A road map to manual segmentation of cerebral structures
Poster No:
2008
Submission Type:
Abstract Submission
Authors:
Fanny Darrault1, Guillaume Dannhoff1,2, Simon Louchez1, Théo Delmaire1, Maëlig Chauvel3,4, Cyril Poupon4, Ivy Uszynski4, Christophe Destrieux1,5, Igor Lima Maldonado1,5, Frédéric Andersson1
Institutions:
1UMR 1253, iBrain, Université de Tours, Inserm, Tours, France, 2CHRU de Strasbourg, Strasbourg, France, 3Max Planck Institute, Leipzig, Germany, 4BAOBAB, NeuroSpin, Paris-Saclay University, CNRS, CEA, Gif-sur-Yvette, Essonne, 5CHRU de Tours, Tours, France
First Author:
Co-Author(s):
Guillaume Dannhoff
UMR 1253, iBrain, Université de Tours, Inserm|CHRU de Strasbourg
Tours, France|Strasbourg, France
UMR 1253, iBrain, Université de Tours, Inserm|CHRU de Strasbourg
Tours, France|Strasbourg, France
Maëlig Chauvel
Max Planck Institute|BAOBAB, NeuroSpin, Paris-Saclay University, CNRS, CEA
Leipzig, Germany|Gif-sur-Yvette, Essonne
Max Planck Institute|BAOBAB, NeuroSpin, Paris-Saclay University, CNRS, CEA
Leipzig, Germany|Gif-sur-Yvette, Essonne
Christophe Destrieux
UMR 1253, iBrain, Université de Tours, Inserm|CHRU de Tours
Tours, France|Tours, France
UMR 1253, iBrain, Université de Tours, Inserm|CHRU de Tours
Tours, France|Tours, France
Igor Lima Maldonado
UMR 1253, iBrain, Université de Tours, Inserm|CHRU de Tours
Tours, France|Tours, France
UMR 1253, iBrain, Université de Tours, Inserm|CHRU de Tours
Tours, France|Tours, France
Introduction:
Manual segmentation is an essential tool in the researcher's technical arsenal. It is a frequent practice necessary for image analysis in many protocols, especially when regions of interest are required (volumetry, connectivity, etc.). Despite its evident relevance, the manual segmentation process has received little attention in the literature. Some works mention addressing strategies for specific structures (Yushkevich et al, 2015), but – to our knowledge – no publication has discussed what should be considered good practices and why. Similarly, in many papers, despite the major impact of the quality of the segmentation on the results, the details of how the segmentation was planned and carried out are often briefly or not mentioned. Although manual segmentation is an expert-dependent procedure, the upstream identification of the best methodological strategies ensures quality and reproducibility. Furthermore, recent advances in neuroimaging, such as ex vivo ultra-high field MRI, enable new acquisition modalities and the visualization of minute structures.
Methods:
This work is the result of a narrative review of the literature with input from expert anatomists with manual segmentation experience. We propose the comprehensive checklist in figure 1 as a guiding tool for manual segmentation of cerebral structures, from planning to reporting.
Results:
As a synthesis of reviewing the literature and our experience with manual segmentation, we emphasized different critical steps (Fig. 2) and gathered tips and tricks to make for an easier and less time-consuming task (Hashempour et al., 2019, Keuken and Forstmann, 2015, Lechanoine et al, 2021). A combination of enlightened choices before the segmentation clarifies how to deal with the limits of anatomical classes that pose the most problems with this technique. Consideration of segmentation usage, subsequent processing, and expected results is necessary to avoid certain pitfalls. Reporting these choices in articles is recommended in order to enhance reproducibility and enable comparison of the results from different studies. By describing precisely what limits were used for each anatomical class, stating if those limits were followed conservatively or liberally, and thoroughly citing the papers and atlases used, a reader will better understand which anatomical area is being referred to. The abovementioned recommendations are summarized in a checklist to be used as a tool for manual segmentation (Fig. 1).
Conclusions:
In this context of new challenges, we propose a general roadmap to optimize both the technique and the reporting of manual segmentation of cerebral structures. If implemented, it would greatly contribute to open and responsible science. Well-conducted and well-described segmentation procedures will increase the validity of research works, which is a significant issue for data reproducibility and results in neuroimaging.
Modeling and Analysis Methods:
Segmentation and Parcellation 1
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
Cortical Anatomy and Brain Mapping 2
Novel Imaging Acquisition Methods:
Anatomical MRI
Keywords:
Segmentation
Other - anatomy; parcellation; manual segmentation
1|2Indicates the priority used for review
Provide references using author date format
Keuken, M.C., et al., A probabilistic atlas of the basal ganglia using 7 T MRI. Data in Brief 4, 577–582. https://doi.org/10.1016/j.dib.2015.07.028
Lechanoine, F., et al., 2021. WIKIBrainStem: an online atlas to manually segment the human brainstem at the mesoscopic scale from ultrahigh field MRI. NeuroImage 236.
Yushkevich, P.A., et al., 2015. Quantitative comparison of 21 protocols for labeling hippocampal subfields and parahippocampal subregions in in vivo MRI: Towards a harmonized segmentation protocol. NeuroImage 111, 526–541. https://doi.org/10.1016/j.neuroimage.2015.01.004