search

khanlab / hippunfold

BIDS App for Hippunfold (automated hippocampal unfolding and subfield segmentation)
https://hippunfold.readthedocs.io
MIT License
47 stars 12 forks source link

Hippunfold on pediatric data? #283

Open oliver-xie opened 3 months ago

oliver-xie commented 3 months ago

Hi Jordan, I am aware that Hippunfold was trained with adult data, and I am curious if you have already validated Hippunfold on kids' data or have plans to do so. Thanks, Oliver

jordandekraker commented 2 months ago

Hi Oliver, sorry for the slow response. I don't know of any project to apply HippUnfold to children. I suspect it will work in some cases, but may have a higher than usual failure rate. If you're interested, you could fine-tune our segmentation model on training data from a subset of images where the results are good, and/or add additional cases by manual segmentation. I'd be happy to help with this, and if you're willing to make your fine-tuned model openly available then it would be a very valuable addition to HippUnfold! (happy to give you credit of course)

Bradley-Karat commented 2 months ago

Hi both, sorry didn't see this earlier - Oliver what is the age range for your data? I recently ran hippunfold on T1w images in a sample of healthy 8-18 year olds. Overall the results looked good, and only required slight manual corrections in a handful of hippocampi.

oliver-xie commented 1 month ago

Hi Jordan and Bradley, sorry for the late reply. I plan to use it on the HCP-D dataset (5-21), with the future goal of applying HippUnfold to pediatric patients. It is good to know that it worked well for 8 yo TDs. My collaborators at UMD recently have started a project to compare ASHS/freesurfer/Hippunfold subfield segmentation in young children (4-8yo), who also do manual segmentation. Depending on their findings, we will see if fine-tuning is needed for HippUnfold for kids. Also, I know at least one more study that uses hippounfold in young kids (https://www.biorxiv.org/content/10.1101/2023.11.03.565423v1).