Project II - 3D Quantitative Analysis of Mouse Models of Structural Birth Defects Through Computational Anatomy

项目摘要 E15.5将从项目1中获得来自85个sci-RNA-seq系中的每一个的6个纯合突变体的胎儿 在建立扩散碘造影增强扫描后，通过高分辨率microCT进行成像 用于胎鼠的方案。还将对多达100只未受影响的同窝出生的胎儿进行成像，以得出 标准样本的合成群体模板（图1），并将用于评估“正常” E15.5时小鼠发育的表型变异（目的1）。模板将被分割到最精细的 可能包含解剖学细节，将包含比现有KOMP 2胎儿更多的解剖结构和细节 阿特拉斯将评估每个sci-RNA-seq细胞系的总体表型变异性（图3），然后将其 对组织体积和形状差异（目标2）进行特定区域统计（图4）分析。我们 还将训练一个深度学习模型来实现胎鼠的自动分割，并比较结果 到目前最先进的，基于模板的分析（目标3）。 我们的目标是利用3D成像、计算机视觉和机器学习的最新进展， 小鼠突变体的3D形态学表征更加准确、定量、可重复和 容易接近结合分子表型和解剖表型将提供一个全面的 表征具有与人类结构性出生缺陷相关的突变的小鼠。

PROJECT SUMMARY E15.5 fetuses of 6 homozygous mutants from each of the 85 sci-RNA-seq lines will be acquired from Project 1 to be imaged via high-resolution microCT following established diffusible iodine contrast enhanced scan protocols for fetal mice. Up to hundred fetuses of unaffected littermates will also be imaged to derive a synthetic population template of normative samples (Figure1) and will be used to assess the ‘normal’ phenotypic variation in mouse development at E15.5 (Aim 1). The template will be segmented to the finest anatomical detail possible and will contain more anatomical structure and detail than the existing KOMP2 fetal atlas. Each sci-RNA-seq line will be assesses for overall phenotypic variability (Figure 3) and then be subjected to region-specific statistical (Figure 4) analysis of tissue volume and shape differences (Aim 2). We will also train a deep-learning model to achieve automated segmentation of fetal mice and compare the results to the current-state-of-the-art, template-based analysis (Aim 3). Our goal is to leverage recent advances in the 3D imaging, computer vision and machine-learning to make the 3D morphological characterization of mouse mutants to be more accurate, quantitative, reproducible and accessible. Combining molecular phenotype and anatomical phenotypes will provide a comprehensive characterization of mice with mutations relevant to human structural birth defects.