search

QTIM-Lab / DeepNeuro

A deep learning python package for neuroimaging data. Made by:
https://qtim-lab.github.io
MIT License
124 stars 36 forks source link

Error-Catching for bad cases in data groups. #23

Closed AnnaBeers closed 5 years ago

AnnaBeers commented 6 years ago

I recently ran into a situation where one modality in one data group in one case out of a hundred cases had an extra dimension. Particularly, a FLAIR sequence had two time-points. This caused an error in patch extraction, as patches were being extracted at the wrong dimension.

We'll face a lot of these unexpected data errors as time goes on. Currently, data is appended list-wise after its processed via augmentation. If an error happens in one data group (e.g. input modalities), but not the other (e.g. ground-truth) the two lists of data will become de-synced, leading to disaster. We need a way to A) catch errors in one data group, and then B) remove all corresponding data from the other data group.

AnnaBeers commented 5 years ago

Merging this issue with https://github.com/QTIM-Lab/DeepNeuro/issues/36