Nondestructive 3D Image Analysis Pipeline to Extract Rice Grain Traits Using X-Ray Computed Tomography.

Abstract

The traits of rice panicles play important roles in yield assessment, variety classification, rice breeding, and cultivation management. Most traditional grain phenotyping methods require threshing and thus are time-consuming and labor-intensive; moreover, these methods cannot obtain 3D grain traits. In this work, based on X-ray computed tomography, we proposed an image analysis method to extract twenty-two 3D grain traits. After 104 samples were tested, the R 2 values between the extracted and manual measurements of the grain number and grain length were 0.980 and 0.960, respectively. We also found a high correlation between the total grain volume and weight. In addition, the extracted 3D grain traits were used to classify the rice varieties, and the support vector machine classifier had a higher recognition accuracy than the stepwise discriminant analysis and random forest classifiers. In conclusion, we developed a 3D image analysis pipeline to extract rice grain traits using X-ray computed tomography that can provide more 3D grain information and could benefit future research on rice functional genomics and rice breeding.

Plant phenotyping relevance

X線CTを用いてイネ穀粒の22種類の3D形質を抽出する画像解析パイプラインを開発し、手動測定との一致性で検証しているため、植物フェノタイピング手法が研究の中心です。

Code and data availability

The paper's MATLAB 3D image analysis pipeline for extracting rice grain traits from X-ray CT is explicitly stated to be publicly available, with an authors' GitHub repository URL matching an allowed URL. The phenotypic data (Supplementary File 1) is only available as a PMC supplement with no allowed URL, so it is not a