BACKGROUND: Manual analysis of (mini-)rhizotron (MR) images is tedious. Several methods have been proposed for semantic root segmentation based on homogeneous, single-source MR datasets. Recent advances in deep learning (DL) have enabled automated feature extraction, but comparisons of segmentation accuracy, false positives and transferability are virtually lacking. Here we compare six state-of-the-art methods and propose two improved DL models for semantic root segmentation using a large MR dataset with and without augmented data. We determine the performance of the methods on a homogeneous maize dataset, and a mixed dataset of > 8 species (mixtures), 6 soil types and 4 imaging systems. The generalisation potential of the derived DL models is determined on a distinct, unseen dataset. RESULTS: The best performance was achieved by the U-Net models; the more complex the encoder the better the accuracy and generalisation of the model. The heterogeneous mixed MR dataset was a particularly challenging for the non-U-Net techniques. Data augmentation enhanced model performance. We demonstrated the improved performance of deep meta-architectures and feature extractors, and a reduction in the number of false positives. CONCLUSIONS: Although correction factors are still required to match human labelled root lengths, neural network architectures greatly reduce the time required to compute the root length. The more complex architectures illustrate how future improvements in root segmentation within MR images can be achieved, particularly reaching higher segmentation accuracies and model generalisation when analysing real-world datasets with artefacts-limiting the need for model retraining.

Department of Computer Science University of Copenhagen Copenhagen Denmark

Dept Biogeochemical Integration Max Planck Institute for Biogeochemistry Jena Germany

Discipline of Botany School of Natural Sciences Trinity College Dublin Ireland

Faculty of Forestry and Wood Technology Mendel University in Brno Brno Czech Republic

IDLab Department of Computer Science University of Antwerp Imec Antwerp Belgium

Institute of Agronomy University of Natural Resources and Life Sciences Vienna Vienna Austria

Institute of Forest Ecology Department of Forest and Soil Sciences University of Natural Resources and Life Sciences Vienna Vienna Austria

School of Electrical and Computer Engineering Ben Gurion University of the Negev Beer Sheva Israel

Vienna Scientific Instruments GmbH Alland Austria

Wyler Department for Dryland Agriculture French Associates Institute for Agriculture and Biotechnology of Drylands Jacob Blaustein Institutes for Desert Research Ben Gurion University of the Negev Sede Boqer Campus Beersheba Israel

Zobrazit více v PubMed

Smit AL, Bengough AG, Engels C, van Noordwijk M, Pellerin S, van de Geijn SC. Root methods: a handbook. Heidelberg: Springer Science Business Media; 2013.

Ryan PR, Delhaize E, Watt M, Richardson AE. Plant roots: understanding structure and function in an ocean of complexity. Ann Bot. 2016;118(4):555–559. doi: 10.1093/aob/mcw192. DOI

Freschet G, Roumet C, Comas L, Weemstra M, Bengough A, Rewald B, et al. Root traits as drivers of plant and ecosystem functioning: current understanding, pitfalls and future research needs. New Phytol. 2021;232(3):1123–1158. doi: 10.1111/nph.17072. PubMed DOI

Schroth G. A review of belowground interactions in agroforestry, focussing on mechanisms and management options. Agrofor Syst. 1998;43(1):5–34. doi: 10.1023/A:1026443018920. DOI

Wijesinghe DK, John EA, Hutchings MJ. Does pattern of soil resource heterogeneity determine plant community structure? An experimental investigation J Ecol. 2005;93(1):99–112.

McCormack ML, Dickie IA, Eissenstat DM, Fahey TJ, Fernandez CW, Guo D, et al. Redefining fine roots improves understanding of below-ground contributions to terrestrial biosphere processes. New Phytol. 2015;207(3):505–518. doi: 10.1111/nph.13363. PubMed DOI

Pierret A, Doussan C, Capowiez Y, Bastardie F, Pagès L. Root functional architecture: a framework for modeling the interplay between roots and soil. Vadose Zone J. 2007;6(2):269–281. doi: 10.2136/vzj2006.0067. DOI

Lynch JP, Strock CF, Schneider HM, Sidhu JS, Ajmera I, Galindo-Castañeda T, et al. Root anatomy and soil resource capture. Plant Soil. 2021;466(1):21–63. doi: 10.1007/s11104-021-05010-y. DOI

Atkinson JA, Pound MP, Bennett MJ, Wells DM. Uncovering the hidden half of plants using new advances in root phenotyping. Curr Opin Biotechnol. 2019;55:1–8. doi: 10.1016/j.copbio.2018.06.002. PubMed DOI PMC

Mooney SJ, Pridmore TP, Helliwell J, Bennett MJ. Developing X-ray computed tomography to non-invasively image 3-D root systems architecture in soil. Plant Soil. 2012;352(1):1–22. doi: 10.1007/s11104-011-1039-9. DOI

Amato M, Basso B, Celano G, Bitella G, Morelli G, Rossi R. In situ detection of tree root distribution and biomass by multi-electrode resistivity imaging. Tree Physiol. 2008;28(10):1441–1448. doi: 10.1093/treephys/28.10.1441. PubMed DOI

Freschet GT, Pagès L, Iversen CM, Comas LH, Rewald B, Roumet C, et al. A starting guide to root ecology: strengthening ecological concepts and standardising root classification, sampling, processing and trait measurements. New Phytol. 2021;232(3):973–1122. doi: 10.1111/nph.17572. PubMed DOI PMC

Withington JM, Elkin AD, Bułaj B, Olesiński J, Tracy KN, Bouma TJ, et al. The impact of material used for minirhizotron tubes for root research. New Phytol. 2003;160(3):533–544. doi: 10.1046/j.1469-8137.2003.00903.x. PubMed DOI

Svane SF, Dam EB, Carstensen JM, Thorup-Kristensen K. A multispectral camera system for automated minirhizotron image analysis. Plant Soil. 2019;441(1–2):657–672. doi: 10.1007/s11104-019-04132-8. DOI

Rahman G, Sohag H, Chowdhury R, Wahid KA, Dinh A, Arcand M, et al. SoilCam: a fully automated minirhizotron using multispectral imaging for root activity monitoring. Sensors. 2020;20(3):787. doi: 10.3390/s20030787. PubMed DOI PMC

Arnaud M, Baird AJ, Morris PJ, Harris A, Huck JJ. EnRoot: a narrow-diameter, inexpensive and partially 3D-printable minirhizotron for imaging fine root production. Plant Methods. 2019;15(1):101. doi: 10.1186/s13007-019-0489-6. PubMed DOI PMC

Sell M, Smith AG, Burdun I, Rohula-Okunev G, Kupper P, Ostonen I. Assessing the fine root growth dynamics of Norway spruce manipulated by air humidity and soil nitrogen with deep learning segmentation of smartphone images. Plant Soil. 2022 doi: 10.1007/s11104-022-05565-4. DOI

Zeng G, Birchfield ST, Wells CE. Automatic discrimination of fine roots in minirhizotron images. New Phytol. 2008;177(2):549–557. doi: 10.1111/j.1469-8137.2007.02271.x. PubMed DOI

Möller B, Chen H, Schmidt T, Zieschank A, Patzak R, Türke M, et al. rhizoTrak: a flexible open source Fiji plugin for user-friendly manual annotation of time-series images from minirhizotrons. Plant Soil. 2019;444:519–534. doi: 10.1007/s11104-019-04199-3. DOI

Smith AG, Han E, Petersen J, Olsen NAF, Giese C, Athmann M, et al. RootPainter: deep learning segmentation of biological images with corrective annotation. New Phytol. 2022 doi: 10.1111/nph.18387. PubMed DOI PMC

Le Bot J, Serra V, Fabre J, Draye X, Adamowicz S, Pagès L. DART: a software to analyse root system architecture and development from captured images. Plant Soil. 2010;326(1):261–273. doi: 10.1007/s11104-009-0005-2. DOI

Lobet G, Pagès L, Draye X. A novel image-analysis toolbox enabling quantitative analysis of root system architecture. Plant Physiol. 2011;157(1):29–39. doi: 10.1104/pp.111.179895. PubMed DOI PMC

Armengaud P. EZ-Rhizo software: the gateway to root architecture analysis. Plant Signal Behav. 2009;4(2):139–141. doi: 10.4161/psb.4.2.7763. PubMed DOI PMC

Bucksch A, Burridge J, York LM, Das A, Nord E, Weitz JS, et al. Image-based high-throughput field phenotyping of crop roots. Plant Physiol. 2014;166(2):470–486. doi: 10.1104/pp.114.243519. PubMed DOI PMC

Galkovskyi T, Mileyko Y, Bucksch A, Moore B, Symonova O, Price CA, et al. GiA Roots: software for the high throughput analysis of plant root system architecture. BMC Plant Biol. 2012;12(1):116. doi: 10.1186/1471-2229-12-116. PubMed DOI PMC

Pound M, French A, Atkinson J, Wells D, Bennett M, Pridmore T. RootNav: navigating images of complex root architectures. Plant Physiol. 2013;162:1802–1814. doi: 10.1104/pp.113.221531. PubMed DOI PMC

Borianne P, Subsol G, Fallavier F, Dardou A, Audebert A. GT-RootS: An integrated software for automated root system measurement from high-throughput phenotyping platform images. Comput Electron Agric. 2018;150:328–342. doi: 10.1016/j.compag.2018.05.003. DOI

Narisetti N, Henke M, Seiler C, Shi R, Junker A, Altmann T, et al. Semi-automated root image analysis (saRIA) Sci Rep. 2019;9(1):19674. doi: 10.1038/s41598-019-55876-3. PubMed DOI PMC

Bradley D, Roth G. Adaptive thresholding using the integral image. Journal of graphics tools. 2007;12(2):13–21. doi: 10.1080/2151237X.2007.10129236. DOI

Frangi R, Niessen WJ, Vincken K, Viergever M. Multiscale vessel enhancement filtering. Med Image Comput Comput Assist Interv. 2000;1496:130–137.

Chapman BE, Parker D. An analysis of vessel enhancement filters based on the hessian matrix for intracranial MRA. Proc Soc Magn Reson Med 2001. p. 607.

Schulz H, Postma JA, van Dusschoten D, Scharr H, Behnke S. Plant root system analysis from MRI images. In: Csurka G, Kraus M, Laramee RS, Richard P, Braz J, editors. Computer vision, imaging and computer graphics theory and application. Berlin: Springer; 2013.

Smith A, Petersen J, Selvan R, Rasmussen C. Segmentation of roots in soil with U-Net. Plant Methods. 2020;16(1):1–15. doi: 10.1186/s13007-020-0563-0. PubMed DOI PMC

Yu G, Zare A, Sheng H, Matamala R, Reyes-Cabrera J, Fritschi FB, et al. Root identification in minirhizotron imagery with multiple instance learning. Mach Vis Appl. 2020;31(6):1–13. doi: 10.1007/s00138-020-01088-z. DOI

Hearst MA, Dumais ST, Osuna E, Platt J, Scholkopf B. Support vector machines. IEEE Intell Syst. 1998;13(4):18–28. doi: 10.1109/5254.708428. DOI

Dargan S, Kumar M, Ayyagari MR, Kumar G. A survey of deep learning and its applications: a new paradigm to machine learning. Arch Comput Methods Eng. 2020;27(4):1071–1092. doi: 10.1007/s11831-019-09344-w. DOI

Shen C, Liu L, Zhu L, Kang J, Wang N, Shao L. High-throughput in situ root image segmentation based on the improved deeplabv3+ method. Front Plant Sci. 2020;11:576791. doi: 10.3389/fpls.2020.576791. PubMed DOI PMC

Wang T, Rostamza M, Song Z, Wang L, McNickle G, Iyer-Pascuzzi AS, et al. SegRoot: a high throughput segmentation method for root image analysis. Comput Electron Agric. 2019;162:845–854. doi: 10.1016/j.compag.2019.05.017. DOI

Ward D, Moghadam P. Scalable learning for bridging the species gap in image-based plant phenotyping. Comput Vis Image Underst. 2020;197–198:103009. doi: 10.1016/j.cviu.2020.103009. DOI

Tan M, Le Q. Efficientnet: Rethinking model scaling for convolutional neural networks. Proceedings of the 36th International Conference on Machine Learning. 97. Long Beach: PMLR; 2019. p. 6105–14.

Hu J, Shen L, Sun G, editors. Squeeze-and-Excitation Networks. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition PP: 7132–7141 doi: 101109/CVPR201800745; 2018 18–23 June 2018.

Bianco S, Cadene R, Celona L, Napoletano P. Benchmark analysis of representative deep neural network architectures. IEEE Access. 2018;6:64270–64277. doi: 10.1109/ACCESS.2018.2877890. DOI

Huynh LD, Boutry N. A U-Net++ With Pre-Trained EfficientNet Backbone for Segmentation of Diseases and Artifacts in Endoscopy Images and Videos. EndoCV@ISBI. Iowa City, Iowa, USA.

Ronneberger O, Fischer P, Brox T. U-Net: convolutional networks for biomedical image segmentation. In: Navab N, Hornegger J, Wells WM, Frangi AF, editors. Medical Image computing and computer-assisted intervention—MICCAI 2015. Cham: Springer International Publishing; 2015. pp. 234–241.

Wang J, Zhang X, Lv P, Zhou L, Wang H. EAR-U-Net: EfficientNet and attention-based residual U-Net for automatic liver segmentation in CT. ArXiv. 2021. abs/2110.01014.

Baheti B, Innani S, Gajre S, Talbar S. Eff-UNet: A Novel architecture for semantic segmentation in unstructured environment. IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) 2020. p. 1473–81.

Shorten C, Khoshgoftaar TM. A survey on image data augmentation for deep learning. J Big Data. 2019;6(1):1–48. doi: 10.1186/s40537-019-0197-0. PubMed DOI PMC

Badrinarayanan V, Handa A, Cipolla R. Segnet: A deep convolutional encoder-decoder architecture for robust semantic pixel-wise labelling. arXiv preprint 2015: arXiv:1505.07293. PubMed

Luke J, Joseph R, Balaji M. Impact of image size on accuracy and generalization of convolutional neural networks. IJRAR. 2019;6(1):70–80.

Kimura K, Kikuchi S, Yamasaki S-i. Accurate root length measurement by image analysis. Plant Soil. 1999;216(1):117–27. doi: 10.1023/A:1004778925316. DOI

Tajbakhsh N, Jeyaseelan L, Li Q, Chiang JN, Wu Z, Ding X. Embracing imperfect datasets: a review of deep learning solutions for medical image segmentation. Med Image Anal. 2020;63:101693. doi: 10.1016/j.media.2020.101693. PubMed DOI

Pound MP, Atkinson JA, Townsend AJ, Wilson MH, Griffiths M, Jackson AS, et al. Deep machine learning provides state-of-the-art performance in image-based plant phenotyping. GigaScience. 2017;6(10):1–10. doi: 10.1093/gigascience/gix083. PubMed DOI PMC

Rangarajan H, Lynch JP. A comparative analysis of quantitative metrics of root architecture. Plant Phenomics. 2021;2021:6953197. doi: 10.34133/2021/6953197. PubMed DOI PMC

Wacker T, Popovic O, Olsen N, Markussen B, Smith A, Svane S, et al. Semi-field root phenotyping: root traits for deep nitrate uptake. Plant Cell Env. 2021;45:823–836. doi: 10.1111/pce.14227. PubMed DOI

Smith L. 2017. Cyclical Learning Rates for Training Neural Networks IEEE Winter on Applications of Computer Vision (WACV) Santa Rosa IEEE winter applications of computer vision. 10.1109/WACV.2017.58.

Smith L, Topin N. Super-convergence: very fast training of neural networks using large learning rates: SPIE; 2019.

Hamwood J, Alonso-Caneiro D, Read SA, Vincent SJ, Collins MJ. Effect of patch size and network architecture on a convolutional neural network approach for automatic segmentation of OCT retinal layers. Biomed Opt Express. 2018;9(7):3049–3066. doi: 10.1364/BOE.9.003049. PubMed DOI PMC

El-Madany TS, Reichstein M, Perez-Priego O, Carrara A, Moreno G, Pilar Martín M, et al. Drivers of spatio-temporal variability of carbon dioxide and energy fluxes in a Mediterranean savanna ecosystem. Agric For Meteorol. 2018;262:258–278. doi: 10.1016/j.agrformet.2018.07.010. DOI

Nair R, Strube M, Hertel M, Kolle O, Rolo V, Migliavacca M. High frequency root dynamics: sampling and interpretation using replicated robotic minirhizotrons. Journal of Experimental Botany . 2023;74(3):769. doi: 10.1093/jxb/erac427. PubMed DOI PMC

Audebert N, Le Saux B, Lefèvre S. Segment-before-detect: vehicle detection and classification through semantic segmentation of aerial images. Remote Sens. 2017;9(4):368. doi: 10.3390/rs9040368. DOI

Frangi A, Niessen WJ, Vincken K, Viergever M. Multiscale Vessel Enhancement Filtering. In: Wells WM, Colchester A, Delp S, editors. International on medical image computing and computer-assisted intervention. Berlin Heidelberg: Springer; 1998.

Yakubovskiy P. Segmentation models pytorch 2020. https://github.com/qubvel/segmentation_models.pytorch. Accessed 6 July 2022.

The BG, Library OpenCV. The openCV Library. Dr Dobb's J Software Tools. 2000;25(120):122–125.

Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, et al. Scikit-learn: machine learning in python. J Mach Learn Res. 2011;12:2825–2830.

Ruder S. An overview of gradient descent optimization algorithms. arXiv preprint 2016: arXiv:1609.04747.

He H, Yang D, Wang S, Wang S, Li Y. Road extraction by using atrous spatial pyramid pooling integrated encoder-decoder network and structural similarity loss. Remote Sens. 2019;11(9):1015. doi: 10.3390/rs11091015. DOI

Kingma DP, Ba J. Adam: A method for stochastic optimization. arXiv preprint 2014: arXiv:1412.6980.

Loshchilov I, Hutter F. Sgdr: Stochastic gradient descent with warm restarts. arXiv preprint 2016:arXiv:1608.03983.

Shamir RR, Duchin Y, Kim J, Sapiro G, Harel N. Continuous dice coefficient: a method for evaluating probabilistic segmentations. arXiv preprint 2019.

Zou KH, Warfield SK, Bharatha A, Tempany CM, Kaus MR, Haker SJ, et al. Statistical validation of image segmentation quality based on a spatial overlap index1: scientific reports. Acad Radiol. 2004;11(2):178–189. doi: 10.1016/S1076-6332(03)00671-8. PubMed DOI PMC

Wang Z, Bovik AC, Sheikh HR, Simoncelli EP. Image quality assessment: from error visibility to structural similarity. IEEE Trans Image Process. 2004;13(4):600–612. doi: 10.1109/TIP.2003.819861. PubMed DOI

Fawcett T. Introduction to ROC analysis. Pattern Recognit Lett. 2006;27:861–874. doi: 10.1016/j.patrec.2005.10.010. DOI

Van der Walt S, Schönberger J, Nunez-Iglesias J, Boulogne F, Warner J, Yager N, et al. scikit-image: image processing in python. PeerJ. 2014;2:e453. doi: 10.7717/peerj.453. PubMed DOI PMC