Automation of tree-ring detection and measurements using deep learning

Poláček, None Miroslav; Arizpe, Alexis; Hüther, Patrick; Weidlich, None Lisa; Steindl, None Sonja; Swarts, Kelly

doi:10.1101/2022.01.10.475709

Cited by 4 publications

(4 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this way, models achieve remarkable results when applied to macroscopic samples. Higher widening of tree ring border for model evaluation has been proposed as an advance (Polaćěk et al, 2023) to a finer analysis of segmentations through spatial overlap-based metrics (Taha & Hanbury, 2015).…”

Section: Discussionmentioning

confidence: 99%

Tree ring segmentation using UNEt TRansformer neural network on stained microsections for quantitative wood anatomy

García-Hidalgo,

García-Pedrero,

Rozas

et al. 2024

Front. Plant Sci.

View full text Add to dashboard Cite

Forests are critical in the terrestrial carbon cycle, and the knowledge of their response to ongoing climate change will be crucial for determining future carbon fluxes and climate trajectories. In areas with contrasting seasons, trees form discrete annual rings that can be assigned to calendar years, allowing to extract valuable information about how trees respond to the environment. The anatomical structure of wood provides highly-resolved information about the reaction and adaptation of trees to climate. Quantitative wood anatomy helps to retrieve this information by measuring wood at the cellular level using high-resolution images of wood micro-sections. However, whereas large advances have been made in identifying cellular structures, obtaining meaningful cellular information is still hampered by the correct annual tree ring delimitation on the images. This is a time-consuming task that requires experienced operators to manually delimit ring boundaries. Classic methods of automatic segmentation based on pixel values are being replaced by new approaches using neural networks which are capable of distinguishing structures, even when demarcations require a high level of expertise. Although neural networks have been used for tree ring segmentation on macroscopic images of wood, the complexity of cell patterns in stained microsections of broadleaved species requires adaptive models to accurately accomplish this task. We present an automatic tree ring boundary delineation using neural networks on stained cross-sectional microsection images from beech cores. We trained a UNETR, a combined neural network of UNET and the attention mechanisms of Visual Transformers, to automatically segment annual ring boundaries. Its accuracy was evaluated considering discrepancies with manual segmentation and the consequences of disparity for the goals of quantitative wood anatomy analyses. In most cases (91.8%), automatic segmentation matched or improved manual segmentation, and the rate of vessels assignment to annual rings was similar between the two categories, even when manual segmentation was considered better. The application of convolutional neural networks-based models outperforms human operator segmentations when confronting ring boundary delimitation using specific parameters for quantitative wood anatomy analysis. Current advances on segmentation models may reduce the cost of massive and accurate data collection for quantitative wood anatomy.

show abstract

Section: Discussionmentioning

confidence: 99%

Tree ring segmentation using UNEt TRansformer neural network on stained microsections for quantitative wood anatomy

García-Hidalgo,

García-Pedrero,

Rozas

et al. 2024

Front. Plant Sci.

View full text Add to dashboard Cite

show abstract

“…Full code that was used for preparation of this application is also available free and open source under the MIT licence at GitHub (https://github.com/Grego r-Mende l-Insti tute/TRG-Image Proce ssing) and version v1.0.2. associated with this article is archived at Zenodo (Poláček et al, 2023a) as well as all datasets used in preparation of this article (Poláček et al, 2023b), including training and validation dataset and their annotation files.…”

Section: Co N Fli C T O F I Nte R E S T S Tate M E Ntmentioning

confidence: 99%

Automation of tree‐ring detection and measurements using deep learning

et al. 2023

Self Cite

View full text Add to dashboard Cite

Core samples from trees are a critical reservoir of ecological information, informing our understanding of past climates, as well as contemporary ecosystem responses to global change. Manual measurements of annual growth rings in trees are slow, labour‐intensive and subject to human bias, hindering the generation of big datasets. We present an alternative, neural network‐based implementation that automates detection and measurement of tree‐ring boundaries from coniferous species. We trained our Mask R‐CNN extensively on over 8000 manually annotated ring boundaries from microscope‐imaged Norway Spruce Picea abies increment cores. We assessed the performance of the trained model after post‐processing on real‐world data generated from our core processing pipeline. The CNN after post‐processing performed well, with recognition of over 98% of ring boundaries (recall) with a precision in detection of 96% when tested on real‐world data. Additionally, we have implemented automatic measurements based on minimum distance between rings. With minimal editing for missed ring detections, these measurements were 98% correlated with human measurements of the same samples. Tests on other three conifer species demonstrate that the CNN generalizes well to other species with similar structure. We demonstrate the efficacy of automating the measurement of growth increment in tree core samples. Our CNN‐based system provides high predictive performance in terms of both tree‐ring detection and growth rate determination. Our application is readily deployable as a Docker container and requires only basic command line skills. Additionally, an easy re‐training option allows users to expand capabilities to other wood types. Application outputs include both editable annotations of predictions as well as ring‐width measurements in a commonly used .pos format, facilitating the efficient generation of large ring‐width measurement datasets from increment core samples, an important source of environmental data.

show abstract

“…In this work, the authors compared the detection of tree ring limits on macroscopic samples, but did not evaluate the discrepancies between both methods of segmentations. Poláček et al (2022) proposed advances in model evaluation by widening the drawn line to transform it in a polygon and comparing the filled areas.…”

Section: Discussionmentioning

confidence: 99%

“…CNN methods reach noteworthy ring detection rate for macroscopic samples . Further developments on ring detection and model validation for dendrochronological samples are recently handled considering ring borders as segments rather than polylines (Poláček et al, 2022). Despite these approaches in automatic ring segmentation and evaluation of models for dendrochronological samples, QWA is focused on cellular and anatomical characteristics of tree rings, requiring specific segmentation models and robust methods of model evaluation.…”

Section: Introductionmentioning

confidence: 99%

Open tools for dendrochronology. Advances in sample digitization and deep learning methods for image segmentation

Hidalgo¹

View full text Add to dashboard Cite

show abstract

Automation of tree-ring detection and measurements using deep learning

Cited by 4 publications

References 31 publications

Tree ring segmentation using UNEt TRansformer neural network on stained microsections for quantitative wood anatomy

Tree ring segmentation using UNEt TRansformer neural network on stained microsections for quantitative wood anatomy

Automation of tree‐ring detection and measurements using deep learning

Open tools for dendrochronology. Advances in sample digitization and deep learning methods for image segmentation

Contact Info

Product

Resources

About