cellprofiler worm toolbox python source code

Cellprofiler worm toolbox python source code windows#

The Hessian and Laplacian segmentation largely avoided false positive predictions but instead often failed to recognize mitochondria, resulting in false negative segmentations in the elongated, mixed, and tubular phenotype (and fragmented for the Laplacian segmentation) ( Figure 2). The Gaussian segmentation produced large sections of false positive predictions in the mixed and tubular phenotype (indicated by yellow arrows).

The Gaussian, Hessian, Laplacian, and Ilastik methods failed to consistently prevent false positive and/or false negative segmentation on all phenotypes ( Figure 2). A cross validation was performed to estimate the performance of the MitoSegNet on an unseen test set and to compare it against other segmentation methods ( Figure 1B). For each tile, 80 augmented copies were generated for training the model.

Each image was split into 4 overlapping tiles. This modification decreased the amount of necessary training time. Our U-Net modification entails the removal of dropout layers at the end of the contracting pathway and instead placing batch normalization layers after every convolutional layer prior to ReLU activation in the contracting pathway. The MitoSegNet model was generated by training a modified U-Net with a training set of 12 1300 × 1030 pixel fluorescent microscopy, maximum-intensity projection images, depicting mitochondria in body wall muscle cells of adult C. elegans worms (mitochondria were visualized using a transgene expressing mitochondrial matrix-targeted GFP under the control of a body wall muscle-specific promoter ( P myo3 ::mitoGFP)) ( Figure 1A and Methods). To analyze mitochondrial morphology, for example, in C. elegans, mitochondria are labeled using either a mitochondria-specific fluorescent dye (such as TMRE) or a transgene expressing a mitochondrial-targeted GFP (mitoGFP) ( As a result, subtle differences in morphology and, hence, phenotypes are often not detected. For this reason, researchers often resorted to the use of a simple qualitative assessment of mitochondrial morphology. Specifically, the diversity of shapes among mitochondria (elongated, fragmented, tubular, as well as ‘mixed’ morphologies) poses a challenge to the automated quantification of mitochondrial morphology. However, such studies have been hindered by the fact that is difficult to assess mitochondrial morphology in different genetic backgrounds or physiological conditions in an unbiased and quantitative manner. For this reason, understanding mitochondrial fusion and fission is not only an important basic biological question but is critical for our ability to understand the pathology of these diseases and to develop novel therapeutics to treat them.

Cellprofiler worm toolbox python source code windows#

We provide MitoSegNet in a toolbox for Windows and Linux operating systems that combines segmentation with morphological analysis. Additionally, MitoSegNet was capable of accurately segmenting mitochondria in HeLa cells treated with fragmentation inducing reagents. We successfully applied MitoSegNet to unseen fluorescence microscopy images of mitoGFP expressing mitochondria in wild-type and catp-6 ATP13A2 mutant C. elegans adults. MitoSegNet outperformed all other methods in both pixelwise and morphological segmentation accuracy. We tested the performance of MitoSegNet against three feature-based segmentation algorithms and the machine-learning segmentation tool Ilastik. Here, we present Mitochondrial Segmentation Network (MitoSegNet), a pretrained deep learning segmentation model that enables researchers to easily exploit the power of deep learning for the quantification of mitochondrial morphology. While the analysis of mitochondrial morphology has emerged as a key tool in the study of mitochondrial function, efficient quantification of mitochondrial microscopy images presents a challenging task and bottleneck for statistically robust conclusions.