I am testing NextBrain in a dataset of older adults with MCI and evidence of cerebrovascular disease. In reviewing the outputs from NextBrain, I noticed that the cortical regions are not very well segmented, especially in comparison to SynthSeg. I have attached screenshots to illusive the differences. Of note, the T1w are generally of good quality with high signal-to-noise ratio. I am wondering if there is anything that can be done to try and improve the NextBrain cortical segmentation OR would you suggest using the cortical volumes from SynthSeg instead? We’re more interested in the subcortical regions anyway.

Thanks so much for the help

!

Narlon