Studies of reproducibility are one of the cornerstones of science and its lack reduces the scientific efforts done up to now. Neuroimaging studies are particularly sensitive to that, because the nature of the measures is complex and there is many segmentation methods with your own techniques. Specifically to the cerebellum, these are important issues, once that structure is tiny and highly irregular. In addition, the current scanners does not have spatial resolution to acquiring the cerebellum adequately. Therefore, we propose to develop a study that assess the reliability and reproducibility of the main segmentation methods of the cerebellum, such as SUIT and CERES, and one developed by us using deep learning techniques.
News published in Agência FAPESP Newsletter about the scholarship: