Learning Representations for Limited and Heterogeneous Medical Data
Learning Representations for Limited and Heterogeneous Medical Data

Author: Wei-Hung Weng

Publisher:

Published: 2022

Total Pages: 0

ISBN-13:

DOWNLOAD EBOOK

Data insufficiency and heterogeneity are challenges of representation learning for machine learning in medicine due to the diversity of medical data and the expense of data collection and annotation. To learn generalizable representations from such limited and heterogeneous medical data, we aim to utilize various learning paradigms to overcome the issue. In this dissertation, we systematically explore the machine learning frameworks for limited data, data imbalance, and heterogeneous data, using cross-domain learning, self-supervised learning, contrastive learning, meta-learning, multitask learning, and robust learning. We present studies with different medical applications, such as clinical language translation, ultrasound image classification and segmentation, medical image retrieval, skin diagnosis classification, pathology metadata prediction, and lung pathology prediction. We first focus on the limited data problem, which is common in medical domains. We learn cross-domain representations for clinical language translation with limited and unpaired medical language corpora using unsupervised embedding space alignment with identical anchors for word translation, and conduct sentence translation using statistical language modeling. Using metrics of clinical correctness and readability, the developed method outperforms a dictionary-based algorithm in both word- and sentence-level translation. For learning better data representations of limited numbers of ultrasound images, we then adopt the self-supervised learning technique and integrate the corresponding metadata as a multimodal resource to introduce inductive biases. We find that the representations learned by the developed approach yield better downstream task performance, such as ultrasound image quality classification and organ segmentation, compared with the standard transfer learning methods. Next, we zoom into the data imbalance problem. We explore the utility of contrastive learning, specifically the Siamese network, to learn representations from an imbalanced fundoscopic imaging dataset for diabetic retinopathy image retrieval. Compared with the standard supervised learning setup, we obtain comparable but interpretable results using the representations learned from the Siamese network. We also utilize meta-learning for skin disease classification with an extremely imbalanced long-tailed skin image dataset. We find that model ensemble with meta-learning models and models trained with conventional class imbalance techniques yields better prediction performance, especially for rare skin diseases. Finally, for heterogeneous medical data, we develop a multimodal multitask learning framework to learn a shared representation for pathology metadata prediction. We use the multimodal fusion technique to integrate the slide image, free text, and structured metadata, and adopt a multitask objective loss to introduce the inductive bias while learning. This yields better prediction power than the standard single-modal single-task training setup. We also apply robust training techniques to learn representations that can tackle a distributional shift across two chest X-ray datasets. Compared with standard training, we find that robust training provides better tolerance when the shift exists, and learns a robust representation for lung pathology prediction. The investigation in this dissertation is not exhaustive but it introduces an extensive understanding of utilizing machine learning in helping clinical decision making under the limited and heterogeneous medical data setting. We also provide insights and caveats to motivate future research directions of machine learning with low-resource and high-dimensional medical data, and hope to make a positive real-world clinical impact.

Efficient Representation Learning for Longitudinal Data in Healthcare Applications
Efficient Representation Learning for Longitudinal Data in Healthcare Applications

Author: Shayan Fazeli

Publisher:

Published: 2023

Total Pages: 0

ISBN-13:

DOWNLOAD EBOOK

Efficient utilization of longitudinal observations is a crucial component in proposing machine learning solutions to problems in healthcare. The temporal nature of numerous problems in this domain, such as understanding fluctuations in physiological signals through time pertinent to health status, renders this avenue of research particularly important for the intersection of Health Analytics and Artificial Intelligence (AI). In the healthcare domain, compared to other fields such as Computer Vision or Natural Language Processing, the data is often available in limited quantities. Additionally, reliable supervision signals for training inference pipelines are scarce. Furthermore, some data modalities and domains are critical to health applications which are, at the same time, considerably less investigated in machine learning research. These challenges are essential bottlenecks to address in improving the efficacy and usability of machine learning-based healthcare solutions. In this dissertation, we investigate the role of longitudinal data in medical and health applications in various related domains. Namely, we consider the domains of 1) Physical Health: Representation learning for monitoring the physical health of an individual useful for in-patient and out-patient setups, with examples being physiological signals, activity data, and posture tracking. 2) Electronic Health Records: The multi-modal and temporal reports in different time resolutions on patients' health trajectories 3) Mental Health: Efficient multi-resolution monitoring of stress and anxiety as an example use-case with important applications, and 3) Public Health: Pandemic analytics and representation of population-level spatio-temporal health data. We suggest novel techniques to address the primary challenges in each task efficiently. In our solutions, we use approaches such as optimizing self-supervised contrastive objectives, knowledge transfer, and adversarial training so as to minimize the reliance on accurate and large-scale supervision signals. We discuss the empirical validation of our suggested solutions and shed light on some of the key future research directions.

Representation Learning in Multi-dimensional Clinical Timeseries for Risk and Event Prediction
Representation Learning in Multi-dimensional Clinical Timeseries for Risk and Event Prediction

Author: Marzyeh Ghassemi

Publisher:

Published: 2017

Total Pages: 114

ISBN-13:

DOWNLOAD EBOOK

There are major practical and technical barriers to understanding human health, and therefore a need for methods that thrive on large, complex, noisy data. In this work, we present machine learning methods that distill large amounts of heterogeneous health data into latent state representations. These representations are then used to estimate risks of poor outcomes, and response to intervention in multivariate physiological signals. We evaluate the reduced latent representations by 1) establishing their predictive value in important clinical tasks and 2) showing that the latent space representations themselves provide useful insight into underlying systems. In particular, we focus on case studies that can provide evidence-based risk assessment and forecasting in settings with guidelines that have not traditionally been data-driven. In this thesis we evaluate several methods to create patient representations, and use these features to predict important outcomes. Representation learning can be thought of as a form of phenotype discovery, where we attempt to discover spaces in the new representation that are markers of important events. We argue that these latent representations are useful markers when they 1) create better prediction results on outcomes of interest, and 2) do not duplicate features that are currently known bio-markers. We present four case studies of learning representations, and evaluate the representations on real predictive tasks. First, we create forward-facing prediction models using baseline clinical features, and those from a Latent Dirichlet Allocation (LDA) model trained with clinical progress notes. We then evaluate the per-patient latent state membership to predict mortality in an intensive care setting as time moves forward. Second, we use non-parametric Multi-task Gaussian Process (MTGP) hyper-parameters as latent features to estimate correlations within and between signals in sparse, heterogeneous time series data. We evaluate the hyper-parameters for forecasting missing signals in traumatic brain injury patients, and predicting mortality in intensive care unit patients. Third, we train switching-state autoregressive models (SSAMs) to model the underlying states that emit patient vital signs over time. We evaluate the time-specific latent state distributions as features to predict vasopressor onset and weaning in intensive care unit patients. Finally, we use statistical and symbolic features extracted from wearable ambulatory accelerometers (ACC) mounted to the neck to classify patient pathology, and stratify patients’ risk of voice misuse. We evaluate the utility of both statistically generated features and symbolic representations of glottal pulses towards patient classification.

Deep Learning for Medical Image Analysis
Deep Learning for Medical Image Analysis

Author: S. Kevin Zhou

Publisher: Academic Press

Published: 2023-12-01

Total Pages: 544

ISBN-13: 0323858880

DOWNLOAD EBOOK

Deep Learning for Medical Image Analysis, Second Edition is a great learning resource for academic and industry researchers and graduate students taking courses on machine learning and deep learning for computer vision and medical image computing and analysis. Deep learning provides exciting solutions for medical image analysis problems and is a key method for future applications. This book gives a clear understanding of the principles and methods of neural network and deep learning concepts, showing how the algorithms that integrate deep learning as a core component are applied to medical image detection, segmentation, registration, and computer-aided analysis. · Covers common research problems in medical image analysis and their challenges · Describes the latest deep learning methods and the theories behind approaches for medical image analysis · Teaches how algorithms are applied to a broad range of application areas including cardiac, neural and functional, colonoscopy, OCTA applications and model assessment · Includes a Foreword written by Nicholas Ayache

On Leveraging Representation Learning Techniques for Data Analytics in Biomedical Informatics
On Leveraging Representation Learning Techniques for Data Analytics in Biomedical Informatics

Author: Xi Hang Cao

Publisher:

Published: 2019

Total Pages: 124

ISBN-13:

DOWNLOAD EBOOK

Representation Learning is ubiquitous in state-of-the-art machine learning workflow, including data exploration/visualization, data preprocessing, data model learning, and model interpretations. However, the majority of the newly proposed Representation Learning methods are more suitable for problems with a large amount of data. Applying these methods to problems with a limited amount of data may lead to unsatisfactory performance. Therefore, there is a need for developing Representation Learning methods which are tailored for problems with ``small data", such as, clinical and biomedical data analytics. In this dissertation, we describe our studies of tackling the challenging clinical and biomedical data analytics problem from four perspectives: data preprocessing, temporal data representation learning, output representation learning, and joint input-output representation learning. Data scaling is an important component in data preprocessing. The objective in data scaling is to scale/transform the raw features into reasonable ranges such that each feature of an instance will be equally exploited by the machine learning model. For example, in a credit flaw detection task, a machine learning model may utilize a person's credit score and annual income as features, but because the ranges of these two features are different, a machine learning model may consider one more heavily than another. In this dissertation, I thoroughly introduce the problem in data scaling and describe an approach for data scaling which can intrinsically handle the outlier problem and lead to better model prediction performance. Learning new representations for data in the unstandardized form is a common task in data analytics and data science applications. Usually, data come in a tubular form, namely, the data is represented by a table in which each row is a feature (row) vector of an instance. However, it is also common that the data are not in this form; for example, texts, images, and video/audio records. In this dissertation, I describe the challenge of analyzing imperfect multivariate time series data in healthcare and biomedical research and show that the proposed method can learn a powerful representation to encounter various imperfections and lead to an improvement of prediction performance. Learning output representations is a new aspect of Representation Learning, and its applications have shown promising results in complex tasks, including computer vision and recommendation systems. The main objective of an output representation algorithm is to explore the relationship among the target variables, such that a prediction model can efficiently exploit the similarities and potentially improve prediction performance. In this dissertation, I describe a learning framework which incorporates output representation learning to time-to-event estimation. Particularly, the approach learns the model parameters and time vectors simultaneously. Experimental results do not only show the effectiveness of this approach but also show the interpretability of this approach from the visualizations of the time vectors in 2-D space. Learning the input (feature) representation, output representation, and predictive modeling are closely related to each other. Therefore, it is a very natural extension of the state-of-the-art by considering them together in a joint framework. In this dissertation, I describe a large-margin ranking-based learning framework for time-to-event estimation with joint input embedding learning, output embedding learning, and model parameter learning. In the framework, I cast the functional learning problem to a kernel learning problem, and by adopting the theories in Multiple Kernel Learning, I propose an efficient optimization algorithm. Empirical results also show its effectiveness on several benchmark datasets.

Graph Representation Learning
Graph Representation Learning

Author: William L. William L. Hamilton

Publisher: Springer Nature

Published: 2022-06-01

Total Pages: 141

ISBN-13: 3031015886

DOWNLOAD EBOOK

Graph-structured data is ubiquitous throughout the natural and social sciences, from telecommunication networks to quantum chemistry. Building relational inductive biases into deep learning architectures is crucial for creating systems that can learn, reason, and generalize from this kind of data. Recent years have seen a surge in research on graph representation learning, including techniques for deep graph embeddings, generalizations of convolutional neural networks to graph-structured data, and neural message-passing approaches inspired by belief propagation. These advances in graph representation learning have led to new state-of-the-art results in numerous domains, including chemical synthesis, 3D vision, recommender systems, question answering, and social network analysis. This book provides a synthesis and overview of graph representation learning. It begins with a discussion of the goals of graph representation learning as well as key methodological foundations in graph theory and network analysis. Following this, the book introduces and reviews methods for learning node embeddings, including random-walk-based methods and applications to knowledge graphs. It then provides a technical synthesis and introduction to the highly successful graph neural network (GNN) formalism, which has become a dominant and fast-growing paradigm for deep learning with graph data. The book concludes with a synthesis of recent advancements in deep generative models for graphs—a nascent but quickly growing subset of graph representation learning.

Deep Learning in Medical Image Analysis
Deep Learning in Medical Image Analysis

Author: Gobert Lee

Publisher: Springer Nature

Published: 2020-02-06

Total Pages: 184

ISBN-13: 3030331288

DOWNLOAD EBOOK

This book presents cutting-edge research and applications of deep learning in a broad range of medical imaging scenarios, such as computer-aided diagnosis, image segmentation, tissue recognition and classification, and other areas of medical and healthcare problems. Each of its chapters covers a topic in depth, ranging from medical image synthesis and techniques for muskuloskeletal analysis to diagnostic tools for breast lesions on digital mammograms and glaucoma on retinal fundus images. It also provides an overview of deep learning in medical image analysis and highlights issues and challenges encountered by researchers and clinicians, surveying and discussing practical approaches in general and in the context of specific problems. Academics, clinical and industry researchers, as well as young researchers and graduate students in medical imaging, computer-aided-diagnosis, biomedical engineering and computer vision will find this book a great reference and very useful learning resource.

Learning Inductive Representations of Biomedical Data
Learning Inductive Representations of Biomedical Data

Author: Samuel G. Finlayson

Publisher:

Published: 2020

Total Pages: 189

ISBN-13:

DOWNLOAD EBOOK

Representation learning with neural networks has catalyzed rapid progress in biomedical pattern recognition. This progress, however, has generally been limited to domains where data are abundant, richly structured, and stable. In contrast, much of biomedicine is marked by limited and poorly structured data and by highly dynamic deployment environments. In particular, many of the most compelling problem areas in biomedicine involve the "long tails" of rare diseases and rare events. In this thesis, I confront the challenge of learning data representations whose utility can extend into dynamic and data-poor biomedical domains. I do so through three primary projects: First, I present a novel method for representation learning with subgraphs. This method, called Subgraph Neural Networks (Sub-GNN), learns disentangled representations of subgraph structure, neighborhood, and position through property-aware routing channels. The work is motivated by the desire for methods that can better contextualize patient phenotypes (encoded as subgraphs) into the broader context of biomedical knowledge, which could allow for better diagnostic generalization to novel disorders involving previously unseen phenotypes. Subgraph neural networks provide a principled framework for doing just this, by leveraging the relational inductive biases of the underlying knowledge graph while still respecting subgraphs as independent entities. Next, I present an approach to learning coordinated representations of small molecules and their associated transcriptional signatures. This approach extends a popular paradigm for drug development (known as connectivity mapping) to operate inductively, making predictions involving drugs that have not previously been experimentally assayed. I benchmark the performance of this approach, studying the circumstances under which it can and cannot achieve strong performance. Finally, I present an analysis of the clinical challenges posed by dataset shift, the phenomenon in which the input data to a deployed machine learning algorithm become mismatched with its training data. After introducing the problem of general dataset shift, I turn to a special case—adversarial examples—which reflect the worst-case generalization conditions for a machine learning system. I then build and test the representational robustness of three high-accuracy machine learning systems, constructing adversarial examples that cause their accuracy to drop to 0% on data that is imperceptibly different from the training data. I discuss the implications of these findings for clinical machine learning, offering specific regulatory recommendations. I conclude my thesis with lessons learned from these projects, and provide an extensive appendix with three additional smaller-scale projects that branched off of my research.

Developing a Protocol for Observational Comparative Effectiveness Research: A User's Guide
Developing a Protocol for Observational Comparative Effectiveness Research: A User's Guide

Author: Agency for Health Care Research and Quality (U.S.)

Publisher: Government Printing Office

Published: 2013-02-21

Total Pages: 236

ISBN-13: 1587634236

DOWNLOAD EBOOK

This User’s Guide is a resource for investigators and stakeholders who develop and review observational comparative effectiveness research protocols. It explains how to (1) identify key considerations and best practices for research design; (2) build a protocol based on these standards and best practices; and (3) judge the adequacy and completeness of a protocol. Eleven chapters cover all aspects of research design, including: developing study objectives, defining and refining study questions, addressing the heterogeneity of treatment effect, characterizing exposure, selecting a comparator, defining and measuring outcomes, and identifying optimal data sources. Checklists of guidance and key considerations for protocols are provided at the end of each chapter. The User’s Guide was created by researchers affiliated with AHRQ’s Effective Health Care Program, particularly those who participated in AHRQ’s DEcIDE (Developing Evidence to Inform Decisions About Effectiveness) program. Chapters were subject to multiple internal and external independent reviews. More more information, please consult the Agency website: www.effectivehealthcare.ahrq.gov)