Chopping my brain into bits – turning my brain into a 3D model on the web

Summary

Last year I had a brain MRI done as part of a research study at the University of Melbourne Centre for Brain, Mind and Markets. Afterwards, the researchers sent me the raw scan, which meant I could inspect my own anatomy rather than just read a radiology summary. Over the past few months I’ve been seeing how far I could push that opportunity: extracting metadata from the original file, running the scan through brain segmentation software, pulling out individual structures, and turning the result into the interactive 3D model at the top. This post is a walkthrough of that pipeline, and of what became legibile from the grayscale slices of my brain. The raw scan is fascinating but not very helpful As a medical student, I’ve been taught the basics of interpreting brain MRIs, so opening up the raw scan in a slice viewer was quite underwhelming. It was essentially the same as any other MRI I’ve seen throughout my education. However, having the scan file itself let me extract a bit more information about how it was made: Field strength 7 Tesla Around twice the strength of a hospital scanner. Sequence MP2RAGE This pulse sequence provides good contrast for brain anatomy. Voxel size 0.75 mm isotropic A voxel = a 3D pixel. Grid dimensions 224 × 320 × 320 The raw file is a 3D block rather than a single image. Total sampled voxels 22.9 million This includes my skull, eyes and neck. Approximate field of view 168 × 240 × 240 mm The physical size of the captured data. I already knew some of this from looking up the machine they used to scan me 1. However, finding out the scan dimensions was useful because they set the ceiling for everything that follows. If my scan had been particularly low-quality, I wouldn’t have even bothered continuing because no algorithm would be able to clearly separate my brain’s structures. The most immediate way to inspect the MRI was just to scroll through it in the three standard anatomical planes. Sagittal Side view through the midline structures Coron...