Trainable Weka Segmentation: a machine learning tool for microscopy pixel classification

Arganda‐Carreras, Ignacio; Kaynig, Verena; Rueden, Curtis; Eliceiri, Kevin W.; Schindelin, Johannes; Cardona, Albert; Seung, H. Sebastian

doi:10.1093/bioinformatics/btx180

Cited by 1,844 publications

(1,375 citation statements)

References 5 publications

Supporting

Mentioning

1,365

Contrasting

Unclassified

Order By: Relevance

“…We use two publicly available datasets for our experiments, which we refer to as DROSOPHILA [9], which consists of two stacks of 20 EM sections with 4×4×40nm resolution (1024× 1024 × 20 pixels), and MOUSE CORTEX [10], which consists of two stacks of 100 EM sections with 6×6×30nm resolution (1024 × 1024 × 100 pixels). We split the parts for which ground truth was available into two stacks of equal size (2×10 sections for DROSOPHILA and 2 × 50 sections for MOUSE CORTEX).…”

Section: Resultsmentioning

confidence: 99%

Structured learning of assignment models for neuron reconstruction to minimize topological errors

Funke

Klein

Moreno-Noguer

et al. 2016

2016 IEEE 13th International Symposium on Biomedical Imaging (ISBI)

Self Cite

View full text Add to dashboard Cite

Structured learning provides a powerful framework for empirical risk minimization on the predictions of structured models. It allows end-to-end learning of model parameters to minimize an application specific loss function. This framework is particularly well suited for discrete optimization models that are used for neuron reconstruction from anisotropic electron microscopy (EM) volumes. However, current methods are still learning unary potentials by training a classifier that is agnostic about the model it is used in. We believe the reason for that lies in the difficulties of (1) finding a representative training sample, and (2) designing an application specific loss function that captures the quality of a proposed solution. In this paper, we show how to find a representative training sample from human generated ground truth, and propose a loss function that is suitable to minimize topological errors in the reconstruction. We compare different training methods on two challenging EM-datasets. Our structured learning approach shows consistently higher reconstruction accuracy than other current learning methods.

show abstract

Section: Resultsmentioning

confidence: 99%

Structured learning of assignment models for neuron reconstruction to minimize topological errors

Funke

Klein

Moreno-Noguer

et al. 2016

2016 IEEE 13th International Symposium on Biomedical Imaging (ISBI)

Self Cite

View full text Add to dashboard Cite

show abstract

“…The percentage shrub canopy and intershrub area was calculated for the five sites using high-resolution satellite imagery from Google Earth (2016d-e), which were cropped to the same dimensions and segmented using the Trainable Weka Segmentation plugin for Fiji (http://fiji.sc/Fiji). Sobel and Gaussian blur filters were used for edge detection and noise suppression, respectively (Arganda-Carreras et al, 2017). The out-of-bag error for all sites classified was <2%.…”

Section: Study Site Descriptionmentioning

confidence: 99%

Soil nitrogen response to shrub encroachment in a degrading semiarid grassland

Turpin-Jelfs¹,

Michaelides²,

Anesio³

2018

Preprint

View full text Add to dashboard Cite

Abstract. Transitions from grass- to shrub-dominated states in drylands by woody plant encroachment represent significant forms of land cover change with the potential to alter the spatial distribution and cycling of soil resources. Yet an understanding of how this phenomenon impacts the soil nitrogen pool, which is essential to primary production in arid and semiarid systems, is poorly resolved. In this study, we quantified how the distribution and speciation of soil nitrogen, as well as rates of free-living biological nitrogen fixation, changed along a gradient of increasing mesquite (Prosopis velutina Woot.) cover in a semiarid grassland of the Southwestern US. Our results show that site-level concentrations of total nitrogen remain unchanged with increasing shrub cover as losses from intershrub areas (sum of grass and bare-soil cover) are proportional to increases in soils under shrub canopies. However, despite the similar carbon-to-nitrogen ratio and microbial biomass of soil from intershrub and shrub areas at each site, site-level concentrations of inorganic nitrogen increase with shrub cover due to the accumulation of ammonium and nitrate in soils beneath shrub canopies. Using the acetylene reduction assay technique, we found increasing ratios of inorganic nitrogen-to-bioavailable phosphorus inhibit rates of biological nitrogen fixation by free-living soil bacteria. Consequently, we conclude that shrub encroachment has the potential to significantly alter the dynamics of soil nitrogen cycling in dryland systems.

show abstract

“…ImageSURF classifiers are built from the training input using an optimised implementation of the random forests algorithm [8] adapted from the FastRandomForest [9] plugin for the WEKA environment [10], which is the default classifier used by Trainable Segmentation [6,7]. Features are stored as size-efficient primitive data structures (byte or short arrays for 8-bit and 16-bit images, respectively) and may be pre-calculated and saved to disk.…”

Section: Implementation and Architecturementioning

confidence: 99%

“…Open-source trainable segmentation tools, such as Ilastik [3] or the Trainable Segmentation [6] plugin for ImageJ [7], address many of these issues by using supervised machine learning algorithms to study 'training set' of pixels, which have been manually assigned class annotations, and create a model (' classifier') to reliably discriminate between these classes. The context of each pixel (e.g., intensity, texture, edges, entropy) can be considered, making the classifier more robust to image artifacts and intensity shifts [3].…”

Section: Introductionmentioning

confidence: 99%

ImageSURF: An ImageJ Plugin for Batch Pixel-Based Image Segmentation Using Random Forests

O'Mara

King

Vickers

et al. 2017

JORS

View full text Add to dashboard Cite

Image segmentation is a necessary step in automated quantitative imaging. ImageSURF is a macro-compatible ImageJ2/FIJI plugin for pixel-based image segmentation that considers a range of image derivatives to train pixel classifiers which are then applied to image sets of any size to produce segmentations without bias in a consistent, transparent and reproducible manner. The plugin is available from ImageJ update site

show abstract

Trainable Weka Segmentation: a machine learning tool for microscopy pixel classification

Abstract: Supplementary data are available at Bioinformatics online.

Cited by 1,844 publications

References 5 publications

Structured learning of assignment models for neuron reconstruction to minimize topological errors

Structured learning of assignment models for neuron reconstruction to minimize topological errors

Soil nitrogen response to shrub encroachment in a degrading semiarid grassland

ImageSURF: An ImageJ Plugin for Batch Pixel-Based Image Segmentation Using Random Forests

Contact Info

Product

Resources

About