DiMSuM dataset (Diffusion MRI of Subjects with Movement artifacts)
Expected data release: Spring 2020
Neonatal and in utero MRI techniques are limited by several pitfalls and processing errors, which must be considered when interpreting such MRI data in research and clinical practice. Our dataset aims to provide real-life reference data for developing image processing algorithms for tackling head motion related artifacts during diffusion tensor MRI (DT-MRI).
The DImSuM dataset consists of two parts:
1. DT-MRI of tractography phantoms. We studied the patterns of in utero fetal head movements and imitated them during the scanning of DT-MRI Phantoms. Our dataset consists of DT-MRI data of two phantoms, with no motion, slight and severe motion, each repeated 3 times in axial and coronal planes. This allows testing reproducibility.
We used DT-MRI phantoms developed and manufactured by HQ Imaging (http://hq-imaging.de/). In these devices, parallel polyester fibers are wound on a circular spindle. The spindle with the fibers is immersed in agarose gel, ensuring high diffusion anisotropy and a controlled diffusion coefficient. The FA of the fibers in both phantoms is around 0.4, which matches the FA value of the corpus callosum of fetuses during mid-gestation.
2. Foetal, in utero DT-MRI data. We acquired 3 axial and one coronal DT scans for 15 fetuses with minimal to severe head motion present during scanning.Our data release includes 50, anonymized data sets in NIFTI format.
The DImSuM dataset is available for academic research projects after registration.
The project is supported by the Stiftung für wissenschaftliche Forschung an der Universität Zürich.