22484 research outputs found
Sort by
Associations Between Striatal Dopaminergic Integrity and Gait Quality
No full textGait quality aspects beyond pace like walking smoothness (harmonic ratio), regularity (entropy rate), and complexity (Lempel-Ziv complexity) may be useful in capturing the motor control strategies important in successful everyday walking. Dopaminergic (DA) integrity in the central nervous system plays a crucial role in motor function. Age-related declines in dopamine within the striatum may contribute to concurrent declines in gait quality in older adults without Parkinson’s. We investigated the relationship between striatal DA integrity and gait quality in an exploratory analysis. In 199 participants (Age: 75.00 years (4.63), Sex: 123 Females (61.81%) BMI: 28.22 (5.41)), DA integrity was measured using [11C]DTBZ PET imaging estimating the striatal binding site density of the type 2 Vesicular Monoamine Transporter (VMAT2). Accelerometers on the L3 segment of the lumbar spine during walking over 15 meters measured entropy rate, Lempel-Ziv complexity, wavelet entropy, and harmonic ratio of the signal in the medio-lateral (ML), anterior-posterior (AP), and vertical anatomical directions. Multiple linear regression estimated the cross-sectional association between the striatal region of interest [11C]DTBZ binding density and each gait measure, adjusted for age, sex, BMI, race, education, and hours walked per week. Before correction for multiple comparisons, higher [11C]DBTZ binding density in the precommissural dorsal caudate was associated with higher entropy rate in the AP anatomical plane (β: 0.033, 95% CI [0.001, 0.066]). Higher [11C]DBTZ binding density in the anterior ventral striatum was associated with lower gait speed (β: -0.112, 95% CI [-0.213, -0.012]). Additionally, before correction, higher [11C]DBTZ binding density in the postcommissural putamen was significantly associated with lower wavelet entropy in the vertical anatomical plane (β: -0.084, 95%CI [-0.153, -0.015]). After Bonferroni correction no relationships between dopaminergic integrity in the striatum and gait quality characteristics were significant. Novel accelerometry gait characteristics and PET imaging are powerful tools for understanding mechanisms of age-related neurological decline. Mitigation of mobility limitations remains an important public health focus
Deep Learning Methods and Datasets for All-atom Protein Structure Prediction
No full textOver the past decade, significant progress has been made in protein structure prediction, largely thanks to tools like AlphaFold2. This paper tackles several lingering challenges in this field. First, we introduce SidechainNet, a dataset and toolkit designed to streamline the handling of protein sequence and structure data for machine learning and increase its accessibility. This initiative addresses the prevalent issue of effectively collecting and organizing data, especially in the realm of protein science, where data quality and availability may vary. Furthermore, leading methods have been observed to fall short in real-world tasks like molecular docking. To enhance the physical realism of predictions made by deep learning models, we implement potential energy as a loss function through OpenMM-Loss. This technique reduces potential energy and clashes in predicted structures, potentially rendering these predictions more viable for various applications. We also scrutinize AlphaFold2 with the aim of refining its sidechain modeling—a crucial aspect of drug discovery. Although we don’t pinpoint a significantly more accurate model, our analysis reveals comparable performance between ResNet and Transformer models in sidechain prediction tasks. In light of these results, we recommend that future efforts concentrate on more holistic sidechain modeling efforts. Finally, we discuss potential future developments and extensions of our methods
New Developments in Inherent Strain Method for Predicting and Mitigating Residual Stress and Distortion in Metal Additive Manufacturing
No full textAs an additive manufacturing process for fabricating metal components, laser powder bed fusion (L-PBF) and wire-arc directed energy deposition (wire-arc DED) have drawn increasing attention in the past few decades due to their advantages such as fast production, high customization, and waste reduction. During fabrication, the rapid, intense, and repeated heat input (laser beam or wire arc) leads to a complex thermal history in parts, resulting in significant residual stress and distortion. These residual stress and distortion can adversely affect product quality by increasing surface roughness, reducing dimensional accuracy, and introducing defects into the parts.
This dissertation is focused on improving the inherent strain (IS) method for predicting residual stress and deformation in parts manufactured by L-PBF and wire-arc DED processes. In addition, two frameworks based on the IS method for recoater interference prediction and distortion compensation in L-PBF are proposed. Chapter 2 introduces a new procedure for implementing the modified inherent strain (MIS) method. This procedure incorporates an additional solution step that uses mechanical properties at elevated temperatures, markedly improving the accuracy of the MIS method on residual stress prediction. Chapter 3 extends the MIS method to include the heat accumulation effect in the wire-arc DED process. This enhancement involves introducing a flashing heating simulation to calculate interpass temperature and applying temperature-dependent ISs in the MIS-based simulation. In Chapter 4, an integrated simulation and experimental framework for predicting potential recoater interference in the L-PBF process is proposed. This framework addresses the previously undefined criterion for recoater interference and incorporates the edge effects when calculating the part deformation. In Chapter 5, a data-driven distortion compensation framework for the L-PBF process is presented. The framework employs a Gaussian process regression (GPR) model and reduced-order modeling to learn from experimentally-validated IS simulation data and generate the compensated shape
Data from: The Chloride/proton Antiporter ClC3 is Critical for Normal Bone Formation in Mice
No full textMaterials, methods, data, and figures for a paper currently under review.
Acid transport is required for bone synthesis by osteoblasts. The osteoblast basolateral surface extrudes acid by Na+/H+ exchange, but proton uptake is unclear. We found high expression of the Cl-/H+ exchanger ClC3 at the bone apical surface. In mammals ClC3 functions in intracellular vesicular chloride transport, but when we found Cl- dependency of H+ transport in osteoblast membranes, we queried whether ClC3 Cl-/H+ exchange functions in bone formation. We used ClC3 knockout animals, and closely-related ClC5 knockout animals: In vitro studies suggested that both ClC3 and ClC5 might support bone formation. Genotypes were confirmed by total exon sequences. Expression of ClC3, and to a lesser extent of ClC5, at osteoblast apical membranes was demonstrated by fluorescent antibody labeling and electron microscopy with nanometer gold labeling. Animals with ClC3 or ClC5 knockouts were viable. In ClC3 or ClC5 knockouts, bone formation decreased ~30% by calcein and xylenol orange labeling in vivo. In very sensitive micro-computed tomography, ClC5 knockout reduced bone relative to wild type, consistent with effects of ClC3 knockout, but was quantitatively smaller. Regrettably, ClC5-ClC3 double knockouts are not viable. We conclude that ClC3 has a direct role in bone formation with overlapping but probably slightly smaller effects of ClC5. The mechanism in mineral formation might include ClC H+ uptake, in contrast to ClC3 abd ClC5 function in cell vesicles or other organs.
Note for mass spectrometry data: These are the essential data for the MS evaluation of ClC3 expression. The KO expresses a shorter protein at about 60% of the level of the wild type. First two are wild type, second two are KO
New Efficient Pruning Algorithms for Compressing and Accelerating Convolutional Neural Networks
No full textRecently, Convolutional Neural Networks (CNNs) are continuously achieving state-of-the-art results in numerous machine-learning tasks. While having impressive performance, the size of current models is also exploding. Motivated by efficient inference, many researchers have been devoted to reducing the storage and computational costs of state-of-the-art models. Channel pruning emerges as a promising solution to reduce the size of the model, and it can achieve acceleration without any post-processing steps. Current channel pruning methods are either time-consuming (reinforcement learning, greedy search, etc.) or depend on fixed criteria of channels resulting in poor results.
In this dissertation work, we propose new methods from the perspective of gradient-guided pruning. We then formulate the pruning problem as a constrained discrete optimization problem. Our discrete model compression work aims to solve this constrained problem by using differentiable gates and propagating gradients through a straight-through estimator. We further improve the results in network pruning via performance maximization by adding a performance prediction loss into the constrained optimization problem. The search for sub-networks is then directly guided by the accuracy of a sub-network. The improvement of supervision leads to better pruning results.
On top of previous works, we propose to further improve our algorithms from different perspectives. The first perspective is to disentangle width and importance for finding the optimal model architecture. From this end, we propose to use an importance generation network and a width generation network to generate the importance and width for each layer. Another challenge in previous works is the huge gap between the model before and after network pruning. To mitigate this gap, we first learn a target sub-network during the model training process, and then we use this sub-network to guide the learning of model weights through partial regularization. Based on the success of previous static pruning methods, we further incorporate dynamic pruning for storage-efficient dynamic pruning
Bringing Agile and Self-Evolvable Intelligence to Weak Embedded Devices
No full textNeural Networks (NNs) can significantly enhance perception and decision-making in resource-constrained devices like drones and wearables. However, limited resources such as memory and computing power hinder modern NN designs, leading to inaccurate predictions and delayed execution.
To first ensure dependable inference, we propose modular NN structures mimicking expert decision-making. We study the effectiveness of our methodology in wireless backscatter systems under noisy channel conditions, where a modular NN is tailored for predictive power adaptation. Despite NN structure advancements, microcontroller-equipped devices still face performance barriers under extreme constraints, such as limited memory (1MB) and low clock frequency (300MHz). To enable more efficient use of limited resources, we propose agile offloading, which uses the patterns of feature importance identified by explainable AI to enhance the offloading efficiency. Due to the non-stationary world, NN models should also be promptly retrained d with new data, allowing it to continuously adapt to environmental dynamics and maintain its accuracy. To achieve this adaptivity, we propose a selective training scheme, where NN substructures can be freely added or skipped at runtime based on their importance with user desired computational costs. We showcase the effectiveness of our scheme on both vision and Large Language Models (LLMs). In addition to retraining upon a stationary structure, we further envision that the NN structure should be runtime expandable to accept more data modalities captured by the device. Such self-evolvability can improve the NN’s generative and reasoning capabilities in more complex tasks like autonomous navigation and human-device interaction. However, as more data modalities are incorporated, continuously enlarged models encounter scalability challenges. To mitigate training costs, we propose connecting unimodal encoders to a flexible set of last LLM blocks, training only such latent connections at runtime. We showcase its improved accuracy-compute efficiency in multimodal question-answering tasks for autonomous driving scenarios
Changing Blues: The Continued Life and Appropriation of Black Women's Blues in Twenty-First Century Popular Culture
No full text“Changing Blues: The Continued Life and Appropriation of Black Women’s Blues in Twenty-First-Century Popular Culture” argues that contemporary creators reimagine or repurpose the politics and methods of 1920s-1930s Black women’s blues music to amplify particular political investments and to attain social value within the twenty-first-century popular culture arena. While specific historical events gave rise to blues women’s tactics, the issues they explored remain relevant today, such as class, sexual expression, and romantic relationships which together make up the sexual-economic. The sexual-economic is a dual concept. It refers to existing systems that oppress Black women through conjoined sexual and economic means and a way of performing resistance that understands the sexual and economic as intertwined concerns for building and imagining radically liberated futures. Focusing on the sexual-economic theme across early twentieth-century blues and twenty-first-century popular culture, this dissertation explores contemporary work from pop stars, blues singers, and filmmakers. While these projects riff on older blues politics, they also reflect their own moment’s concerns. The blues’ sonic slipperiness, I argue, is a double-edged sword: giving it the ability to adapt but also putting it at risk for commodification and codification that waters down or erases its radical provocations. In adopting a moving or unmoored orientation towards blues politics, we can readjust in parallel to the blues’ movements across time and space, simultaneously holding on to blues women’s often-silenced contributions while letting go of how and where the blues must sound. This approach will allow us to understand not only the blues’ functioning today but how Black women’s artistic and critical contributions are put to work within popular culture
The burden of animal bites in Allegheny County, Pennsylvania, 2022
No full textBackground: Animal bites are a public health concern, with thousands of people seeking treatment in the emergency room annually. Dog encounters cause more severe injuries and account for most bite-related hospitalizations. Complications may result from animal bites including rabies, a viral disease that is fatal once symptoms appear and damage bones and joints. People of all ages may be bitten by animals, but children are disproportionally affected. The objective of this analysis was to describe the frequency of animal bite reports and the burden of animal bite related hospitalizations in Allegheny County, PA for the year 2022.
Methods: We reviewed animal bite reports received by the Allegheny County Health Department (ACHD) from healthcare providers via fax or web-based data forms. The data were entered into an Oracle database and exported into an Excel file. SAS was used to clean and analyze data. Hospitalization data were obtained from the PA Health Care Cost Containment Council via ACHD. Primary and secondary diagnosis codes and external cause of injury codes related to animal bites were used to identify bite-related hospitalizations. Descriptive statistics were generated in SAS and used to create charts and graphs in Microsoft Word and Excel. Data from previous reports were used to show trends from 2013 through 2022.
Results: ACHD received 2,189 unique animal bite reports in 2022, a 23% increase since 2013. Most bites were attributed to dogs (77.8%) or cats (17.5%). In 2022 there were 92 hospitalizations related to animal bites among Allegheny County residents. Most hospitalizations were attributed to dog bites (56.5%) and cats (14.9%). Females (60.9%) compromised more bite-related hospitalizations than males (39.1%). The mean length of stay in the hospital was 3 days.
Conclusion: The incidence rates of animal bites continues to rise in Allegheny County and remains a public health concern. Animal bites pose a risk of bacterial infection and exposes people to the risk of rabies transmission. Surveillance should continue, along with interventions, including educational initiatives to address this public health concern
Long-term Stability, Test-Retest Reliability, and Lexical Properties of Core Nouns from Novel Discourse Stimuli
No full textBackground: Core Lexicon Analysis (CLA) measures the use of keywords normed against the neurologically intact population during discourse. It is a promising tool to evaluate and monitor discourse production for people with aphasia (PWA), because it requires little transcription and training, and has been shown to have good interrater reliability and concurrent validity (Kim & Wright, 2020b). However, there are no studies on its test-retest reliability.
Aims: This study aims to investigate the test-retest reliability of CLA, specifically for core nouns, for a set of novel discourse stimuli. We also aimed to explore the effect of sample length on the test-retest reliability of CLA and characterize the phonological and semantic properties of the core nouns.
Methods & Procedures: Thirty-one non-brain damaged (NBD) community dwelling adults produced discourse in picture description and narrative storytelling tasks using novel visual stimuli. Fifteen participants saw discourse stimuli again after a mean of 149. Core noun lists, core noun scores, parameters related to test-retest reliability such as intraclass correlation coefficients (ICCs) and standard errors of measurement (SEMs), and sample lengths (defined as speaking durations) were obtained. Lexical properties of core nouns were extracted from corpus databases and normative studies (Balota et al., 2007; Fergadiotis et al., 2019; McRae et al., 2005; Johns, 2021).
Result: The ICCs for this set of stimuli ranged from –0.517 (95% Confidence Level [CI]: –0.875 to 0.033) to 0.695 (95% CI: [0.292, 0.887]), which are classified as “poor to moderate” according to Koo & Li (2016). Consistent with the ICCs, the group percent agreements ranged from 0 and 0.8, indicating considerable variability over time. Spearman’s ρ between sample lengths and ICCs for individuals ranged from 0.14 to 0.72, indicating a small to moderate positive correlation between the two variables.
Conclusions: We conducted the first study on the test-retest reliability of CLA. Under our experimental conditions, the test-retest reliability of core nouns for picture description and story narrative tasks was poor, though it seems to increase slightly as discourse duration increases
Measuring the Test-Retest Reliability of Communicative Congruence and Communicative Dysphoria
No full textVarious communicative qualities can convey meaningful information about a speaker’s identity and personality. When the qualities and associated connotations of someone's communication behaviors do not align with their own self-identities and experiences, they may experience distress or discomfort. Two constructs describe this phenomenon. Communicative congruence (CC) is the experience of alignment between one’s sense of self and their voice, speech, and communication. Communicative dysphoria (CD) is significant emotional, motivational, and/or cognitive distress due to communicative incongruence. Both constructs are framed within the Cybernetic Big Five Theory of personality.
CC and CD were previously measured in a sample of 366 individuals. In order to understand the behavior of these constructs over time, this study repeated 2 measures of CC and 2 measures of CD in subset of the 366 participants at three time points. Test-retest reliability was assessed using an interclass correlation coefficient. Findings indicated that communicative congruence and communicative dysphoria will be more reliable when measured in the short-term, and that CC has greater test-retest reliability than CD across measurement approaches. Clinical application of CC and CD measurement include as an outcome measure and prognostic indicator. Future work is necessary to understand cross-group variance and clinical value