FlyLimbTracker: An active contour based approach for leg segment tracking in unmarked, freely behaving Drosophila

Uhlmann, Virginie; Ramdya, Pavan; Delgado-Gonzalo, Ricard; Benton, Richard; Unser, Michaël

doi:10.1371/journal.pone.0173433

Cited by 38 publications

(30 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…FLLIT does not require a contact surface for detection 19,20 or leg markers for tracking 21 , is not sensitive to variations in recording parameters and is not rule-based 23 . These strengths permit its application to other animals, which we show by application to spiders.…”

Section: Discussionmentioning

confidence: 99%

“…State-of-the-art methods include foot-printing-based approaches that report contact points with a detection surface 19,20 , and leg marking-based techniques that track distinct marked spots on the legs 21 . Semi-automated algorithms have been developed to aid high-resolution leg tracking in freely-moving, unmarked flies [22][23][24] , but these require a considerable degree of user annotation and/or user-led optimisation. Therefore, these methods were not feasible for use on the large volume of data required to quantify rapid and fine tremors in suspended legs.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Automated leg tracking reveals distinct conserved gait and tremor signatures inDrosophilamodels of Parkinson’s Disease and Spinocerebellar ataxia 3

Tan

Govindarajan

et al. 2018

Preprint

View full text Add to dashboard Cite

Genetic models in Drosophila have made invaluable contributions to our understanding of the molecular mechanisms underlying neurodegeneration. In human patients, some neurodegenerative diseases lead to characteristic movement dysfunctions, such as abnormal gait and tremors. However, it is currently unknown whether similar movement defects occur in the respective fly models, which could be used to model and better understand the pathophysiology of movement disorders. To address this question, we developed a machine-learning image-analysis programme -Feature Learning-based LImb segmentation and Tracking (FLLIT) -that automatically tracks leg claw positions of freely moving flies recorded on highspeed video, generating a series of body and leg movement parameters. Of note, FLLIT requires no user input for learning. We used FLLIT to characterise fly models of Parkinson's Disease (PD) and Spinocerebellar ataxia 3 (SCA3).Between these models, walking gait and tremor characteristics differed markedly, and recapitulated signatures of the respective human diseases.Selective expression of mutant SCA3 in dopaminergic neurons led to phenotypes resembling that of PD flies, suggesting that the behavioural phenotype may depend on the circuits affected, rather than the specific nature of the mutation. Different mutations produced tremors in distinct leg pairs, indicating that different motor circuits are affected. Almost 190,000 video frames were tracked in this study, allowing, for the first time, high-throughput analysis of gait and tremor features in Drosophila mutants. As an efficient assay of mutant gait and tremor features in an important model system, FLLIT will enable the analysis of the neurogenetic mechanisms that underlie movement disorders.(n = 10), mir-263a KO (n = 11), ple-Gal4>SCA3-flQ27 (n = 14), ple-Gal4>SCA3-flQ84 (n = 15). P values were calculated using a non-parametric Mann-Whitney test except for park 1 and mir-263a KO which shared the same control (yw), hence, P was calculated using a non-parametric Kruskal-Wallis test with Dunn's multiple comparisons post-hoc test (See Fig. S5).See also Videos 2 -6.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Automated leg tracking reveals distinct conserved gait and tremor signatures inDrosophilamodels of Parkinson’s Disease and Spinocerebellar ataxia 3

Tan

Govindarajan

et al. 2018

Preprint

View full text Add to dashboard Cite

show abstract

“…Animal pose estimation. The proposed approach fills a void between state of the art human pose estimation algorithms, which often rely on large quantities of manually labeled samples (see [9] for a recent review), and their counterparts in animal pose estimation [37,4,6,5,38,39]. Among these animal pose estimation algorithms, Deep Lab Cut (DLC) [4], Leap Estimates Animal Pose (LEAP) [6], and Deep Pose Kit (DPK) [5], stand out as they can achieve near human-level accuracy using a modest number of labels.…”

Section: S1 Related Workmentioning

confidence: 99%

Deep Graph Pose: a semi-supervised deep graphical model for improved animal pose tracking

Buchanan

Whiteway

et al. 2020

Preprint

View full text Add to dashboard Cite

Noninvasive behavioral tracking of animals is crucial for many scientific investigations. Recent transfer learning approaches for behavioral tracking have considerably advanced the state of the art. Typically these methods treat each video frame and each object to be tracked independently. In this work, we improve on these methods (particularly in the regime of few training labels) by leveraging the rich spatiotemporal structures pervasive in behavioral video --- specifically, the spatial statistics imposed by physical constraints (e.g., paw to elbow distance), and the temporal statistics imposed by smoothness from frame to frame. We propose a probabilistic graphical model built on top of deep neural networks, Deep Graph Pose (DGP), to leverage these useful spatial and temporal constraints, and develop an efficient structured variational approach to perform inference in this model. The resulting semi-supervised model exploits both labeled and unlabeled frames to achieve significantly more accurate and robust tracking while requiring users to label fewer training frames. In turn, these tracking improvements enhance performance on downstream applications, including robust unsupervised segmentation of behavioral "syllables," and estimation of interpretable "disentangled" low-dimensional representations of the full behavioral video.

show abstract

“…Due to the challenges associated with 3D pose estimation, most studies have favored the simplicity and higher throughput afforded by 2D pose estimation using one camera and a single viewpoint [18,19,6,20,10,5]. However, projected 2D poses can result from multiple distinct 3D poses, and thus true 3D joint configurations remain unknown.…”

Section: Introductionmentioning

confidence: 99%

“…Third, we apply pretrained LiftPose3D networks to predict realistic 3D poses from completely different experimental systems including from a previously published dataset consisting of a single viewpoint behavioral video [19]. Thus, we can effectively resurrect old data for new kinds of kinematic analyses.…”

mentioning

confidence: 99%

LiftPose3D, a deep learning-based approach for transforming 2D to 3D pose in laboratory animals

Gosztolai¹,

Günel²,

Abrate³

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

Markerless 3D pose estimation has become an indispensable tool for kinematic studies of laboratory animals. Most current methods recover 3D pose by multi-view triangulation of deep network-based 2D pose estimates. However, triangulation requires multiple, synchronised cameras per keypoint and elaborate calibration protocols that hinder its widespread adoption in laboratory studies. Here, we describe LiftPose3D, a deep network-based method that overcomes these barriers by reconstructing 3D poses from a single 2D camera view. We illustrate LiftPose3D’s versatility by applying it to multiple experimental systems using flies, mice, and macaque monkeys and in circumstances where 3D triangulation is impractical or impossible. Thus, LiftPose3D permits high-quality 3D pose estimation in the absence of complex camera arrays, tedious calibration procedures, and despite occluded keypoints in freely behaving animals.

show abstract

FlyLimbTracker: An active contour based approach for leg segment tracking in unmarked, freely behaving Drosophila

Cited by 38 publications

References 35 publications

Automated leg tracking reveals distinct conserved gait and tremor signatures inDrosophilamodels of Parkinson’s Disease and Spinocerebellar ataxia 3

Automated leg tracking reveals distinct conserved gait and tremor signatures inDrosophilamodels of Parkinson’s Disease and Spinocerebellar ataxia 3

Deep Graph Pose: a semi-supervised deep graphical model for improved animal pose tracking

LiftPose3D, a deep learning-based approach for transforming 2D to 3D pose in laboratory animals

Contact Info

Product

Resources

About