I removed ID numbers and dates from the headers and processed everything with free, open-source research tools:

• Brain regions: FastSurfer (deep-learning segmentation, DKT atlas) on the 0.5 mm T1 scan
• Hypothalamus, cerebellar lobules, corpus callosum: FastSurfer’s HypVINN, CerebNet and CC modules
• Folded and inflated cortex: SimNIBS TopoFit surfaces, inflated FreeSurfer-style
• Skull & eyes: SimNIBS charm (atlas-based head segmentation) on T1 + T2, then split into bones
• Optic nerves & eye muscles: traced through the orbit fat on the T1 scan
• Skin: thresholding of the T1 scan
• Inner ear: fluid in the 0.3 mm T2 SPACE scan, found automatically
• Veins: vesselness filter on the SWI scan, with the phase used to reject artefacts; sinuses from charm
• 3D meshes: marching cubes, smoothing and Draco compression, about 4 million triangles in total

The viewer is a single HTML file using three.js. Source code, data and the full methodology are on GitHub.

Things that aren’t perfect

• Bone is basically invisible on MRI (it shows up black), so the skull is estimated by an atlas-based model. The overall shape is right; fine details are approximate.
• A few “veins” near the skull base are really artefacts from the air in my sinuses.
• The eye muscles are roughly in the right places, but their exact shapes are approximate.
• Automatic labels can be a few millimetres off at the borders.