背景：临床通过耳廓再造修复先天性小耳畸形，构建高保真、结构稳定且安全的耳软骨支架一直是耳廓再造的研究核心。耳软骨支架的构建方式不断改进，先后经历了肋软骨手工雕刻法、软骨组织工程和3D生物打印法。相比前两种方式，3D生物打印技术可以通过计算机辅助设计（computer-aided design，CAD）精确控制耳软骨支架的宏观和微观结构，并且不需要截取患者肋软骨，减轻了对患者的伤害，因而是目前精确构建耳软骨支架的主要研究方向。图像采集是影响3D生物打印结构精度的一个主要因素，3D激光扫描、计算机断层扫描（computer tomography，CT）和常规的磁共振成像（magnetic resonance imaging，MRI）扫描序列对耳软骨成像效果较差，不适合分割耳软骨，所以构建的模型精度上仍有一定误差。本研究旨在探索适合耳软骨成像的扫描序列，并构建高精度的耳软骨支架模型，为3D生物打印提供模型基础。

方法：用3.0 T MRI（Philips，Achieva）对40位健康志愿者的单侧耳廓进行多模态MRI成像，包括UTE、3D-T2、PDW和PROSET。首先由两位有经验的分割者（Rater1和Rater2）根据UTE和3D-T2手动重复分割耳软骨，并计算耳软骨的体积（Cg.V）、表面积（Cg.S）和厚度（Cg.Th）。之后基于以上形态学指标，对组内（Intra-rater）分割结果进行配对t检验，并分析组内和组间（Inter-rater）分割的精度误差（precision error，PE）、组内相关系数（intraclass correlation coefficient，ICC）、Pearson相关系数和Dice相似性系数（Dice similarity coefficient，DSC），PE和ICC用于评估分割的再现性，ICC是衡量和评价观察者间信度和复测信度的信度系数指标之一。DSC可以进一步评价分割结果的相似性。然后结合UTE、3D-T2和PDW对耳朵完成精细结构分割。最后对耳软骨模型进行修复和切片处理，提供可以用于3D生物打印的耳软骨支架模型。

结果：本研究找到一种适合耳软骨分割的扫描方案。UTE和3D-T2可以用于分割耳软骨，组内Cg.V和Cg.Th无显著性差异，组内和组间分割的Cg.V、Cg.S和Cg.Th的精度误差（PE_%CV）均小于5%，ICC分别高于0.95、0.89、0.69，Pearson相关系数依次高于0.95、0.95、0.78，DSC可以达到80%。研究表明该扫描方案的精度可以表征个体间的耳软骨形态差异，组内和组间的耳软骨分割结果具有良好的重复性。此外，尽管Cg.Th的相关性较低，但不同的评估者仍然可以重复识别完整的耳软骨轮廓。

  本研究根据对耳软骨有特异性成像效果的MRI序列，构建出具有高保真性的耳软骨支架模型。此外，本研究结合UTE、3D-T2和PDW将外耳廓分成十三个精细解剖结构，提出将耳朵精细结构用于辅助耳廓重建的手术规划和术后评估。

结论：本研究表明，基于UTE和3D-T2的手动分割结果的精度足以检测患者之间的耳软骨形状差异，能够作为耳软骨3D生物打印模型构建的图像采集方案。基于个体特异性的耳软骨成像序列，本课题构建出能够用于3D生物打印的耳软骨支架模型，此外还提供了一种结合耳朵精细结构来辅助临床医生制定手术方案和术后评估的方法。

关键词：耳软骨分割、3D生物打印模型构建、耳廓再造

Background：Constructing a high-fidelity, structurally stable and safe auricular cartilage scaffold has always been the core of auricle reconstruction to repair congenital microtia. Construction methods of the ear cartilage scaffold have been continuously improved, including hand crafting of rib cartilage, cartilage tissue engineering and 3D bioprinting. Compared with the first two methods, 3D bioprinting technology can accurately control the macroscopic and microscopic structure of ear cartilage scaffold through computer-aided design (CAD) without intercepting the patient's costal cartilage, reducing the injury to the patients. Therefore, it is the primary research direction of accurate construction of the ear cartilage scaffold model. Image acquisition is one of the major factors affecting the accuracy of 3D bioprinting structures. 3D laser scanning, computer tomography (CT) and conventional magnetic resonance imaging (MRI) sequences have poor imaging effects on auricular cartilages, which are not suitable for the segmentation of ear cartilages, resulting some precision errors in the constructed models. This study aims to explore scanning protocols suitable for auricular cartilage and to construct a high-precise auricular cartilage scaffold providing the model basis for 3D bioprinting.

Methods：MRI was performed on the unilateral auricles of 40 healthy volunteers with 3.0 T Philips (Achieva), including UTE, 3D-T2, PDW and PROSET. Firstly, the auricular cartilages were manually segmented by two experienced raters (Rater1 and Rater2) with UTE and 3D-T2，and cartilage volume (Cg.V), surface area (Cg.S), and thickness (Cg.Th) were calculated .Then, based on the above morphological indexes, the paired t-test was performed on the intra-rater segmentation results, and the precision errors (PE), intraclass correlation coefficients（ICC）, Pearson correlation coefficients and Dice similarity coefficients（DSC）of intra-rater and inter-rater segmentation were analyzed. PE and ICC were used to evaluate the segmentation reproducibility. ICC was one of the reliability coefficient indicators for evaluating the inter-observer reliability and test-retest reliability. DSC can further evaluate the similarity of segmentation results. Fine structure segmentation was then performed on the ear based on UTE, 3D-T2 and PDW. Finally, the ear cartilage model was repaired and sliced to provide an ear cartilage scaffold model for 3D bioprinting.

Results：This study discovered a scanning scheme suitable for the ear cartilage segmentation. UTE and 3D-T2 could be used to segment ear cartilages. There was no significant difference between Cg.V and Cg.Th in intra-rater. Inter- and intra-rater precision errors（PE_%CV）of Cg.V, Cg.S and Cg.Th were less than 5%, intraclass correlation coefficients (ICC) were higher than 0.95, 0.89, 0.69, respectively, and Pearson’s correlation coefficients were higher than 0.95, 0.95, 0.78.DSC could reach 80%. This study demonstrated that the scanning protocol could characterize the variation of ear cartilage morphology between individuals, and good inter- and intra-rater segmentation reproducibility of auricular cartilages were achieved. Besides, although Cg.Th had a low correlation, different raters could detect intact ear cartilage contours.

  A high-fidelity auricle cartilage scaffold model was constructed based on MRI sequences with specific imaging effects on the ear cartilage. In addition, based on UTE, 3D-T2 and PDW, the external auricle was divided into thirteen fine structures. This study proposed the use of ear fine structures to assist in the surgical planning and postoperative evaluation of auricle reconstruction.

Conclusions：This study demonstrated that the precision of the manual segmentation results based on UTE and 3D-T2 was sufficient to detect patient-specific variation in auricular cartilage shape, and the proposed scanning protocol could be used as an image acquisition strategy for the construction of auricular cartilage scaffold model in 3D bioprinting. Based on the individual-specific ear cartilage imaging sequences, this study constructed ear cartilage scaffold models for 3D bioprinting, and also provided a method for combining the fine structures of the ear to assist the clinician in developing a surgical plan and postoperative evaluation.

Key words：auricular cartilage segmentation，3D bioprinting model construction，auricle reconstruction