Science

Our physicists and engineers just implemented an open-source AI model called TotalSegmentator. It’s a machine learning model that automatically draws 3D outlines around organs and tissues on CT scans — like the lungs, liver, even individual muscles.

DOI: 10.1038/s41467-023-36337-4

Normally, this kind of segmentation takes hours to days of manual work. TotalSegmentator does it in minutes. I just check it for small errors and fix them if needed.

Because I’m  working with whole-body PET/CT, this is an absolute game-changer. I’ve basically shaved off 1–3 weeks of boring data wrangling per dataset using open-source AI.

Bottom line: what used to support maybe 1–2 studies can now power 10+ new analyses from a single scan. Total segmentation = total acceleration.

The only sad part? The best way to say thanks, is to cite the authors and/or star their GitHub repo — which feels a bit underwhelming for saving me 3 weeks of soul-crushing VOI drawing.