Dartmouth Institute for Health Policy and Clinical Practice
Dartmouth Digital Commons (Dartmouth College)Not a member yet
8214 research outputs found
Sort by
Facilitating Design of Effective, Scalable Digital Mental Health Care
Get PDFThis thesis tackles the challenges of designing effective, scalable digital mental health care solutions by integrating psychological research with human-computer interaction (HCI) principles. An exploration of self-care technologies (SCTs) for mental health is conducted to highlight challenges to comparative analysis, like fragmented research outcomes. Two research questions are posed to investigate how mental health interventions can be designed to be effective and overcome practical challenges to implementation. A theoretical framework in clinical psychology is adapted to bridge the gap between psychology and HCI. A methodology for conducting standardized, structured comparative analysis of mental health interventions is identified. The identified method is used to demonstrate the study of a conventional in-person social anxiety intervention. This study is followed by an analysis of a corresponding, digitized social anxiety intervention, highlighting how interventions can change through digitization.
Further, the influence of intervening third variables is observed, and plausible solutions to implementation challenges are suggested. The discussion further illustrates the importance of HCI principles in improving the utilization of interventions through user engagement and provides theoretical grounding for incorporating principles into interventions.
A replicable model for designing future mental health interventions is presented, and offers actionable steps to close the gap between theoretical promise and real-world impact in digital mental health care. A website-based tool is created to facilitate data collection and allow data to be queried post-analysis. This work provides a structure for conducting future investigations of digital health technologies. This dissertation concludes that a first-principles approach can improve mental health care interventions and support targeted development of precise solutions through digital health tools
MULTIMODAL MACHINE LEARNING TO ENHANCE PATIENT PROGNOSIS & MANAGEMENT IN CANCER CARE
Get PDFPrecision medicine is a data-driven approach that tailors treatments to individualized need. Emerging in the late twentieth century, it emphasizes customizing therapy based on each person’s unique characteristics and became a major national focus in the US starting in 2015.
Cancer remains a leading cause of death worldwide. With high mortality rates and resource- intensive treatment process, it burdens the healthcare systems. One of the inherent challenges in cancer care is its biological heterogeneity, making diagnosis and treatment planning highly complex and calls for personalized cancer management.
Today, diverse data sources are available to support cancer management and treatment decisions. Despite this, cancer management largely depend on manual interpretation of medical images and consideration of a limited number of risk factors. Several challenges remain: (a) it is infeasible for clinicians to comprehensively integrate all available data for decision-making; (b) patients respond differently to the same treatments, leading to uncertainty in outcomes and suboptimal prognoses; and (c) cancer diagnosis and care are resource intensive and heavily reliant on specialist expertise, which may not be readily accessible in resource-limited seƫngs.
This dissertation presents three novel pipelines that advance precision oncology through multimodal deep learning frameworks.
The first pipeline targets non-small cell lung cancer (NSCLC), developing a multimodal model that combines histology whole-slide images, clinical data, and next-generaƟon sequencing data to predict resistance to osimertinib. The model demonstrated strong performance in risk prediction, stratification, and interpretability, offering clinical value in decision-making.
The second pipeline introduces a contrasƟve learning-enhanced framework for survival prediction in lower-grade glioma patients. By integrating histopathology images, somatic mutation, and clinical variables, the model achieved a high c-index and robust risk stratification, outperforming models without contrastive learning.
The third pipeline addresses diagnostic and grading challenges in bone lesion classification using a contrastive learning-enhanced model trained with histology and X-ray images. It achieved strong performance using only X-ray images at test time, reducing reliance on invasive biopsy and histology imaging. Together, these studies demonstrate the value of contrastive and multimodal learning in improving clinical prediction, contributing novel model architectures, training strategies, and applications that expand the scope and impact of precision cancer care
POTSDAM: Pareto Optimization Targeting Security, Data, and Mediation
Get PDFGiven the growing amount and variety of data handled by modern systems, it is crucial to guarantee the accuracy and protection of input data without errors or malicious intentions. The need to improve security in software programs often conflicts with the assurance of maximum performance, making developers and maintainers hesitant to incorporate more testing.
LangSec (Language-Theoretic Security) is a security approach that treats input validation as a formal language recognition problem, ensuring that only well-defined, unambiguous inputs are processed to eliminate exploitable parsing flaws. This dissertation explores integrating LangSec principles with Pareto optimization to enhance safety and robustness in digital environments. We introduce three novel methods for embedding LangSec techniques into software applications and data streams, striking a balance between security and performance. These approaches offer practical, adaptable solutions for improving the security and efficiency of existing software and network protocols, advancing the resilience of modern computing systems
SECURE AND TRUSTWORTHY AI/ML
No full textMachine learning (ML) models and artificial intelligence (AI) systems are widely vulnerable to different adversarial and privacy attack vectors. Adversaries with different capabilities target AI/ML systems to break down overall functionality (i.e., adversarial attacks) or leak sensitive information (i.e., privacy attacks). To ensure a trustworthy AI/ML system, it is crucial to characterize these vulnerabilities and develop defense techniques. This dissertation comprises five published papers and one draft paper that focus on analyzing vulnerabilities of AI/ML systems and introducing innovative techniques for defenses. The first dissertation work (IEEE CSF) focuses on systematizing novel model inversion (MI) privacy attacks against ML models, including their attack taxonomy, foundational aspects, challenges, and future directions. The second dissertation work (Springer SaSeIoT) explores the use of AI for robust user authentication (e.g., validating users under attacks) in the context of an IoT device, leveraging continuous biometrics, i.e., breathing patterns. The third dissertation work (IEEE SaTML) investigates ML model vulnerabilities against privacy attacks (e.g., tabular data) under a realistic setup with limited adversarial capabilities and shows that ML models can indeed leak sensitive information even in those restricted scenarios.
These new ML attack vectors then inspired a suite of novel ML adversarial and privacy attack defenses. Specifically, the fourth dissertation work (IEEE ICASSPW) focuses on the mitigation of state-of-the-art (SOTA) audio adversarial attacks through multilayer lateral completion networks. The fifth dissertation work (ECCV) introduces novel defenses against MI attacks based on the sparse coding architecture (SCA), which shows 1.1-18.3 times more robustness against MI attacks while not significantly compromising accuracy. This novel work inspires the last piece of dissertation work to further improve MI defense by designing privacy-preserving modeling techniques to systematically eliminate highly sensitive features during training to achieve even better robustness
Characterization of the Erv41-Erv46 complex as a retrograde receptor for misfolded secretory proteins
No full textThe Endoplasmic Reticulum (ER), the site of secretory protein biosynthesis
provides a favorable environment to promote polypeptide folding, protein
oligomerization and export. Although the ER contains many chaperones and other
factors that assist in protein folding, this process is error prone. Eukaryotes have
evolved protein quality control checkpoints to maintain homeostasis in the secretory
pathway, however these processes are imperfect and may result in proteotoxicity
when toxic levels or aggregates of proteins arise. Mutations in proteins involved in
trafficking in the early secretory pathway are closely associated with disease states.
Therefore, it is of great interest to further understand how misfolded proteins and the
potentially toxic effects of such proteins are mitigated in the early secretory pathway.
The Erv41-Erv46 complex is a retrograde cargo receptor that has been
shown to retrieve ER resident proteins from the Golgi back to the ER. Intriguingly,
other studies have shown that deletion of the Erv41 subunit which destabilizes the
complex results in increased turnover of a model misfolded substrate, CPY* and
unfolded protein response (UPR) down-regulation. With this knowledge, we propose
that Erv41-Erv46, in addition to retrieval of ER resident cargo, also functions in
retrograde trafficking of misfolded proteins. Through this body of work, we support a
model in which Erv41-Erv46 preferentially retrieves misfolded proteins with luminal
lesions from the Golgi to the ER. Cycloheximide chase and immunoprecipitation (IP)
experiments show that Erv41-Erv46 deficient cells fail to retrieve CPY*, resulting in
its vacuolar degradation or secretion. Additionally, Erv41-Erv46 directly and
preferentially binds this misfolded protein over its native, folded counterpart, CPY in
IP and in vitro reconstitution assays. These data and findings from this study
support a model in which Erv41-Erv46 directly binds misfolded proteins at the Golgi
for return to the ER
Modeling the Cystic Fibrosis Intestinal Environment to Study Microbial Interactions
Get PDFThe microbiome of the cystic fibrosis (CF) intestine is altered significantly in structure and function beginning in early life. Key taxa associated with health, digestion and immune-training are reduced in abundance, including but not limited to Bacteroides, Faecalibacterium, Akkermansia, Ruminococcus and other short-chain fatty acid producers. Conversely, there is an expansion of microbes with pathogenic potential, such as Escherichia, Veillonella, Pseudomonas and Staphylococcus. The shifts in microbiota composition direct community-wide changes with differences in beta diversity by genotype, whereby children with CF (cwCF) exhibit significantly different gut microbial compositions compared to nonCF children. Additionally, shifts in microbiota structure often translate to shifts in microbiota function, with microbial metabolism as a large focus in most studies. Taken together, an imbalance in the intestinal microbiome as observed early in life in cwCF contributes to an observed delay in maturation of their intestinal microbiome compared to nonCF cohorts, underscoring the need to identify key drivers of microbial imbalances which may or may not overlap with other intestinal diseases (i.e., inflammatory bowel disease).
In this thesis, I introduce a novel growth medium to represent the nutritional environment of the CF colon. Through the implementation of this medium in growth assays, I show that physiological features relevant to the CF intestine (i.e., excess mucin, fats, bile, altered pH, alternative electron acceptors, inflammatory by-products, antibiotic perturbation) are sufficient in directing community-wide changes in microbiome structure that mimic CF intestinal microbiota signatures. Furthermore, these taxonomical shifts translate to altered microbial metabolism, with short chain fatty acid abundance as proxy.
I apply this medium to study the mechanism of altered microbial interactions, with a focus on Escherichia coli and Bacteroides vulgatus. In vitro growth assays analyzed via linear regression identify significant drivers of dysbiosis in our system: excess bile and excess fat (in the form of glycerol). Genetic mutants of E. coli involved in microbial antagonism of B. vulgatus in the context of glycerol reveal a significant role for colibactin biosynthesis, linking a genotoxin associated with colorectal cancer to signatures of gut microbial dysbiosis in children with CF.
The findings presented in this thesis begin to elucidate host physiological features and microbial genetic factors involved in modulating markers of dysbiosis across CF cohorts. While many altered physiologies and treatments inherent to the CF genotype have been correlated with microbial imbalances, there are limited mechanistic studies in the field. Development and implementation of the novel in vitro medium introduced in this thesis sets the stage for future in vitro studies to further our understanding of CF gut biology
Yes, Instagram Has Led to Overcrowding in Yosemite. That\u27s OK: A Professional Climber Reflects on the Bad and Good of a Tourism Spectacle
Get PDFSocial media has Instagrammed the outdoors but also cultivated community. The crowds who flock to the park in February to see Horsetail Fall deserve to enjoy public lands, too