Precise 3D geometric phenotyping and phenotype interaction network construction of maize kernels.

Accurate identification of maize kernel morphology is crucial for breeding and quality improvement. Traditional manual methods are limited in dealing with complex structures and cannot fully capture kernel characteristics from a phenome perspective. To address this, our study aims to develop a high-throughput 3D phenotypic analysis method for maize kernels using Micro-CT-based point cloud data, thereby enhancing both accuracy and efficiency. We introduced new phenotypic indicators and developed a kernel phenome interaction network to better characterize the diversity and variability of kernel traits. Using a natural population of maize, high-resolution 2D slice data from Micro-CT scans were converted into 3D point cloud models for detailed analysis. This process led to the proposal of five new indicators, such as the endosperm density uniformity index (ENDUI) and endosperm integrity index (ENII), and the construction of their corresponding phenome interaction network. The study identified 27 3D morphological feature parameters, significantly improving the accuracy of kernel phenotypic analysis. These new indicators enable a more comprehensive evaluation of trait differences between subgroups. Results show that ENDUI and ENII are central to the phenome interaction networks, revealing synergistic relationships and environmental adaptation strategies during kernel growth. Additionally, it was found that length traits significantly impact the volumes of the embryo and endosperm, with linear regression coefficients of 0.599 and 0.502, respectively. This study not only advances maize kernel morphology research but also offers a novel method for phenotypic analysis. By enriching the phenotypic diversity of maize kernels, it contributes to breeding programs and grain processing improvements, ultimately enhancing the quality, and utilization value of maize kernels.