Segmentation of roots in soil with U-Net

Abstract

Background Plant root research can provide a way to attain stress-tolerant crops that produce greater yield in a diverse array of conditions. Phenotyping roots in soil is often challenging due to the roots being difficult to access and the use of time consuming manual methods. Rhizotrons allow visual inspection of root growth through transparent surfaces. Agronomists currently manually label photographs of roots obtained from rhizotrons using a line-intersect method to obtain root length density and rooting depth measurements which are essential for their experiments. We investigate the effectiveness of an automated image segmentation method based on the U-Net Convolutional Neural Network (CNN) architecture to enable such measurements. We design a data-set of 50 annotated chicory (Cichorium intybus L.) root images which we use to train, validate and test the system and compare against a baseline built using the Frangi vesselness filter. We obtain metrics using manual annotations and line-intersect counts. Results Our results on the held out data show our proposed automated segmentation system to be a viable solution for detecting and quantifying roots. We evaluate our system using 867 images for which we have obtained line-intersect counts, attaining a Spearman rank correlation of 0.9748 and an $$r^2$$r2 of 0.9217. We also achieve an $$F_1$$F1 of 0.7 when comparing the automated segmentation to the manual annotations, with our automated segmentation system producing segmentations with higher quality than the manual annotations for large portions of the image. Conclusion We have demonstrated the feasibility of a U-Net based CNN system for segmenting images of roots in soil and for replacing the manual line-intersect method. The success of our approach is also a demonstration of the feasibility of deep learning in practice for small research groups needing to create their own custom labelled dataset from scratch.

Genomic prediction of yield and root development in wheat under changing water availability

“…The root images were made along the upwards facing side of the minirhizotrons, therefore enable photography of roots covering a soil depth interval of 0.7 m to 2.7 m [6]. The subsequent image analysis delivered an estimate of living root length in each image using the U-Net Neural Network (CNN) architecture to provide automated image segmentation of root structures [47]. A detailed description of the image analysis strategy can be found in a previous study [31].…”

High-Throughput in situ Root Image Segmentation Based on the Improved DeepLabv3+ Method

“…However, this traditional manual segmentation method is greatly affected by human subjectivity, and the segmentation time is longer, approximately 2 to 3 h for an image, making it an inefficient method. Therefore, a highefficiency and high-accuracy in situ root image segmentation method is needed to support in situ root phenotype research (Smith et al, 2020). Improving image quality is the most important issue in in situ root system research.…”

“…The first application of pixel-level semantic segmentation tasks is the fully convolutional network (FCN) approach (Long et al, 2014), which uses an encoder-decoder structure to automatically extract target features and classify all pixels in an image one by one. In research on root image segmentation based on deep learning, Smith et al (2020) proposed a U-Net-based root segmentation system; this proposed network architecture is also composed of an encoder-decoder structure. Compared with the traditional machine learning method using Frangi vessel enhancement filter (Frangi et al, 1998), U-net can segment the root morphology in soil images with higher accuracy.…”

Three Dimensional Root CT Segmentation Using Multi-Resolution Encoder-Decoder Networks

“…U-Net is a widely cited and deployed encoder-decoder structure that was originally proposed for medical image segmentation [22]. Within the field of root image analysis, Smith et al [23] proposed using U-Net for segmentation of roots from soil in 2D rhizotron images, comparing their method to the Frangi vesselness [24] filter, an image-based approach originally designed for segmentation of (e.g. blood) vessels that have similar structure to roots.…”