Towards Reliable Measurement of Cerebellar Morphology: A Comparative Assessment of Segmentation Pipelines.
Authors
Affiliations (8)
Affiliations (8)
- Université Paris-Saclay, CEA, CNRS, NeuroSpin, BAOBAB, Gif-sur-Yvette, France.
- Université Paris-Est Créteil, INSERM, IMRB, Translational Neuropsychiatry, Créteil, France.
- Centre d'acquisition et de traitement des images, US52-UAR2031, CEA, ICM, Sorbonne Université, CNRS, INSERM, AP-HP, Ile-de-France, France.
- Child Mind Institute, New York, NY, USA.
- Université Paris-Est Créteil, INSERM, IMRB, Translational Neuropsychiatry, Créteil, France. [email protected].
- Université Paris-Saclay, CEA, NeuroSpin, UNIACT, Gif-sur-Yvette, France. [email protected].
- Fondation FondaMental, Paris, France. [email protected].
- Mondor University Hospital, AP-HP, Créteil, France. [email protected].
Abstract
Characterizing cerebellar morphology is fundamental for accurately mapping its structure and function across individuals, yet remains challenging due to its densely foliated architecture. Although multiple automated segmentation pipelines exist, the measurement reproducibility of these tools has not been comprehensively benchmarked. We conducted a systematic assessment of reliability for cerebellar morphology estimates using four commonly used pipelines: one classic parcellation method (CERES), two deep-learning methods (ACAPULCO, DeepCERES), and one voxel-based morphometry toolbox (SUIT). Leveraging the HNU dataset, which provides MRI scans for ten timepoints per individual over a month, we evaluated the test-retest reliability for each of the four pipelines using ReX, an integrative tool for quantifying and optimizing measurement reliability and individual differences. We quantified intra- and inter-individual variability, as well as the Intraclass Correlation Coefficient (ICC), for each pipeline at both global and region-of-interest levels. We used two pools of data: a common dataset of 12 subjects x 5 sessions to benchmark all the pipelines, and method-specific datasets including pools of subjects with 5 valid sessions for each pipeline to assess the generalizability of our findings. Overall, all pipelines yielded highly reliable segmentation volumes (ICC > 0.87). DeepCERES demonstrated the most consistent performance across all metrics and subject pools, while SUIT achieved the highest ICC on the common dataset but showed decreased reliability when evaluated on a larger subject pool. Importantly, our analysis highlighted substantial heterogeneity in reliability across lobules for each method. Lobule X consistently showed reduced reliability whereas lobules I-V, VI, Crus I, IX, and VIIIB were reliably estimated across all pipelines. While all pipelines demonstrated strong reliability, DeepCERES showed the most robust and consistent performance overall. Substantial lobule-specific variability and method-dependent failure rates underscore the need for reliability-aware pipeline selection to optimize cerebellar morphology estimation in research.