In new study in The Crop Journal, scientists develop cutting edge vascular system image analysis pipeline for crops
This accurate, deep learning pipeline can detect the vascular system of plants with high detail, helping agricultural practices
Peer-Reviewed PublicationIMAGE: A DEEP-LEARNING PIPELINE BATCH-PROCESSES A SERIES OF COMPUTER TOMOGRAPHIC IMAGES OF THE MAIZE STEM TO VISUALIZE AND ANALYZE ITS VASCULAR CHARACTERISTICS. view more
CREDIT: CHUNJIANG ZHAO VIA THE CROP JOURNAL
A plant’s vascular system is essential for maintaining stem structure, providing mechanical support, and for delivering resources to various plant organs. However, the structure and distribution of these vascular bundles varies greatly across individual plants, and this poses a great challenge in automating the process of their identification and quantification. Therefore, the ability to deliver a rapid and accurate quantitative and functional evaluation of these plant systems is vital to agricultural research.
The histological study of sectioned plant tissue (done under a microscope) is the foundation of plant anatomy and microscopy techniques. Spanning light-based, fluorescence, and electron-based techniques, this is the backbone of inner tissue plant research. However, the physical and chemical treatments during sample preparation for these methods can alter the natural structure of specimens. Micro-computed tomography (micro-CT), however, can deliver high-resolution imagery with minimal preprocessing that is non-destructive to tissue specimens. Unfortunately, some manual adjustments are necessary for CT image reconstruction which results in errors and variation introduced unwittingly by the observer.
In terms of capability, Deep Convolutional Neural Network (CNN) approach, a data-driven feature extraction technique, has only recently achieved state-of-the-art performance in detecting objects in segmented images, allowing it to be routinely used in image-based phenotyping in plant phenomics (i.e., a study of the plant’s phenotype and its evolution). Now, a group of researchers at Beijing Academy of Agriculture and Forestry Sciences (BAAFS), led by Dr. Jianjun Du, has developed a CNN based deep-learning pipeline that can rapidly produce accurate analyses of vascular bundle architecture. “We believe we can break new ground in understanding the relationship between vascular bundle architecture at the single-plant level and the traits involved in water transport,” says Dr. Chunjiang Zhao, corresponding author of a study detailing their methods and findings. The study was published online on 27 May 2022 in The Crop Journal.
The team was particularly interested in studying how plants that express plasticity in the structure of their stems can quickly adapt to their environment. It is this plasticity that allows for changing structure without hindering growth and development. To study it, they grew maize under natural and drought conditions and utilized information extracted from CT images to examine the vascular bundles for different stem internodes (which is the part of the stem between two nodes, or branching areas), evaluate architectural differences in stem structure, and investigate the relationship between flow rates and structural traits. Their pipeline processed images and detected vascular bundles in the plants, identified specific zones (the periphery, the epidermis, and the inner zones) within the bundles, categorized bundles into phenotypes based on specific traits (quality, quantity, size, and shape), and performed a statistical analysis of these traits in different stem internodes. They also conducted sap flow experiments to study the traits of vascular bundles in maize at the single-plant level to gain an insight into the water use efficiency of the different phenotypes.
“We could achieve an image processing time of three seconds, and for the first time we shed light on the thickness of the epidermis (which is the outermost layer) of the maize stem. Our pipeline is incredibly accurate too. During testing, it enumerated vascular bundles across all types of internodes and quantified size-related traits with an R2–which indicates consistency–of over 0.98,” Dr. Du explains the novelty of their study.
In addition, the sap flow experiments showed that the rate of flow was affected not only by the structure of the vascular bundles, but also by environmental and meteorological conditions.
So why are these findings so important? “We believe we have laid the foundation for deeper studies on identification of genes essential to determining water use efficiency, and development of crop breeds that can ensure national food security. The pipeline certainly allows for future work to establish the relationship between sap flow and the specific traits of vascular bundles,” says Dr. Du with a smile. Their research might lead to improved, more resilient crops in the future!
***
Reference
DOI: https://doi.org/10.1016/j.cj.2022.04.012
Authors: Jianjun Du, Ying Zhang, Xianju Lu, Minggang Zhang, Jinglu Wang, Shengjin Liao, Xinyu Guo, and Chunjiang Zhao
Affiliations:
Beijing Key Lab of Digital Plant, Research Center of Information Technology, Beijing Academy of Agriculture and Forestry Sciences, China
About Professor Chunjiang Zhao
Dr, Chunjiang Zhao, a member of the Chinese Academy of Engineering, received his Ph.D from China Agriculture University in 1993. He is currently a professor at the Beijing Academy of Agricultural and Forestry Sciences and develops technology for efficient information acquisition, quantitative analysis, diagnostic decision making, and intelligent equipment control. He has published over 400 articles and founded the International Symposium on Intelligent Information Technology in Agriculture to promote cooperation between scientists and engineers. He holds advisory roles in the Ministry of Agriculture of China, the Natural Science Foundation, and the Ministry of Science and Technology.
JOURNAL
The Crop Journal
METHOD OF RESEARCH
Experimental study
SUBJECT OF RESEARCH
Not applicable
ARTICLE TITLE
A deep learning-integrated phenotyping pipeline for vascular bundle phenotypes and its application in evaluating sap flow in the maize stem
No comments:
Post a Comment