Clinica's image processing pipeline QC #41
Description
Authors
Matthieu Joulot [email protected]
Ju-Chi Yu [email protected]
Summary
The goal of the project was to evaluate different QC metrics and visuals for pipelines currently existing in Clinica, which deal with registration or segmentation. That way, we would be able to find a few good metrics to separate the good images from the bad or moderately good ones, which the user would check using some visual we would generate for him.
We were able to output some first results for registration, and keeping only two metrics, the correlation coefficient and the dice made from HD BET which gave us this graph. which enables a correct separation of the categories.
Figure 1: Scatterplot of HD BET Dice probability and correlation ratio between reference and moving images. Red color means both metrics are below the threshold, blue or purple means one of the metrics is below the threshold, green means both metrics are above the threshold.
We'll look further into this based on this, so that we can hopefully have some good idea of what we want to implement in the future version of Clinica which will include QC.
References (Bibtex)
@Article{routier2021clinica,
title={Clinica: An open-source software platform for reproducible clinical neuroscience studies},
author={Routier, Alexandre and Burgos, Ninon and D{'\i}az, Mauricio and Bacci, Michael and Bottani, Simona and El-Rifai, Omar and Fontanella, Sabrina and Gori, Pietro and Guillon, J{'e}r{'e}my and Guyot, Alexis and others},
journal={Frontiers in Neuroinformatics},
volume={15},
pages={689675},
year={2021},
publisher={Frontiers Media SA}
}