search

Deep-MI / BrainPrint

A python-only derivative of the original BrainPrint scripts
MIT License
11 stars 9 forks source link

Integration with SynthSeg #6

Closed jaivardhankapoor closed 1 year ago

jaivardhankapoor commented 1 year ago

Hi,

Is there ongoing work on integrating the library with SynthSeg neural net-based segmentation software? this would possibly massively speed up analyses.

m-reuter commented 1 year ago

Hi, everything that delivers a FreeSurfer-like output (with the same label ids etc) should work.

Note, that we distribute FastSurfer which is performing the segmentation in less than 1 minute and is recommended over SynthSeg for T1 images of resolutions from .6 to around 1.2mm as it outperforms SynthSeg by a margin. For lower resolutions and other sequences than T1, SynthSeg is best.

jaivardhankapoor commented 1 year ago

Thanks! I'm curious to know where the evaluations of FastSurfer and SynthSeg are performed, as I did not see any comparisons in any of the two accompanying publications.

m-reuter commented 1 year ago

Comparison-SynthSeg-FastSurfer.pdf Here we show a head-to-head comparison of different versions of FastSurfer and Synthseg on the Mindboggle (manual) and a few other test datasets.