Final Report Summary - MECHANOSENSORYTREES (Role of Mechanosensory Touch-Based Cues on Arborization of Neuronal Dendritic Trees)
Caenorhabditis elegans (also known as C. elegans) is perhaps the best understood metazoan in terms of anatomy, genetics, development, and behavior. This transparent, free-living nematode has been widely used as a model organism ever since its introduction over 40 years ago by the Nobel laureate Sydney Brenner. In particular, C. elegans is commonly used to investigate fundamental questions in biology, including behavioral genetics, neuroscience, drug screening and development as well as modelling different aspects of human diseases among others.
Notably, this organism relies on mechanosensory functions to interact with its environment. Thus, C. elegans is an attractive candidate to study mechanosensation since its nervous system includes sensory neurons that are both morphologically and functionally similar to those in humans.To uncover the genetic basis of behavioral traits in C. elegans, a widespread strategy has been to study experimentally locomotion defects in mutants. In reverse genetics, strains with known mutations are phenotyped to determine whether or not the gene of interest has an effect on motility. Since traditional approaches to classifying patterns of C. elegans movement have often been based on manual annotation, motility phenotyping is often imprecise or qualitative, as well as time consuming. As a result, there is a necessity for algorithms that can automatically analyze and classify C. elegans motility phenotypes quantitatively. One ongoing bottleneck lies in reliably providing users with automatic segmentations from single images or sequences (i.e. videos) that feature complex and/or dynamic visual cues. Such computer vision tasks serve as a first and necessary step prior to extracting nematode phenotypes of interest, a necessary step in view of high-throughput motility assays.
In the present CIG framework, our goals have been aimed at further understanding the locomotive behavior of the model organism C. elegans in an effort to characterize altogether motility phenotyping traits of the influence. More broadly, this work fits into the scope of uncovering how touch-based sensory input determines dendritic patterns during development. Overall, our efforts have been geared at leveraging computer vision techniques, using statistical learning methods that rely on the use of both intensity- and texture-based image features extracted
using image filters or other image processing methods, integrated within a probabilistic framework. Our major deliverables over the course of the CIG are as follows.
First, we have have developed and validated an automatic segmentation strategy that combines the use of intensity, texture and temporal features within a probabilistic framework to deliver coarse segmentation approximations, which are then refined using a Markov Random Field (MRF) model. We have evaluated our approach on a number of well-established motility environments (i.e. crawling, swimming, etc.) and shown reliable segmentations for the purpose of key motility phenotypes extraction in both traditional and extremely challenging nematode essays. Overall, we have demonstrated that our segmentation approach on challenging environments provide significant performance gains compared to traditional methods (e.g. simple thresholding techniques).
Furthermore, we have developed a statistical computer vision method for segmenting dendritic tree structures from so-called Maximum Intensity Projection (MIP) images of C. elegans obtained using laser-scanning confocal microscopy technique. Our approach makes use of texture-based features that are invariant to orientation changes in an effort to characterize noisy tubular-like image patches. These features are then used in a probabilistic model that provides a coarse tree segmentation before further fine-tuning using post-processing steps. We have quantitatively shown that our method delivers reliable segmentations for various noisy MIP imaging scenarios and widely outperforms traditional intensity-based methods. In addition, we have shown that our method performs at least as well, if not better to more sophisticated methods when extracting the dendritic tree outline. These efforts were pursued to help unburden manual labor on the user end and to appeal to a growing community of researchers interested in characterization of neuronal arborization in C. elegans.
In a final step, we have developed a fully-automated approach to characterize C. elegans’ phenotypes that does not require the definition of nematode-specific features. Here, we have made use of the popular computer vision Scale-Invariant Feature Transform (SIFT) from which we construct histograms of commonly-observed SIFT features to represent nematode motility. Our method was first evaluated on a synthetic dataset simulating a range of nematode crawling gaits followed by two distinct datasets of crawling C. elegans with mutants affecting neuromuscular structure and function. Our algorithm was shown to detect differences between strains, results capture similarities in locomotory phenotypes that lead to clustering that is consistent with expectations based on genetic relationships.
Overall, our efforts have aimed at helping unburden manual labor on the user end (e.g. scientist) and to appeal to a growing community of researchers interested in characterization of neuronal arborization in C. elegans. Here, the ultimate end point is to provide experts/scientists with a software (i.e. Matlab) that can be used to conduct phenotyping analysis of C. elegans locomotion data.
A detailed list of the dissemination activities, including published journal articles and conference presentations are available at: http://biofluids.technion.ac.il(öffnet in neuem Fenster)
Notably, this organism relies on mechanosensory functions to interact with its environment. Thus, C. elegans is an attractive candidate to study mechanosensation since its nervous system includes sensory neurons that are both morphologically and functionally similar to those in humans.To uncover the genetic basis of behavioral traits in C. elegans, a widespread strategy has been to study experimentally locomotion defects in mutants. In reverse genetics, strains with known mutations are phenotyped to determine whether or not the gene of interest has an effect on motility. Since traditional approaches to classifying patterns of C. elegans movement have often been based on manual annotation, motility phenotyping is often imprecise or qualitative, as well as time consuming. As a result, there is a necessity for algorithms that can automatically analyze and classify C. elegans motility phenotypes quantitatively. One ongoing bottleneck lies in reliably providing users with automatic segmentations from single images or sequences (i.e. videos) that feature complex and/or dynamic visual cues. Such computer vision tasks serve as a first and necessary step prior to extracting nematode phenotypes of interest, a necessary step in view of high-throughput motility assays.
In the present CIG framework, our goals have been aimed at further understanding the locomotive behavior of the model organism C. elegans in an effort to characterize altogether motility phenotyping traits of the influence. More broadly, this work fits into the scope of uncovering how touch-based sensory input determines dendritic patterns during development. Overall, our efforts have been geared at leveraging computer vision techniques, using statistical learning methods that rely on the use of both intensity- and texture-based image features extracted
using image filters or other image processing methods, integrated within a probabilistic framework. Our major deliverables over the course of the CIG are as follows.
First, we have have developed and validated an automatic segmentation strategy that combines the use of intensity, texture and temporal features within a probabilistic framework to deliver coarse segmentation approximations, which are then refined using a Markov Random Field (MRF) model. We have evaluated our approach on a number of well-established motility environments (i.e. crawling, swimming, etc.) and shown reliable segmentations for the purpose of key motility phenotypes extraction in both traditional and extremely challenging nematode essays. Overall, we have demonstrated that our segmentation approach on challenging environments provide significant performance gains compared to traditional methods (e.g. simple thresholding techniques).
Furthermore, we have developed a statistical computer vision method for segmenting dendritic tree structures from so-called Maximum Intensity Projection (MIP) images of C. elegans obtained using laser-scanning confocal microscopy technique. Our approach makes use of texture-based features that are invariant to orientation changes in an effort to characterize noisy tubular-like image patches. These features are then used in a probabilistic model that provides a coarse tree segmentation before further fine-tuning using post-processing steps. We have quantitatively shown that our method delivers reliable segmentations for various noisy MIP imaging scenarios and widely outperforms traditional intensity-based methods. In addition, we have shown that our method performs at least as well, if not better to more sophisticated methods when extracting the dendritic tree outline. These efforts were pursued to help unburden manual labor on the user end and to appeal to a growing community of researchers interested in characterization of neuronal arborization in C. elegans.
In a final step, we have developed a fully-automated approach to characterize C. elegans’ phenotypes that does not require the definition of nematode-specific features. Here, we have made use of the popular computer vision Scale-Invariant Feature Transform (SIFT) from which we construct histograms of commonly-observed SIFT features to represent nematode motility. Our method was first evaluated on a synthetic dataset simulating a range of nematode crawling gaits followed by two distinct datasets of crawling C. elegans with mutants affecting neuromuscular structure and function. Our algorithm was shown to detect differences between strains, results capture similarities in locomotory phenotypes that lead to clustering that is consistent with expectations based on genetic relationships.
Overall, our efforts have aimed at helping unburden manual labor on the user end (e.g. scientist) and to appeal to a growing community of researchers interested in characterization of neuronal arborization in C. elegans. Here, the ultimate end point is to provide experts/scientists with a software (i.e. Matlab) that can be used to conduct phenotyping analysis of C. elegans locomotion data.
A detailed list of the dissemination activities, including published journal articles and conference presentations are available at: http://biofluids.technion.ac.il(öffnet in neuem Fenster)