Multi-atlas based auto-segmentation for hippocampus
Hu Ha · Jisuanji gongcheng yu sheji · 2014
A new multi-atlas based segmentation method for hippocampus from T1weighted magnetic resonance brain images is proposed to study some nervous system disease.A spatial prior is introduced by registration to achieve an accurate auto-segmentation result.To improve the registration precision,a new residual-complexity-based method which is more robust and more accurate is chosen.After the atlases are refined based on the deforming field,STAPLE fusion method is adopted to get the final segmented result.Finally,the experimental results indicate that this algorithm is suitable and accurate enough for segmentation of hippocampus.