DSpace@RPI (Rensselaer Polytechnic Institute)
Not a member yet
6809 research outputs found
Sort by
Does Imageable Language Make Your Tweets More Persuasive?
No full textImageability is a psycholinguistic property of words that indicates how quickly and easily a word evokes a mental image or other sensory experience. Highly imageable words are easier to read and comprehend, and, as a result, their use in communications, such as social media, makes messages more memorable, and, potentially, more impactful and influential. In this paper, we explore the relationship between the imageability of messages in social media and their influence on the target audience. We focus on messages surrounding important public events and approximate the influence of a message by the number of retweets the message receives. First, we propose novel ways to determine an imageability score for a text, utilizing combinations of wordlevel imageability scores from the MRCPD+ lexicon, as well as word embeddings, image caption data, and word frequency data. Next, we compare these new imageability score functions to a variety of simple baseline functions in correlation between tweet imageability and number of retweets in the domain of the 2017 French Presidential Elections. We find that the imageability score of messages is correlated with the number of retweets in general, and also when normalized for topic and novelty; thus, imageable language is potentially more influential. We consider grouping tweets into imageability score ranges, and find that tweets within higher ranges of imageability scores receive more retweets on average compared to tweets within lower ranges. Lastly, we manually annotate a small number of tweets for imageability and show that our imageability score functions agree well with the human annotators when the agreement between human raters is high
Multirate digital signal processing for time interleaved analog to digital converters
No full textMay 2017School of EngineeringIn today’s increasingly higher speed and higher data rate world, high speed Analog to Digital Converters (ADCs) are very much in demand. They are needed in multiple applications including communications, radars, software defined radio, high speed digital oscilloscopes to cite a few. Time interleaving two or more ADCs running at a lower sampling rate enables achieving a higher sampling rate. Interleaving N sub-ADCs operating at a sampling rate Fs would ideally result in a Time Interleaved ADC (TIADC) with an N fold increased sampling rate. However, interleaving comes with a set of challenges. Offset, Gain and timing skew mismatches between the sub-ADCs cause undesired degradations in the performance of the TIADC. The purpose of this research work is to design and implement a programmable Digital Signal Processing (DSP) module that will help mitigate the offset, Gain and timing skew mismatches encountered in TIADCs. The core building block of the Digital Signal Processing module implemented in the first part of this thesis work is a fractional delay filter for detecting timing skew mismatches preceded by Offset and Gain mismatch compensation modules. Various building blocks of the Digital Signal Processing module for minimizing the effect of TIADC mismatches are implemented using the IBM CMOS 9HP standard cell library which is part of the IBM BICMOS 9HP technology kit. In the second part, a scalable, extremely efficient timing skew detection and compensation technique using a single FIR filter once, is developed; enabling close to 50% reduction in hardware resource utilization compared to current state of the art timing skew calibration techniques. This efficient timing skew calibration technique is then expanded to incorporate a current state of the art gain mismatch calibration resulting in a very efficient all digital background gain and timing skew mismatch calibration algorithm. Future work includes the integration of the Digital Signal Processing module with a 2-way TIADC designed using the Heterojunction Bipolar Transistor (HBT) devices which are part of the IBM BICMOS 9HP kit. One could also envision implementing the timing skew detection and mitigation technique developed in this research work for a 4-way, 8-way TIADC and beyond.Ph
Whyis 2: An Open Source Framework for Knowledge Graph Development and Research
No full textWhyis is the first open source framework for creating custom provenance-driven knowledge graph applications, or KGApps, support- ing three principal tasks: knowledge curation, inference, and interaction. It has been used in knowledge graph projects in materials science, health informatics, and radio spectrum policy. All knowledge in Whyis graphs are encapsulated in nanopublications, which simplifies and standardizes the production of qualified knowledge in knowledge graphs. The architecture of Whyis enables what we consider to be essential requirements for knowledge graph construction, maintenance, and use. These require- ments include support for automated and manual curation of knowledge from diverse sources, provenance traces of all knowledge, domain-specific user interaction, and generalized distributed knowledge inference. We coin the term “Nanoscale knowledge graph” to refer to nanopublication-driven knowledge graphs. Knowledge graph developers can use Whyis to configure custom sets of knowledge curation pipelines using custom data importers and semantic extract, transform, and load scripts. The flexible, nanopublication-based architecture of Whyis lets knowledge graph developers integrate, extend, and publish knowledge from heterogeneous sources on the web. Whyis KGApps and are easily developed locally, managed using source control, and deployable via continuous integra- tion, server deployment scripts, and as docker containers.This work was funded by the National Institute of Environmental Health Sciences (NIEHS) Award 0255-0236-4609 / 1U2CES026555-01, National Science Foundation (NSF) Award DMR-1310292, IBM Research AI through the AI Horizons Network, the National Spectrum Consortium (NSC) project number NSC-17-7030, and by the Gates Foundation through Healthy Birth, Growth, and Development knowledge integration (HBGDki). Any opinions, findings and conclusions or recommendations expressed in this material are those the authors and do not necessarily reflect the views of AFRL, IBM, or the Gates Foundation
Modeling, system identification, and parameter estimation for electrified aircraft systems and hydroelectric power plants
No full textDecember 2022School of EngineeringIn recent years, there has been a push to electrify everyday technologies to help meet emissionsreduction and promote more sustainable energy consumption practices. In this work, three systems that
consider the different aspects of electrification development are modeled, identified, and analyzed. The
methods applied for modeling, identification and analysis aim to help in making electrification more
sustainable throughout the entire engineering life cycle. This work considers a model-based systems
engineering framework, where the systems are represented by physics-based models to analyze with
system identification methods and for various trade studies. Hydroelectric power plants are well-established systems that have been in operation for decades, wherefield measurement data collected from the electric power grid is used to validate and identify power plant
models. These models have been defined in a standardized and custom manner, allowing us to apply our
methodology to a system with models that have been in use for a long time. The parameters of the models
are estimated using field data, showing the value in model maintenance and standardization. This work
also introduces a methodology for validating the individual components of a power plant, making it
necessary to re-identify and re-validate only the affected components. This methodology is incredibly
valuable when performing re-validation, as considering only the invalid component reduces the
complexity of the optimization problem. Next, we model and study an electric vertical take-off and landing (eVTOL) aircraft. These systems are inan earlier stage of development than Hydroelectric power plants, where physical prototypes and products
exist for some eVTOL systems, but have been limited in wide-scale application and therefore operational
measurement data is not readily available (e.g. hobbyist drones). Many existing eVTOL systems are small,
so this work shows how we can expand on the existing modeling technology and study multi-domain
dependencies when eVTOL are scaled to provide human-scale transport. The modeling approach enables
the development of a library to model quadcopters using physics-based components through a flexible
modeling framework. It is then used to study an eVTOL drivetrain to determine the effects of battery
configuration and motor modeling fidelity on dynamic response, showing necessary considerations
needed for eVTOL design. The third system discussed is an electrified aircraft system that is still in the early phases of design anddevelopment. It is necessary to expand upon existing research to develop sub-domain components for
the aircraft as there is no cohesive physical prototype of the specific aircraft or its subsystems available
to validate the models. A novel system architecture was developed for the fully-electric aircraft concept,
showing the different considerations needed to design the electrical and thermal systems as well as each
of the individual subsystems. This also required the development of novel cryogenic component models,
which were then studied in the aircraft configuration under fault conditions. This fault study shows how
the sizing of other components in a novel aircraft system can be utilized to mitigate the impact of faults.Ph
Deep learning for video-based assessment of surgical skills
No full textDecember 2022School of EngineeringSurgical skill assessment is crucial for training and certification. The current gold standard is based on real-time proctoring in the operating room (OR) – developed by Halsted over a century ago. To assess skills, proctors use objective rating scales such as the Objective Structured Assessment of Surgical Skills (OSATS) or time and error metrics of the Fundamentals of Laparoscopic Surgery (FLS) program. The current assessment techniques suffer from drawbacks, including being subjective with poor inter-rater reliability, distribution errors, recall bias, and halo effects. Video-based assessment (VBA) can ensure patient safety by allowing surgeons to provide post-hoc feedback. Still, VBA is not real-time, and post-hoc evaluation may lead to burnout. Also, it is prone to subjective interpretation. Deep Learning (DL) can circumvent these limitations. However, current models use video snippets, ignoring long-term representations. This prevents them from providing meaningful feedback. Besides, models predominantly use motor actions, e.g., hand motion, for assessment which is not holistic. Finally, they are data-hungry and restricted to their training domain. This prevents them from adapting to new tasks without retraining via data-intensive sets. This thesis addresses these limitations to achieve feasible real-life implementation of VBA.
Firstly, we propose a deep learning (DL) model – Video-Based Assessment Network (VBA-Net) – that can automatically and objectively assess surgical skills in real-time from complete videos while generating statistically-verifiable feedback via Class Activation Maps (CAMs). We also benchmarked the VBA-Net and achieved state-of-the-art performance in the commonly-used public dataset, JIGSAWS.
Secondly, we fuse neural activations from the prefrontal cortex – shown to differentiate skill levels – with motor action data for holistic skill assessment. We also compare motor actions with neural activations. For comparison, we use a t-test on the distributions of performance metrics, e.g., accuracy and R2, after 100 repetitions. The VBA-Net generates slightly better assessment performance via neural activations (p<.05) than motor actions, and multimodality leads to the best performance (p<.05). This shows the saliency of neural activations and the advantage of holistic inputs in skill assessment.
Lastly, we develop an Adaptive VBA-Net (A-VBANet) to deliver domain-agnostic skill classification via one-shot learning without retraining. A-VBANet successfully adapts to independent physical simulators with a single video-based sample. Further, using the simulator data, A-VBANet successfully adapts to the OR task – laparoscopic cholecystectomy – only with one sample.
The technology developed in this thesis has the potential to significantly impact patient care by providing automated assessment tools for surgeons based on videos. Large-scale adoption will require the development of surgical video repositories. However, the development of meta-learning approaches will mitigate the need for intensive annotation and enable the assessment of intra-operative videos based on simulator data, which is abundant and easier to obtain.Ph
Eat4Genes: A Bioinformatic Rational Gene Targeting App and Prototype Model for Improving Human Health
No full textEat4Genes is a prototype diet recommendation web app for patients, healthcare providers, and researchers that aids in the selection of a healthy diet to help treat and prevent numerous health conditions. Our approach is focused on the strategic use of diet to regulate key risk gene expression, which we call dietary rational gene targeting (DRGT). To create the source DRGT dataset, we analyzed data of three types: an expert-curated list of conditions and the desirable modulation associated with risk genes to improve those conditions; an expert-curated list of rigorous scientific studies that quantify how the consumption of dietary agents (nutrients) influences gene expression in humans; and gene expression results for each of the studies. For each study, we added analyses of differentially expressed genes, study quality, and nutrient concentration. We developed a ranking system for studies that emphasize in vivo over in vitro studies, as well as whole foods and extracts over isolated phytochemicals at reasonable concentrations of nutrients. The Eat4Genes web app provides an engaging and informative interface that enables users to perform analysis by condition or by gene. We also present user scenarios from physician's and researcher's perspectives. The prototype Eat4Genes DRGT dataset and web app represent important steps towards translating DRGT and dietary research into a precision nutrition approach that is lower cost and healthier compared to pharmaceutical approaches
Three essays on ethical issues in natural language use for the design and implementation of artificial intelligence (ai) systems
No full textAugust 2023School of ManagementThis dissertation consists of three essays on ethical issues in natural language use for the design and implementation of artificial intelligence (AI) systems. Several ethical concerns regarding the use of AI to support human decision-making have emerged. The three studies highlight concerns related to confirmation bias and algorithmic fairness in both human and AI systems. Algorithmic fairness refers to efforts to ensure that AI systems are designed and deployed in a manner that does not discriminate against a select group of people, specifically, underrepresented group members. Confirmation bias is characterized by making decisions based on existing beliefs. The dissertation examines these themes in the context of hiring and news veracity recommendations. In the first two essays of this dissertation, solutions to address algorithmic discrimination are proposed. In the third essay, strategies are proposed to impact fairness judgments about algorithms in the context of news veracity assessments to improve the acceptance of algorithmic recommendations. Essay 1. Human resources (HR) platforms using advanced information technology (IT) and artificial intelligence (AI) tools to assist with HR tasks are rapidly being adopted. Considering these platforms have the potential to remove human bias (e.g., confirmation bias) from personnel selection by standardizing the evaluation process, as well as the potential to codify discrimination in seemingly objective algorithms, there is a need to better understand their impact on diversity, equity, and inclusion (DEI). To this end, we use a real-world dataset of 2,506 applicants’ interview responses on an HR platform to examine how applicants’ job-irrelevant dialect use (i.e., African American English; AAE) affects selection decisions. We find that (1) greater use of AAE reduces an applicant’s chances of getting hired, (2) the negative impact of AAE use is stronger in unstructured interview questions, and (3) that this negative effect is strongest among Black applicants. In addition, we find machine learning models are more discriminatory against Black applicants when predicting selection outcomes with unstructured questions, and removing AAE words from machine learning models can enhance their fairness by making them less discriminatory against Black applicants. These findings have important implications for personnel selection processes in organizations seeking to improve social justice and algorithmic fairness.
Essay 2. Machine learning (ML) and artificial intelligence (AI) are increasingly playing a role in personnel assessment and selection. However, the use of ML and AI to support human resources (HR) tasks in organizations has a short history rife with significant challenges. For example, algorithms may reinforce existing inequalities in human processes - which may be a result of confirmation bias - when the data used to train them reflects such inequalities. To contribute insights to address this challenge, we propose a loose coupling algorithmic fairness framework that utilizes multiple sources of ground truth labels (i.e., decentralization) thereby decoupling the relationship between predictors and target outcomes (i.e., reducing directness) in ML pipelines. We use a real-world dataset of 2,506 applicants' interview responses to predict candidate selection on a hiring platform to investigate this approach. Models based on our framework estimate the similarity of candidate responses to interview questions to human-validated exemplar answers collected from HR websites, and compared to directly predicting historical hiring decisions, the proposed approach (1) leads to fairer outcomes for underrepresented group members, and (2) is less able to predict candidate race. These results highlight how to enhance the procedural and distributive fairness of ML and AI systems in organizations through human and algorithmic collaboration.
Essay 3. Advancements in information and internet technologies have contributed to the pervasiveness of information sharing on platforms such as social media, and blogs, among others. Unfortunately, this has led to the effortless dissemination of false information (or Fake news) on these platforms. While several fact-checking tools and resources have been developed to prevent the spread of Fake News, existing studies have found that these tools are not always effective due to cognitive biases such as confirmation bias. That is, people do not accept the recommendation of these fact-checking tools if the recommendations do not align with their beliefs. To address this challenge, the third study proposes a conceptual framework explaining how the similarity between AI tools and users, users’ perceived fairness of AI tools, and the autonomy AI tools provide users during interaction, are key factors that can improve the acceptance of algorithmic advice in the presence of confirmation bias. The study espouses that similarity will lead to perceived fairness, while perceived fairness will improve the acceptance of algorithmic advice. Additionally, the relationship between the perceived fairness and the decision to accept or reject an algorithm’s recommendation will be moderated by the extent of autonomy provided to the user by the AI tool during the recommendation task. Future work will focus on gathering data to test the arguments advanced in the study.Ph
Exploring the association of glycemic control and tissue functionality in diabetes: a mechanistic and data analytic approach
No full textMay 2023School of EngineeringDiabetes is a global epidemic affecting approximately 537 million people worldwide, which can lead to multiple complications. Non-enzymatic glycation (NEG), a post-translational modification process that is highly accelerated in diabetes-related persistent hyperglycemia (high blood glucose concentration) and oxidative stress, has been identified to be one of the important causes for multiple tissues and organs damage. One of the comorbid conditions is bone fragility and individuals with type 1 diabetes (T1D) and type 2 diabetes (T2D) have higher risk of fracture than people without diabetes. This elevation of fracture risk in diabetes cannot be solely explained by bone mass or fall incidences and is currently underestimated by standard of care tools that measure bone mineral density (BMD). Therefore, the bone quality in T1D and T2D independently of BMD needs to be investigated to understand the pathogeneses and mechanisms of the increased bone fragility in diabetes. With the comprehensive understanding of the mechanisms which diabetes induces skeletal fragility, the risk of fracture in people with diabetes can be properly evaluated and managed. Similarly to bone fragility, diabetes could also significantly affect the severity and clinical outcomes of several respiratory diseases, for example, COVID-19. At the time when the hospital and intensive care units (ICUs) were at full capacity/overcrowded with COVID-19 cases, such as seen during the early stage of the pandemic, proper evaluation of diabetes severity and its impact for COVID-19 outcomes at the time of admission could help proper monitoring and management of subsequent care. In this work, we discuss the examination of the cortical and trabecular bone compositional and organizational quality at the material level under the influence of T1D and T2D. Using microindentation, Raman spectroscopy, Fourier transform infrared spectroscopy, and small angle X-ray scattering, we were able to identify and measure the alterations in T2D mineral crystal nanoscale morphology and organic matrix composition, as well as changes in T1D bound-water content. Using data analytics, the association between longitudinal HbA1c, commonly used medications, and two-year fracture risk was assessed in a large cohort with 157,439 T2D individuals, providing important clinical input on assessment, management and reduction of fracture risk in people with T2D through monitoring and management of the longitudinal glycemic control, and the use of metformin and/or DPP4 inhibitors. In a different cohort, we also identified that circulating biomarkers of glycation and oxidative stress including carboxymethyl-lysine (CML), pentosidine, and f2-isoprostanes are independently associated with fracture risk among T2D individuals. The role of these circulating biomarkers in T2D fracture risk was also determined at given BMD levels. Furthermore, using similar data analytic methods in another large T2D cohort, we identified that the two- to three-year longitudinal glycemic control is most significantly associated to COVID-19-related severity in people with T2D, and the combined use of metformin and insulin, as well as the use of corticosteroids are effective to prevent T2D patients from becoming critically ill from COVID-19.Ph
Sensorless frame-to-volume multimodal image fusion via deep learning
No full textDecember 2022School of EngineeringProstate cancer is the leading cause of death for men in the western world. The fusion of transrectal ultrasound (US) and magnetic resonance (MR) images for guiding the biopsy can facilitate the clinical diagnosis of prostate cancer. Intra-operative US scan provides real-time 2D prostate images, and pre-operative MR image volume offers good sensitivity and specificity for lesion localization. During a biopsy procedure, clinicians semi-manually set up the MR-US correspondence to superimpose the pre-identified lesions from MR onto the real-time US frames for navigation. Although there exist several image fusion methods, we are still facing a few technical challenges to make this technology more accessible for disadvantaged population. For example, current tracking-based approaches require motion sensors to be attached to the US probes which increases the hardware costs. These methods usually require clinical experts to manually align the US images with MR images for setting up the cross-modality correspondence, which significantly limits the patient throughput. In this work, we propose a solution for real-time multi-modality 2D-US frame to 3D-MR fusion which is fully automated by DL techniques. We develop this project aiming to remove the hardware constraint and perform automatic multi-modal cross-dimensional image fusion with minimum human supervision. The proposed method can largely reduce the hardware complexity while increase the inference speed and accuracy. The innovation of this project is three fold: (1) We propose to automatically reconstruct 3D US volume from 2D frames without using any external US probe tracking devices. The trained neural networks can reveal the inter-frame relationship by extracting context information between neighboring frames in a US sweep video. Without the tracking devices, our sensorless volume reconstruction allows clinicians to move the probe with less constraint without the concerns of blocking tracking signals. Additionally, it also reduces hardware costs. We develop a systematic pipeline for the task of 3D US reconstruction, including data acquisition/preprocessing, model design and training, volume reconstruction performance evaluation, and learning capacity analysis. (2) We introduce a deep learning based method for registering 2D US frames and 3D US volume to bridge the dimensional gap for the US/MR fusion. During this process, we combine both the video context from real-time US scans and volumetric information from 3D reconstructed US volumes to estimate the location of the current 2D US frame in 3D space. While existing methods require external tracking devices to map the location of a US frame in the reconstructed US volume, in our developed technology, such mapping can be accomplished fully automatically without additional hardware. (3) We further bridge the image modality gap by proposing an automatic registration method between the reconstructed 3D US volume and 3D pre-operative MR volume. Unlike traditional image registration, our forward-pass method does not require iterative optimization, thus greatly reducing computational time. Considering all previous image correspondences, including 2D-US to 3D-US and 3D-US to 3D-MR, we can propagate the transformation to achieve 2D-US to 3D-MR registration without hardware constraints. We validate our method on a clinical dataset with 618 subjects and test its potential on real-time 2D-US to 3D-MR fusion tasks. The proposed frame-to-volume multi-modal image fusion pipeline achieved the average target navigation error of 1.93 mm with a registration speed of 5 to 14 frames per second.Ph
Intrinsic cell chirality of human embryonic stem cells and derived lineages
No full textAugust 2018School of EngineeringLeft-right (LR) asymmetry is an imperative part of embryonic development. The disruption of LR axis development can result in defects in organ morphology, positioning and physiological functions. The mechanisms behind LR symmetry breaking are under debate in the field with theories ranging from intracellular cytoskeletal chirality in the single cell embryo to morphogen gradients induced by cilial flow to unifying theories wherein LR asymmetry is established through an interplay of many different models. Recent studies have demonstrated that LR asymmetry at the cellular level, i.e., intrinsic cell chirality, plays an important role in embryonic development. The goal of this dissertation is to explore the possible role of intrinsic cellular chirality in the development of LR asymmetric organs and mechanisms behind its action. Current studies in embryonic development rely heavily on animal models, which do not always recapitulate human development. We propose to develop and utilize an in vitro model system consisting of human embryonic stem cells (hESCs) and a biomaterial-based chirality assay to assess intrinsic cell chirality and study embryonic LR asymmetry development. Evaluation of the differentiation of hESCs into heart, gut and brain tissues shows lineage-dependent chiral biases in 3D cell rotation at various stages of differentiation. The chiral biases correlate with the directionality of asymmetric looping during in vivo organ development. Pre-treatments of hESCs with activators and inhibitors of key LR asymmetry signaling pathways, Nodal and canonical Wnt, demonstrate that these pathways regulate inherent cellular chirality. Taken together, this research suggests that cell chirality regulated by developmental signaling pathways determines the LR asymmetry of organ development.Ph