The shape of kernels and cracks, in a nutshell

We explore the shape of walnut shells and kernels. We analyze 1264 individual 3D X-ray CT scan reconstructions of walnuts, corresponding to 150 walnut accessions. We exploit the nondestructiveness of X-rays to digitally segment and measure the 4 main tissues of interest for each walnut, namely shell, kernel, packing tissue, and sealed air. From these, we extract a total of 38 size- and shape-specific descriptors, many of them unexplored in the current literature. We focus on several allometric relationships of interest, from which we draw theoretical upper and lower bounds of possible walnut and kernel sizes. We then study correlations and variations of these morphological descriptors with qualitative data corresponding to traits of commercial interest like ease of kernel removal and shell strength., All plant materials represent walnut breeding lines, germplasm, and cultivars maintained by the Walnut Improvement Program at the University of California, Davis. A total of 150 walnuts accessions were harvested into mesh bags at hull split, oven-dried overnight at 95F, and then air-dried for several weeks before moving into cold storage at 35F. 5 to 16 individuals were selected for each accession, for a total of 1301 individual walnuts to be scanned at Michigan State University. The walnuts were scanned in 171 batches. The scans were produced using the North Star X3000 system and the included efX-DR software. The X-ray source was set at 75 kV and 100~\micro{A}, with 720 projections per scan, at 3 frames per second and with 3 frames averaged per projection. The data was obtained in continuous mode. The 3D X-ray CT reconstruction was computed with the efX-CT software, obtaining voxel-based images with voxel size of 75.9 microns., All the files are either 3D TIFF images, CSVs with associated metadata, or JPGs for visualization and inspection purposes.
All the TIFF files were manipulated as 3D numpy arrays with Python. A GitHub repository is referenced with plenty of Jupyter notebooks that provide more detail on how the images were handled and how the CSVs and images were produced.

我们探究核桃壳与核桃仁的形态特征。本研究分析了1264份独立的核桃三维X射线计算机断层扫描（CT）重建数据，对应150份核桃种质资源（accessions）。利用X射线的非破坏性，我们对每个核桃的4种主要目标组织进行数字分割与测量，即壳、仁、填充组织及封闭空气。基于这些数据，我们提取了共计38个尺寸与形状特异性描述符，其中多数在现有文献中尚未被探索。我们聚焦于若干关键异速生长关系（allometric relationships），并由此推导核桃及核桃仁可能尺寸的理论上下限。随后，我们研究了这些形态学描述符与商业相关性状（如取仁难易度、壳强度）定性数据之间的相关性及变异规律。

所有植物材料均为加州大学戴维斯分校核桃改良项目所保存的核桃育种系、种质及品种。150份核桃种质资源于裂壳期采收并装入网袋，经95F烘箱过夜干燥后，风干数周，再转入35F冷库贮藏。每份种质选取5至16个单果，共计1301个核桃个体在密歇根州立大学完成扫描，扫描分为171批次。扫描采用北极星X3000系统及配套的efX-DR软件，X射线源设置为75千伏（kV）与100微安（μA），每扫描包含720个投影，帧率为每秒3帧，且每个投影取3帧平均。数据以连续模式采集，三维X射线CT重建通过efX-CT软件完成，得到体素大小为75.9微米的体素图像。

所有文件类型包括三维TIFF图像、含关联元数据的逗号分隔值文件（CSV），以及用于可视化与检查的JPG图像。所有TIFF文件均通过Python语言处理为三维NumPy数组（NumPy array）。研究引用了一个GitHub仓库，其中包含大量Jupyter笔记本，详细阐述了图像处理方法及CSV与图像的生成过程。