Automatic segmentation of subcortical nuclei on childrens cerebral MRI based on atlas
Objective To observe the value of automatic segmentation of subcortical nuclei on children's cerebral MRI based on atlas.Methods Head MRI data of 105 healthy children aged 0-12 years were retrospectively analyzed.The standard brain template and atlas were used to automatically segment the thalamus,caudate nucleus,putamen and globus pallidus on 3D T1WI using nonlinear registration algorithm.The automatic segmentation efficacy of subcortical nuclei were compared between nonlinear registration based on atlas and linear registration,and the consistency between automatic segmentation based on atlas and manual segmentation was evaluated using intra-class correlation coefficient(ICC)and Bland-Altman diagram.Results Dice similarity coefficient(DSC)of automatic segmentation based on atlas for segmenting thalamus,caudate nucleus,putamen and globus pallidus was 0.80-0.93,with 95％Hausdorff distance(HD95)of 1.00-8.00 mm,among which of putamen had the best the segmentation efficacy(all P＜0.05).DSC and HD95 of nonlinear registration based on atlas for segmenting subcortical nuclei on children's MRI were both better than those of linear registration(all P＜0.001),and the segmentation range was in good coefficient with that of manual segmentation(all ICC≥0.80,most of the differences were within 95％coefficient limit).Conclusion Subcortical nuclei on children's cerebral MRI could be accurately and rapidly automatically segmented based on atlas.
万方数据
维普
