miqa

miqa copied to clipboard

MIQA currently has specific types (T1, T2, etc.). Real world studies have aliases: t1 = T1 = t1w = T1W = T1w. MIQA would be more useful if it was...

curtislisle

I already closed two. The rest are here: https://github.com/OpenImaging/miqa/milestones

dzenanz

Scan decisions are currently an enumerated type. MiQA would be more powerful if scan decisions can be dynamically changed for each project. Does this require making scan-decisions a model? Support...

curtislisle

Add active learning

2

Create a new web component and expose a few parameters which control re-learning: - Number of epochs - Percentage of data to use for re-training. When only a subset of...

dzenanz

aashish24

This is our specific aim 3.

dzenanz

Validate learning instructions

5

Instructions are here: https://github.com/OpenImaging/miqa/blob/master/miqa/learning/README.md

aashish24

NN currently uses mean to combine patch outputs. Check whether some other way to combine the tiles gives better results.

dzenanz

dzenanz

Use MRIQC or MRQy to establish classical image quality metrics

2

Classical image quality metrics can be additional information, in addition to outputs of our trained neural network models. MRIQC: https://github.com/poldracklab/mriqc MRQy: https://github.com/ccipd/MRQy

dzenanz