Fig. 3: Data analysis flowchart.

A Patient and multi-site healthy children enrolled. B Patients’ behaviors were assessed in the neurosurgical clinic using the Child Behavior Checklist. C Structural MRI (T1) data acquisition and processing. Patients’ brainstem was excluded before segmentation. D All participants’ 28 cerebellar regions were obtained by Automatic Cerebellum Anatomical Parcellation using U-Net with Locally Constrained Optimization. E Normative trajectory in the GPR Model. In each model, a Gaussian model of the volume of each brain region was established with age and sex as covariates. F The Z-score profile of individual cerebellar regions’ volume. G Pearson correlation matrix of cerebellar regions and behavioral problems. H The deviations of cerebellar volume predict behaviors. I Feature weight maps for 26 cerebellar regions. GPR Gaussian Process Regression.