Fast and Efficient Root Phenotyping via Pose Estimation

  1. Elizabeth M. Berrigan
  2. Lin Wang
  3. Hannah Carrillo
  4. Kimberly Echegoyen
  5. Mikayla Kappes
  6. Jorge Torres
  7. Angel Ai-Perreira
  8. Erica McCoy
  9. Emily Shane
  10. Charles D. Copeland
  11. Lauren Ragel
  12. Charidimos Georgousakis
  13. Sanghwa Lee
  14. Dawn Reynolds
  15. Avery Talgo
  16. Juan Gonzalez
  17. Ling Zhang
  18. Ashish B. Rajurkar
  19. Michel Ruiz
  20. Erin Daniels
  21. Liezl Maree
  22. Shree Pariyar
  23. Wolfgang Busch
  24. Talmo D. Pereira

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Image segmentation is commonly used to estimate the location and shape of plants and their external structures. Segmentation masks are then used to localize landmarks of interest and compute other geometric features that correspond to the plant’s phenotype. Despite its prevalence, segmentation-based approaches are laborious (requiring extensive annotation to train) and error-prone (derived geometric features are sensitive to instance mask integrity). Here, we present a segmentation-free approach that leverages deep learning-based landmark detection and grouping, also known as pose estimation. We use a tool originally developed for animal motion capture called SLEAP (Social LEAP Estimates Animal Poses) to automate the detection of distinct morphological landmarks on plant roots. Using a gel cylinder imaging system across multiple species, we show that our approach can reliably and efficiently recover root system topology at high accuracy, few annotated samples, and faster speed than segmentation-based approaches. In order to make use of this landmark-based representation for root phenotyping, we developed a Python library ( sleap-roots ) for trait extraction directly comparable to existing segmentation-based analysis software. We show that pose-derived root traits are highly accurate and can be used for common downstream tasks including genotype classification and unsupervised trait mapping. Altogether, this work establishes the validity and advantages of pose estimation-based plant phenotyping. To facilitate adoption of this easy-to-use tool and to encourage further development, we make sleap-roots , all training data, models, and trait extraction code available at: https://github.com/talmolab/sleap-roots and https://osf.io/k7j9g/ .

Article activity feed

  1. Version published to 10.34133/plantphenomics.0175
  2. PREreview

    This Zenodo record is a permanently preserved version of a Structured PREreview. You can view the complete PREreview at https://prereview.org/reviews/10400528.

    Does the introduction explain the objective of the research presented in the preprint? Yes
    Are the methods well-suited for this research? Highly appropriate The paper describes a new method for root image analysis, applying state-of-the-art deep learning methods. A few suggestions / questions related to the methods : - the imaging setup described in the manuscript produce, for each plant and each time point, 75 frame at different angles. My understanding is that each frame is then analysed independently and data aggregated afterward. My question is therefore how strongly does the presented pipeline rely on the 75-frame input ? Would it work with only one image per time point ? In other words, how much of the whole pipeline accuracy could be attributed to the imaging setup (75 frames) or the image analysis pipeline (pose estimation)? - can the image analysis pipeline take into account time series information? For instance, can you follow specific roots over time as they grow? - Preaching for my own church here, but did you consider making the output of the annotation pipeline compatible with the Root System Markup Language format (info : I am an author of the RSML paper). This would allow compatibilities with other existing image analysis tools, data analysis pipeline (archiDART) or computational root models. It seems to me it the current data structure should be compatible. I would also like to highlight that the authors did a very good job making all the data and code available through public repositories.
    Are the conclusions supported by the data? Somewhat supported The results are very well presented and analysed. In particular, the results (both accuracy and speed) are also well compared to one other existing tools (RootPainter + RhizoVisio) and ground-truth data. In that regard, the conclusion of the paper (faster and reliable image analysis pipeline) are very well supported. However, I have the feeling that not enough emphasis is given to the fact that the paper describes an image analysis pipeline developed only for a very specific plant culture and image acquisition setup (gel containers on turntables). As such, the paper does not describe a complete new phenotyping method for roots, just the image analysis pipeline. This is perfectly fine, as the image analysis is the current bottleneck in root phenotyping, but it shoudl be more explicit. In addition, the image analysis was tested on 1 type of image only (again, from the gel containers) in the paper, which limits the extrapolations to other setups. Personally, I have the feeling it could be successfully used with other images, such as the ones coming from aeroponic experiments or even shovelomics data, but this has to be tested. The current paper could be a bit more explicit about this.
    Are the data presentations, including visualizations, well-suited to represent the data? Highly appropriate and clear Very high quality figures and visualisations. Nothing to say, very well done!
    How clearly do the authors discuss, explain, and interpret their findings and potential next steps for the research? Somewhat clearly See comments on the "Conclusion"
    Is the preprint likely to advance academic knowledge? Somewhat likely The proposed method is a much needed improvement on current root image analysis methods. However, it should be tested wether it could be applied to images from different setup, which would highly increase its potential use in the community
    Would it benefit from language editing? No Already very good :)
    Would you recommend this preprint to others? Yes, it's of high quality The paper is very well structured and written, the figures of high quality. The presented image analysis pipeline is very interesting and well describe, although, as mentioned in a previous section, its use might be restricted only to a limited number of experimental setup.
    Is it ready for attention from an editor, publisher or broader audience? Yes, after minor changes I would suggest the authors to adjust the title, abstract and discussion to match more closely the actual scope of the manuscript. For instance, I would suggest to focus on "image analysis" rather than "phenotyping" and mention more clearly the fact that it is design for young plants grown in a specific experimental setup.

    Competing interests

    The author declares that they have no competing interests.

  3. Version published to 10.1101/2023.11.20.567949v1 on bioRxiv