A Look Back: Celebrating🎉 2023’s 10 Awesome DeepLabCut Papers

Created by Dall-E and ChatGPT4.

As we begin 2024, let’s take a moment to reflect on and applaud👏 the incredible strides made in research using DeepLabCut🐭 during the past year. In 2023, our community has flourished with innovative studies, and we’ve selected ten exceptional papers that highlight the versatile applications of DeepLabCut in scientific research.

Join us as we journey through these amazing discoveries, showcasing the dedication and collaborative spirit of researchers👩‍🔬🧑‍🔬 worldwide. Each paper is a shining example of how creativity and scientific inquiry come together to advance our understanding across various fields.

Get ready to explore these fascinating stories that have made 2023 a year of remarkable scientific achievements🚀. Cheers to the power of collaboration, discovery, and the exciting prospects that await us in the new year! 🍾🔬

1. Fast Prediction in Marmoset Reach-to-Grasp Movements for Dynamic Prey

📚Shaw, L., Wang, K. H., & Mitchell, J. (2023). Fast prediction in marmoset reach-to-grasp movements for dynamic prey. Current Biology, 33(12). https://doi.org/10.1016/j.cub.2023.05.032

• This study focuses on the sophisticated, visually guided reaching behaviors of marmosets, specifically their ability to predict and grasp moving prey like insects.

Imagine trying to catch a swiftly moving cricket🦗 with just your hands — sounds challenging, right? That’s what marmosets do, and they’re astonishingly good at it! Researchers have unlocked some fascinating insights into these primates’ lightning-fast reflexes.

Utilizing an innovative reach-to-grasp task with live crickets in a laboratory setting, the research leverages high-speed video cameras and machine vision algorithms for marker-free tracking of both marmosets and crickets.

The Discovery: Faster than a Blink

The study finds that marmosets can operate with incredibly short visuo-motor delays, indicating a critical role of visual prediction in facilitating rapid movement adjustments during prey capture.

Μarmosets can predict and react to the crickets’ movements in just 80 milliseconds. That’s less than a tenth of a second — faster than we blink!

• DeepLabCut played a pivotal role in accurately tracking the complex and dynamic movements of marmosets and crickets, allowing for detailed analysis of their interaction.

High-speed, multi-camera video reconstructs reaching for dynamic prey.

This isn’t just about marmosets being quick. It’s about understanding the brain’s amazing ability to predict and react. By studying these little acrobats, scientists are uncovering secrets of movement and perception that can help in areas like robotics and even medicine.

2. Multilevel dynamic adjustments of geckos (Hemidactylus frenatus) climbing vertically: head-up versus head-down

📚 Schultz, J. T., Labonte, D., & Clemente, C. J. (2023). Multilevel dynamic adjustments of geckos ( hemidactylus frenatus ) climbing vertically: Head-up versus head-down. Journal of The Royal Society Interface, 20(201). https://doi.org/10.1098/rsif.2022.08

• This study focuses on the intricate dynamics of gecko climbing. The researchers explore how these animals, equipped with direction-dependent adhesives, navigate the challenges of changing force orientations while climbing vertically in both upward and downward directions.

Imagine a tiny gecko, not just scampering across a flat surface but climbing a vertical wall — both upwards and downwards!

The researchers used a vertical racetrack with high-speed cameras for dorsal and lateral views, and a force plate integrated into the track to collect ground reaction forces, providing detailed insights into the geckos’ climbing dynamics.

Limbs in Limelight: The Gecko’s Climbing Choreography

Limb Functionality Swap: The study observes that the geckos’ limbs switch roles based on their position relative to the center of mass (COM). Limbs above the COM generate adhesive forces, while those below produce compressive forces.

Kinematic Adjustments: The research highlights kinematic changes across various levels: limbs, feet, and toes. Significant adjustments were noted in the hind limbs, indicating their flexible role in different climbing directions.

• DeepLabCut enabled the researchers to track and analyze the geckos’ movements with unprecedented precision. By marking 34 key points per frame on the geckos, DeepLabCut provided detailed kinematic data at multiple hierarchical levels: whole-body, limb, foot, and toe.

Display of the experimental set-up, the labels used for automated tracking and the calculation of the different kinematic variables on the different levels of the gecko (whole-body, limb, foot and toe level).

This isn’t just a cool party trick of nature. Understanding gecko locomotion opens doors in robotics, particularly in designing robots that can traverse complex terrains and surfaces.

3. Predictive Saccades and Decision Making in the Beetle-Predating Saffron Robber Fly

📚Talley, J., Pusdekar, S., Feltenberger, A., Ketner, N., Evers, J., Liu, M., Gosh, A., Palmer, S. E., Wardill, T. J., & Gonzalez-Bellido, P. T. (2023). Predictive saccades and decision making in the beetle-predating saffron robber fly. Current Biology, 33(14). https://doi.org/10.1016/j.cub.2023.06.019

• This study investigates the predictive visual tracking and decision-making process in the saffron robber fly, Laphria saffrana, a specialized beetle predator. The research demonstrates how Laphria employs a unique visual strategy to differentiate beetles from other insects using visual flicker cues generated by wingbeats.

Laphria is not just any hunter; it is a master of prediction and perception. This beetle-predating fly has mastered a remarkable strategy to single out beetles from a myriad of other flying insects. How does it achieve this? The key lies in the subtle flickers of light, the rhythmic reflections cast off by the wingbeats of beetles. Laphria has attuned itself to these flickers, using them as a beacon to distinguish its preferred prey.

The Ingenious Strategy of Laphria:

Visual Tracking Strategy: Laphria uses predictive saccades as part of its visual tracking strategy, alternating between rapid saccades and longer fixation periods. These saccades are predictive rather than purely reactive, allowing the fly to anticipate the future position of its target.

Predatory Behavior Based on Wingbeat Frequency: The fly categorizes its prey (specifically beetles) by detecting the frequency of light flashes reflected from their wingbeats. This unique sensory mechanism helps Laphria to distinguish between beetles and other flying insects.

Influence of Angular Position and Velocity: During the fixation phase, Laphria gathers information about the prey’s angular position and velocity, which informs the amplitude and timing of the subsequent predictive saccade.

Prey Categorization Using Visual Flicker: The fly’s predatory behavior is triggered by flickering stimuli resembling the wingbeat-induced light patterns of beetles. Experiments using LED panels mimicking these flickers confirmed Laphria’spreference for specific frequencies, reflecting its natural prey’s wingbeat rates.

• DeepLabCut (DLC) was crucial for accurately digitizing and analyzing the fly’s movements. It enabled automated tracking of key features like the antenna, head, and thorax in high-speed videos. DLC’s precision in detecting rapid antenna movements (saccades) and identifying fixation periods contributed to understanding the fly’s tracking behavior and gaze strategies during hunting.

Laphria’s predictive saccades are informed by prey’s subtended location and velocity.

Understanding Laphria’s visual tracking strategy opens doors to potential applications in robotics and artificial intelligence. This fly’s ability to process visual information rapidly and make split-second decisions can inspire the design of advanced robotic systems, especially in fields requiring quick sensory processing and decision-making

4. Algorithmic Assessment of Shoulder Function Using Smartphone Video Capture and Machine Learning

📚Darevsky, D.M., Hu, D.A., Gomez, F.A. et al. Algorithmic assessment of shoulder function using smartphone video capture and machine learning. Sci Rep 13, 19986 (2023). https://doi.org/10.1038/s41598-023-46966-4

• This study introduces a novel, cost-effective approach for diagnosing rotator cuff tears, a common source of shoulder pain. This research pivots from traditional, expensive imaging methods to an accessible solution using smartphones and machine learning. It involves a simple string-pulling task, interpreted through advanced algorithms, to assess shoulder function with high accuracy.

Imagine holding a string and performing a hand-over-hand motion, similar to pulling a rope. This simple action is recorded on a smartphone. The video isn’t just a recording; it provides information into the complex world of shoulder mobility and function.

The Magic of Machine Learning

Smartphone Videos and Machine Learning Synergy: By recording this task through smartphones, the researchers harness machine learning to analyze subtle movement nuances. This innovative approach demonstrates over 90% accuracy in classifying human patients with rotator cuff tears.

Cross-Species Analysis: The study’s methodology extends beyond humans, incorporating mice models, thus offering a comprehensive view of shoulder function assessment across species.

Accessible and Affordable Diagnosis: This method offers a potential in making diagnostics for rotator cuff injuries more accessible and cost-effective, using technology available to most people.

• DeepLabCut was used to analyze shoulder movement from smartphone videos. It was employed to precisely track specific features of shoulder movement during the string-pulling task. The trajectories and kinematics extracted through DLC played a crucial role in assessing shoulder function, enabling the algorithm to accurately identify patterns indicative of rotator cuff tears.

Post-processing kinematic trajectories to extract biomarkers of shoulder function.

The study is showcasing the importance of innovative thinking in medical diagnostics. By marrying simple physical tasks with sophisticated machine learning and widespread smartphone technology, it opens new avenues for accessible, efficient healthcare solutions.

5. Neural Encoding of Danger and Threat Specificity in the Prefrontal Cortex

📚Martin-Fernandez, M., Menegolla, A.P., Lopez-Fernandez, G. et al. Prefrontal circuits encode both general danger and specific threat representations. Nat Neurosci 26, 2147–2157 (2023). https://doi.org/10.1038/s41593-023-01472-8

• This study investigates how the dorsomedial prefrontal cortex (dmPFC) of mice processes and encodes both general danger and specific threats. Utilizing a novel behavioral paradigm, the research reveals that the dmPFC simultaneously forms a general representation of danger while distinctly encoding each specific threat.

Imagine being in a forest and suddenly sensing danger — your brain is quickly trying to figure out if it’s a general threat or something specific, like a snake or a bear.

The Dual Role of the dmPFC

Dual Encoding in dmPFC: The dmPFC encodes a general sense of danger and specific threat identities, critical for adaptive behavioral responses.

Neural Responses to Threats: Neurons in the dmPFC exhibit mixed-threat responses, neither solely generalizing across threats nor being entirely specific.

Optogenetic Validation: Optogenetic inhibition of dmPFC neurons impairs the selection of adaptive defensive responses, underscoring the region’s key role in behavioral adaptation to threats.

It turns out this area is like a sophisticated alarm system, simultaneously setting off a general alert (danger is near!) and pinpointing specific threats (it’s that snake again!).

• DeepLabCut was used for tracking and analyzing the movements of mice during various behavioral experiments. This involved quantifying their movements in response to different threatening situations, which were essential for understanding how the dorsomedial prefrontal cortex (dmPFC) encodes both a general representation of danger and specific representations of individual threats.

New MIO behavioral paradigm.

This study provides crucial insights into how the brain encodes and differentiates between general and specific aspects of danger. These findings offer a deeper understanding of the neural mechanisms underlying fear and anxiety.

6. Brainstem Circuits Encoding Start, Speed, and Duration of Swimming in Adult Zebrafish

📚Berg, E. M., Mrowka, L., Bertuzzi, M., Madrid, D., Picton, L. D., & El Manira, A. (2023). Brainstem circuits encoding start, speed, and duration of swimming in adult zebrafish. Neuron, 111(3). https://doi.org/10.1016/j.neuron.2022.10.034

• This study provides a comprehensive understanding of how the brainstem circuits in adult zebrafish encode the initiation, speed, and duration of swimming. The research focuses on the medial longitudinal fasciculus nucleus (nMLF), revealing the roles of two distinct glutamatergic neuron subpopulations (vGlut1+ and vGlut2+) in managing various aspects of swimming behavior.

Picture a zebrafish gliding through water — ever wonder how its brain orchestrates this seamless movement? It turns out, the zebrafish’s graceful swimming is all about teamwork among specialized brain cells!

Dive into the Brain of a Swimmer

Neuron Function Differentiation: The study identifies vGlut2+ neurons as key in encoding the start and duration of swimming, while vGlut1+ neurons encode rapid increases in swim speed and strength.

Behavioral Impact: Ablation studies demonstrate that impairing vGlut2+ neurons affects slow swimming, whereas vGlut1+ neuron impairment affects fast swimming.

Neural Activation Patterns: Calcium imaging and electrophysiological studies show distinct activation patterns for these neuron groups in relation to swimming speed and amplitude.

• DeepLabCut was used to analyze the tail movements of zebrafish during calcium imaging sessions. This tool allowed for precise tracking and quantification of the fish’s tail activity. The data obtained from DeepLabCut were then aligned with fluorescence changes in the fish’s brain neurons, linking specific neuronal activities with corresponding tail movements.

Encoding swim onset, speed, and duration in the nMLF.

This study explores the complex role of specific brainstem neurons in managing the nuanced aspects of swimming in zebrafish. It demonstrates how different neurons are responsible for distinct elements of locomotion, providing a clearer picture of the neural basis of movement in aquatic vertebrates.

Next time you see a zebrafish swimming in its tank💦, remember, it’s not just swimming🏊‍♀️; it’s executing a complex neural symphony, with each neuron playing its part to perfection🎶.

7. Unraveling the Complexities of Pain and Recovery in Mice

📚Bohic, M., Pattison, L. A., Jhumka, Z. A., Rossi, H., Thackray, J. K., Ricci, M., Mossazghi, N., Foster, W., Ogundare, S., Twomey, C. R., Hilton, H., Arnold, J., Tischfield, M. A., Yttri, E. A., St. John Smith, E., Abdus-Saboor, I., & Abraira, V. E. (2023). Mapping the neuroethological signatures of pain, analgesia, and recovery in mice. Neuron, 111(18). https://doi.org/10.1016/j.neuron.2023.06.008

• This study explores the intricate dynamics of pain, analgesia, and recovery in mice. By employing advanced electrophysiological and machine learning techniques, the researchers provide a deeper understanding of how mice experience and adapt to pain across various stages and how common analgesics affect these processes.

Exploring the World of Mouse Pain and Healing

Limb Kinematics and Pain: Observations reveal significant adjustments in limb movements correlating with various pain states, highlighting a complex interaction between physiological and behavioral responses to pain.

Role of 3D Pose Analytics: The use of 3D pose analytics proved crucial in identifying specific behavioral sequences indicative of different pain conditions.

Analgesics and Behavioral Change: The study found that while analgesics address hypersensitivity to stimuli, they do not revert the animals’ behaviors to their pre-injury state.

Neuroethological Signatures: The research uncovers unique neuroethological signatures associated with different stages of pain and recovery.

One of the study’s striking revelations was about pain relief. We often think taking a painkiller brings things back to normal, but the mice told a different story. Even when the pain seemed to fade, their movements didn’t entirely return to how they were before.

• DeepLabCut was employed to precisely track and analyze the movements of mice’s hind paws. DLC facilitated the differentiation of various pain responses in mice, such as distinguishing between allodynic (response to non-painful stimuli) and hyperalgesic (increased sensitivity to painful stimuli) behaviors.

PAWS and B-SOiD automated pain assessment platforms detect defensive coping behaviors associated with pain sensation during inflammation.

This study presents a deep analysis of the complexities surrounding pain and recovery in mice, emphasizing the adaptability of their behavioral responses to different pain stages. The insights gained extend the understanding of pain mechanics in animals, offering valuable implications for future research in pain management and animal welfare.

8. Neural mechanisms underlying uninstructed orofacial movements during reward-based learning behaviors

📚Li, W.-R., Nakano, T., Mizutani, K., Matsubara, T., Kawatani, M., Mukai, Y., Danjo, T., Ito, H., Aizawa, H., Yamanaka, A., Petersen, C. C. H., Yoshimoto, J., & Yamashita, T. (2023). Neural mechanisms underlying uninstructed orofacial movements during reward-based learning behaviors. Current Biology, 33(16). https://doi.org/10.1016/j.cub.2023.07.013

• This study investigates the neural basis of spontaneous orofacial movements in mice during reward-based learning tasks. It highlights how optogenetic stimulation of dopamine neurons in the ventral tegmental area (VTA-DA) triggers whisker and nose movements independently of goal-directed behaviors. The research reveals that a Pavlovian association between sensory cues and VTA-DA stimulation produces distinct cue-locked and VTA-DA-aligned orofacial movements.

The team’s exploration began with a curious discovery: activating specific brain cells — the VTA-DA neurons — in mice makes their whiskers and nose move spontaneously, as if responding to an invisible conductor. This discovery was akin to finding a hidden button in the brain, controlling these subtle movements.

The Brain’s Reward Symphony

Optogenetic Stimulation of VTA-DA Neurons: Stimulation leads to spontaneous whisker and nose movements.

Cue-VTA-DA Association: Results in distinct cue-locked and VTA-DA-aligned orofacial movements.

Role of Accumbal D1 Receptors: These receptors mediate VTA-DA-aligned, but not cue-locked, motion.

Involvement of Whisker Motor Cortex: wM1 is critical for both types of movements, with distinct neuronal populations representing each movement type.

• DeepLabCut played a critical role in analyzing nose movements of mice. The software was used for precise tracking and measurement of nose area, length, and lateral movements in response to specific stimuli. This analysis was essential in understanding the detailed dynamics of orofacial movements in mice, particularly in the context of reward-based learning behaviors.

Orofacial movements induced by transient activation of VTA-DA neurons.

The study offers significant insights into the neural mechanisms controlling uninstructed orofacial movements in mice during reward-based learning tasks. By delineating the roles of the VTA-DA neurons, nucleus accumbens, and wM1, it uncovers a complex neural network orchestrating spontaneous facial movements in response to reward cues.

9. A Neural Substrate for Short-Term Taste Memories

📚Juen, Z., Villavicencio, M., & Zuker, C. S. (2023). A neural substrate for short-term taste memories. Neuron. https://doi.org/10.1016/j.neuron.2023.10.009

• This study investigates the neural mechanisms underlying short-term taste memories in the gustatory cortex. The researchers used behavioral assays and neural activity recordings in mice to explore how the brain retains taste information over short periods.

Imagine a mouse in the wild, making split-second decisions on which foods to eat. This ability hinges on a highly sensitive sense of taste and a sophisticated mechanism for holding onto these taste experiences, even for just a few seconds. This study shows that this area stays active, holding onto the taste even after the food is gone, much like savoring a flavor after a meal.

Unraveling the Mystery of Taste Memories

Persistence of Gustatory Cortex Activity: The gustatory cortex exhibits persistent activity that outlasts the taste stimulus, suggesting its role in maintaining short-term taste memory.

Activity Decay and Memory: The decay of persistent activity in the gustatory cortex correlates with the fading of short-term taste memory.

Manipulation of Memory Trace: Early termination of persistent activity erases short-term memory, while extending it enhances memory retention.

• DeepLabCut was employed to precisely track and analyze facial movements of head-restrained mice. This technique enabled the detailed quantification of subtle orofacial movements, focusing on 13 key features including the nose, mouth, whisker pad, and eyes. The application of DeepLabCut was crucial for understanding the behavioral responses of mice to taste stimuli.

Persistent activity in taste cortex.

This research provides a deep understanding of the neural basis of short-term taste memories, revealing the gustatory cortex’s key role.

10.Social odor discrimination and its enhancement by associative learning in the hippocampal CA2 region

📚Hassan, S. I., Bigler, S., & Siegelbaum, S. A. (2023). Social odor discrimination and its enhancement by associative learning in the hippocampal ca2 region. Neuron, 111(14). https://doi.org/10.1016/j.neuron.2023.04.026

• This study investigates the role of the hippocampal CA2 region in processing social sensory information, specifically focusing on the discrimination of social odors and the impact of associative learning. Using two-photon calcium imaging in awake mice, the researchers demonstrate that hippocampal CA2 pyramidal neurons can distinguish between social odors (urine) from different mice. The study reveals that associative learning, where a social odor is paired with a reward, enhances the discrimination ability of these neurons.

The Social Scent Symphony

Discrimination of Social Odors: CA2 neurons exhibit the capacity to discriminate between different social odors.

Enhancement through Learning: Associative learning enhances the neurons’ ability to differentiate rewarded social odors from unrewarded ones.

Generalization of Odor Categories: CA2 neurons enable generalized classification between social and non-social odors.

• DeepLabCut was employed to analyze the movement trajectories of mice during free exploration. DLC provided detailed visualizations of how mice interacted with various odor sources, contributing to the understanding of their exploratory behaviors and responses to different stimuli. The trajectories, combined with manual analysis of active exploration times, offered comprehensive insights into the mice’s behavior in response to odors.

Adaptive social odor-reward association learning in mice using a Go/No-Go operant conditioning task.

The study underscores the importance of the hippocampal CA2 region in social cognition, particularly in discriminating and encoding social odors. It provides insights into the neural mechanisms underlying social memory formation, enhanced through associative learning.

As we look back on 2023, it was a remarkable year for science, filled with intriguing discoveries. From understanding 🦎geckos’ gravity-defying climbs to unveiling the secrets of zebrafish swimming💦, researchers have been busy unraveling nature’s mysteries. In the background, tools like DeepLabCut have been instrumental, helping scientists track and analyze movements in these studies.

🚀 There’s a universe of cool studies out there, and we’ve only scratched the surface. Got a favorite that sparks joy and curiosity?

🌟 Don’t keep it to yourself — share it with us @deeplabcut !

Let’s keep the conversation going and the inspiration flowing.

Here’s to a year of fascinating science and the promise of more discoveries in 2024. Happy New Year to our curious and ever-expanding community of researchers and science enthusiasts! 🌟🐭🎉