Stimulatory effect of monoacylglycerol lipase inhibitor MJN110 on locomotion and step kinematics demonstrated by high-precision 3D motion capture in mice

Ignatowska‐Jankowska, Bogna M.; Özer, Ayşen Gürkan; Kuck, Alexander; Niphakis, Micah J.; Ogasawara, Daisuke; Cravatt, Benjamin F.; Uusisaari, Marylka Yoe

doi:10.1101/2023.06.25.546437

“…We used a subset of the FL2 dataset including recordings of 10 mice performing the floor exploration task under 3 pharmacological conditions (drugs A, B and vehicle). As in the original analysis of this dataset [9], we detected steps on both sides (cf definition on Fig. 7a) and performed this detection before and after imputation.…”

Section: Imputed Data Enable Finer Analysis Of Drug-induced Differenc...mentioning

confidence: 99%

“…In the floor exploration task, step detection needs to differentiate between real locomotory steps and other leg movements. We therefore used kinematic rather than positional features of ankles as in [9] to detect steps. A fully-resolved step is defined as a velocity peak bordered with acceleration and deceleration events, reflecting take-off and touch-down (cf Fig.…”

Section: Imputed Data Enable Finer Analysis Of Drug-induced Differenc...mentioning

confidence: 99%

“…We tested the different neural network solutions on seven 3D kinematics datasets: two 3D mouse datasets -FL2, in which mice perform floor exploration and CLB, in which mice climb on the outisde of a mesh wheel [9] -, a 2D dataset with 2 mice (2-Mice-2D), a 3D Drosophila dataset (DF3D) [53], a 3D Human motion CMU motion capture dataset (Human) [54], a 3D rat dataset (Rat7M ) [4], and a 3D dataset with two Zebrafish (2-Fish) [31]. The 3D mouse datasets comprise recordings of 1 minute of 22 mouse individuals under pharmacological interventions affecting their behavior [9]. Each mouse is recorded freely behaving in two different set-ups: an open field exploration (FL2 dataset, 101 recordings) and a climbing vertical mesh (CLB dataset, 102 recordings).…”

Section: Datasetsmentioning

confidence: 99%

See 1 more Smart Citation

Deep Imputation for Skeleton Data (DISK) for Behavioral Science

Rose,

Michaluk,

Blindauer

et al. 2024

Preprint

Self Cite

0

View full text Add to dashboard Cite

Pose estimation methods and motion capture systems have opened doors to quantitative measurements of animal kinematics. However, these methods are not perfect and contain missing data. Our method, Deep Imputation for Skeleton data (DISK), leverages deep learning algorithms to learn dependencies between keypoints and their dynamics to impute missing tracking data. We developed an unsupervised training scheme, which does not rely on manual annotations, and tested several neural network architectures for the imputation task. We found that transformer outperforms other architectures including graph convolutional networks that were developed specifically for skeleton-based action recognition. We demonstrate the usability and performance of our imputation method on seven different animal skeletons including two multi-animal set-ups. With an optional estimated imputation error, DISK enables behavior scientists to assess the reliability of the imputed data. The imputed recordings allow to detect more episodes of motion, such as steps, and to obtain more statistically robust results when comparing these episodes between experimental conditions. While animal behavior experiments are expensive and complex, tracking errors make sometimes large portions of the experimental data unusable. DISK allows for filling in the missing information and for taking full advantage of the rich behavioral data. This stand-alone imputation package, freely available at https://github.com/bozeklab/DISK.git, is applicable to results of any tracking method (marker-based or markerless) and allows for any type of downstream analysis.

show abstract

“…We tested the different neural network solutions on seven 3D kinematics datasets: two 3D mouse datasets -FL2, in which mice perform floor exploration and CLB, in which mice climb on the outisde of a mesh wheel [9] -, a 2D dataset with 2 mice (2-Mice-2D), a 3D Drosophila dataset (DF3D) [53], a 3D Human motion CMU motion capture dataset (Human) [54], a 3D rat dataset (Rat7M ) [4], and a 3D dataset with two Zebrafish (2-Fish) [31].…”

Section: Datasetsmentioning

confidence: 99%

“…Marker-based techniques use motion capture cameras to directly assess the marker 3D positions and are typically more precise in determining keypoint locations [4,9,10].…”

Section: Introductionmentioning

confidence: 99%

Deep Imputation for Skeleton Data (DISK) for Behavioral Science

Rose,

Michaluk,

Blindauer

et al. 2024

Preprint

0

View full text Add to dashboard Cite

Pose estimation methods and motion capture systems have opened doors to quantitative measurements of animal kinematics. However, these methods are not perfect and contain missing data. Our method, Deep Imputation for Skeleton data (DISK), leverages deep learning algorithms to learn dependencies between keypoints and their dynamics to impute missing tracking data. We developed an unsupervised training scheme, which does not rely on manual annotations, and tested several neural network architectures for the imputation task. We found that transformer outperforms other architectures including graph convolutional networks that were developed specifically for skeleton-based action recognition. We demonstrate the usability and performance of our imputation method on seven different animal skeletons including two multi-animal set-ups. With an optional estimated imputation error, DISK enables behavior scientists to assess the reliability of the imputed data. The imputed recordings allow to detect more episodes of motion, such as steps, and to obtain more statistically robust results when comparing these episodes between experimental conditions. While animal behavior experiments are expensive and complex, tracking errors make sometimes large portions of the experimental data unusable. DISK allows for filling in the missing information and for taking full advantage of the rich behavioral data. This stand-alone imputation package, freely available at https://github.com/bozeklab/DISK.git, is applicable to results of any tracking method (marker-based or markerless) and allows for any type of downstream analysis.

show abstract

High-resolutionin vivokinematic tracking with injectable fluorescent nanoparticles

Ulutas,

Pradhan,

Koveal

et al. 2024

Preprint

0

View full text Add to dashboard Cite

Behavioral quantification is a cornerstone of many neuroscience experiments. Recent advances in motion tracking have streamlined the study of behavior in small laboratory animals and enabled precise movement quantification on fast (millisecond) timescales. This includes markerless keypoint trackers, which utilize deep network systems to label positions of interest on the surface of an animal (e.g., paws, snout, tail, etc.). These approaches mark a major technological achievement. However, they have a high error rate relative to motion capture in humans and are yet to be benchmarked against ground truth datasets in mice. Moreover, the extent to which they can be used to track joint or skeletal kinematics remains unclear. As the primary output of the motor system is the activation of muscles that, in turn, exert forces on the skeleton rather than the skin, it is important to establish potential limitations of techniques that rely on surface imaging. This can be accomplished by imaging implanted fiducial markers in freely moving mice. Here, we present a novel tracking method called QD-Pi (Quantum Dot-based Pose estimation in vivo), which employs injectable near-infrared fluorescent nanoparticles (quantum dots, QDs) immobilized on microbeads. We demonstrate that the resulting tags are biocompatible and can be imaged non-invasively using commercially available camera systems when injected into fatty tissue beneath the skin or directly into joints. Using this technique, we accurately capture 3D trajectories of up to ten independent internal positions in freely moving mice over multiple weeks. Finally, we leverage this technique to create a large-scale ground truth dataset for benchmarking and training the next generation of markerless keypoint tracker systems.

show abstract

Stimulatory effect of monoacylglycerol lipase inhibitor MJN110 on locomotion and step kinematics demonstrated by high-precision 3D motion capture in mice

Cited by 3 publications

References 71 publications

Deep Imputation for Skeleton Data (DISK) for Behavioral Science

Deep Imputation for Skeleton Data (DISK) for Behavioral Science

Deep Imputation for Skeleton Data (DISK) for Behavioral Science

High-resolutionin vivokinematic tracking with injectable fluorescent nanoparticles

Contact Info

Product

Resources

About