Biology subjects

Nguyen, S. L.

Publications and source records attributed to Nguyen, S. L..

5 recordsLinked to original sources

bioRxiv2025

Physics-Informed Neural Networks for Real-Time Deformation-Aware AR Surgical Tracking

Soft tissue deformation severely degrades registration accuracy in AR-assisted surgery. We propose a physics-informed neural network (PINN) that integrates biomechanical priors into real-time depth-based registration. The model embeds finite element elasticity constraints directly into the loss function, allowing neural predictions to remain physically plausible under deformation. Validated on liver and brain phantoms with induced deformations up to 20 mm, the method achieved mean registration error of 1.1 mm, compared with 2.9 mm for conventional ICP and 1.8 mm for FEM-only solvers. Frame rates remained at 22 fps on GPU hardware. Results demonstrate that embedding physics constraints within deep learning significantly enhances robustness in dynamic surgical contexts.

bioengineering↗

bioRxiv2021

A large-scale neural network training framework for generalized estimation of single-trial population dynamics

Recent technical advances have enabled recording of increasingly large populations of neural activity, even during natural, unstructured behavior. Deep sequential autoencoders are the current state-of-the-art for uncovering dynamics from these datasets. However, these highly complex models include many non-trainable hyperparameters (HPs) that are typically hand tuned with reference to supervisory information (e.g., behavioral data). This process is cumbersome and time consuming and biases model selection toward models with good representations of individual supervisory variables. Additionally, it cannot be applied to cognitive areas or unstructured tasks for which supervisory information is unavailable. Here we demonstrate AutoLFADS, an automated model-tuning framework that can characterize dynamics using only neural data, without the need for supervisory information. This enables inference of dynamics out-of-the-box in diverse brain areas and behaviors, which we demonstrate on several datasets: motor cortex during free-paced reaching, somatosensory cortex during reaching with perturbations, and dorsomedial frontal cortex during cognitive timing tasks. We also provide a cloud software package and comprehensive tutorials that enable new users to apply the method without dedicated computing resources.

neuroscience↗

bioRxiv2020

INTEGRINS MEDIATE PLACENTAL EXTRACELLULAR VESICLE TRAFFICKING TO LUNG AND LIVER IN VIVO

Membrane-bound extracellular vesicles (EVs) mediate intercellular communication in all organisms, and those produced by placental mammals have become increasingly recognized as significant mediators of fetal-maternal communication. Here, we aimed to identify maternal cells targeted by placental EVs and elucidate the mechanisms by which they traffic to these cells. Exogenously administered pregnancy-associated EVs traffic specifically to the lung; further, placental EVs associate with lung interstitial macrophages and liver Kupffer cells in an integrin-dependent manner. Localization of EV to maternal lungs was confirmed in unmanipulated pregnancy using a transgenic reporter mouse model, which also provided in situ and in vitro evidence that fetally-derived EVs, rarely, may cause genetic alteration of maternal cells. These results provide for the first time direct in vivo evidence for targeting of placental EVs to maternal immune cells, and further, evidence that EVs can alter cellular phenotype.

cell biology↗

bioRxiv2020

The Neurovascular Basis of Postictal Amnesia

Long-lasting confusion and memory difficulties during the postictal state remain a major unmet problem in epilepsy that lacks pathophysiological explanation and treatment. We previously identified that long-lasting periods of severe postictal hypoperfusion/hypoxia, not seizures per se, are associated with memory impairment after temporal lobe seizures. While this observation suggests a key pathophysiological role for insufficient energy delivery, it is unclear how the networks that underlie episodic memory respond to vascular constraints that ultimately give rise to amnesia. Here, we focused on cellular/network level analyses in the CA1 of hippocampus in vivo to determine if neural activity, network oscillations, synaptic transmission, and/or synaptic plasticity are impaired following kindled seizures. Importantly, the induction of severe postictal hypoperfusion/hypoxia was prevented in animals treated by a COX-2 inhibitor, which experimentally separated seizures from their vascular consequences. We observed complete activation of CA1 pyramidal neurons during brief seizures, followed by a short period of reduced activity and flattening of the local field potential that resolved within minutes. During the postictal state, constituting tens of minutes to hours, we observed no changes in neural activity, network oscillations, and synaptic transmission. However, long-term potentiation of the temporoammonic pathway to CA1 was impaired in the postictal period, but only when severe local hypoxia occurred. Lastly, we tested the ability of rats to perform object-context discrimination, which has been proposed to require temporoammonic input to differentiate between sensory experience and the stored representation of the expected object-context pairing. Deficits in this task following seizures were reversed by COX-2 inhibition, which prevented severe postictal hypoxia. These results support a key role for hypoperfusion/hypoxia in postictal memory impairments and identify that many aspects of hippocampal network function are resilient during severe hypoxia except for long-term synaptic plasticity.

neuroscience↗

bioRxiv2018

Visualizing murine placental extracellular vesicle data with tidyNano: a computational framework for analyzing and visualizing nanoparticle data in R

Extracellular vesicles (EVs) are increasingly recognized as important mediators of intercellular communication, and in mammals are generally classified as ~50-150nm exosomes, ~100-1000nm microvesicles, and apoptotic bodies, each arising as a result of different biological processes. EVs carry protein, lipids, and nucleic acids within the circulation, to target cells whereupon they mediate physiological changes. Due to their small size, quantification and characterization by conventional microscopy is not possible. However, nanoparticle tracking analysis (NTA) has provided a method to determine the fluid concentration and size of extracellular vesicles and other nanoparticles. While NTA provides statistical summaries of samples in an experiment, a recurring difficulty is the organization, manipulation, and management of raw particle count data because of the large size of datasets, and resultant vulnerability to user error. To address these limitations, we developed tidyNano, an R package that provides functions to import, clean, and quickly summarize raw NanoSight (Malvern Panalytical) data for efficient calculation of statistics and visualization. Here, we provide a framework for importing raw nanoparticle data and provide functions to facilitate rapid and efficient analysis, visualization and calculation of summary statistics. tidyNano was used to analyze murine plasma extracellular vesicles across gestation by aggregating and summarizing samples based on technical, biological and gestational parameters. In addition, we developed shinySIGHT, a Shiny web application that allows for interactive exploration and visualization of EV data. Using this package, we analyzed data generated from 36 samples of EV derived from the plasma of mice across gestation.\n\nPeripheral EV concentration increased linearly across pregnancy, with trending increases as early as gestation day (GD) 5.5 and significant rises at GD14.5, and 17.5 relative to EV concentrations in nonpregnant females. Thus, the data highlight the utility of the mouse as a model of EV biology in pregnancy. Further, the package provides a mechanism for seamless analysis of EV data generated by NanoSight. Importantly, this package provides a straightforward framework by which diverse types of large datasets can be simply and efficiently analyzed. tidyNano and shinySIGHT are freely available under the MIT license and is hosted on GitHub (https://nguyens7.github.io/tidyNano/).

cell biology↗