基于统计形状模型的国人骨盆解剖

Objective To construct a three⁃dimensional statistical shape model of the pelvis and analyze the individual variation and gender differences of the three⁃dimensional shape of the pelvis. Methods We collected CT data from 201 Chinese individuals and used deep learning to automatically reconstruct three⁃dimensional models of the pelvis. Through three⁃dimensional model registration, dense correspondence mesh mapping, and the use of statistical shape modelling (SSM) and principal component (PC) analysis method, we extracted models of variations (MoV) of pelvic shape changes and statistically compared the shape MoV between males and females. Results We analysed the top 10 principal components of shape variations, which accounted for 86. 1% of the total variability. Among them, PC01, PC02, and PC04 showed significant differences between genders (P<0. 001), accounting for a total variability of 60. 1%. PC08 and PC10 demonstrated pelvic asymmetry, accounting for a total variability of 3. 8%. Conclusion We constructed a three⁃ dimensional statistical shape model of the pelvis in Chinese individuals, revealing the morphological variation and sex differences of Chinese pelvis.