Tutorial on
Mobility and Health: Can We Use Wireless Sensors and Mobility Data for Health Assessment?
Instructor
|
Hesham Ali
College of Information Science and Technology, University of Nebraska at Omaha
United States
|
|
|
|
Brief Bio
Hesham H. Ali is a Professor of Computer Science and the director of the University of Nebraska Omaha (UNO) Bioinformatics Core Facility. He served as the Lee and Wilma Seemann Distinguished Dean of the College of Information Science and Technology at UNO between 2006 and 2021. He has published numerous articles in various IT areas, including scheduling, distributed systems, data analytics, wireless networks, and Bioinformatics. He has been serving as the PI or Co-PI of several projects funded by NSF, NIH and Nebraska Research Initiative in the areas of data analytics, wireless networks and Bioinformatics. He has also been leading a Research Group that focuses on developing innovative computational approaches to model complex biomedical systems and analyze big bioinformatics data.
|
Abstract
Abstract
The last several years have witnessed major advancements in the development of sensor technologies and wearable devices with the goal of collecting various types of data in many application domains. Based on such technologies, many commercial products have swamped the market and found their way on the wrists, ancles and belts of many users. Although these developments are certainly welcomed, so much left to be done to take full advantage of the data gathered by such devices. The most critical missing component is the lack of advanced data analytics. In the case of health monitoring, like many aspects of healthcare, the focus has been primarily on producing devices with data collection capabilities rather than developing advanced models for analyzing the available data.
In this tutorial, we attempt to fill this gap by presenting new data analytics tools that connect mobility and heath. We utilize graph modeling and complex networks along with how to effectively employ population-based approaches in implementing such tools. We demonstrate how the proposed tools can be applied to analyze big mobility data and reveal new useful health-related features in several case studies. We also utilize graph-theoretic mechanisms to zoom in and out of the network models and extract different types of information at various granularity levels. The proposed approach paves the way towards a new decision support system that leads to new discoveries in Biomedical research and healthcare applications.
Keywords
Mobility data, health assessment, data analytics, wearable devices, preventive healthcare, graph modeling, complex networks.
Aims and Learning Objectives
The fields of Biomedical Informatics and big data analytics have been attracting a lot of attention in recent years. The use of wireless devices to collect various types of critical data continues to grow both in the commercial world as well as in the research domain. The impact of such devices remains limited though, primarily due to the lack of sophisticated data analytics tools to allow for the extraction of useful information out of the collected data. The proposed tutorial will address these issues with a particular focus on the following objectives:
1- Provide an overview of available wearable devices and research studies associated with the use of wireless sensors in domains related to mobility and healthcare.
2- Introduce the main ideas associated with obtaining mobility patterns or signatures using raw data collected from wearable devices and wireless sensors to fully characterize the mobility parameters and to assess to some degree the health level of individuals.
3- Introduce the basic concepts of using correlation and similarity networks to store, visualize and analyze data associated with different applications in the domain biomedical informatics and show the potential of using these networks as a key component of an advanced decision support system for healthcare.
4- Introduce the participants to how graph models and integrated networks can be developed using mobility data to estimate health levels of various groups. The goal of the proposed model is to classify health levels, track their health variability pattern, and predict potential health hazards.
5- Show how the proposed approach can be used specifically for the early diagnosis of health issues associated with early childhood development and aging conditions.
Target Audience
The tutorial is intended primarily for computational scientists who are interested in wireless networks and data analytics. It is also of interest to Biomedical and Engineering researchers since the focus of the main application domains of the proposed methodology is health informatics and engineering. In particular, the tutorial targets researchers interested in how mobile devices and wireless technologies can used to support the new direction of health care that focused on predictive and preventative approaches.
Prerequisite Knowledge of Audience
Biomedical scientists and engineers with some background in computational concepts who are interested in how new technologies can support health care and medical information systems represent another group of intended audience. Basic background in computer science and wireless networks would be helpful but not necessary. The main concepts will be introduced in a highly accessible manner.
Detailed Outline
The proposed tutorial is designed for a 1.5-hour session that could potentially be extended to a three-hour session time permitting. The tutorial focuses on the use of mobile devices and mobility data in health monitoring and assessment; introducing the concepts of mobility signatures developed using data collected from wireless sensors; using correlation networks and graph theoretic tools to properly analyze mobility data and assess health levels; and studying how correlation networks can be used to link mobility studies with biomedical research. The tutorial will also address how to use clustering algorithms and advanced graph theoretic tools to provide advanced big data analysis of the collected data used to build the correlation networks, and how to integrate different types of data including mobility data and genetic information to provide a comprehensive analysis for health data for individuals.
1. Survey of current wireless technologies in healthcare - Brief discussion on the various research studies and commercial wireless devises developed with the goal of monitoring health activities and measure various mobility parameters such as number of steps, distance covered, and active periods.
2. How to obtain mobility signatures using raw mobility data – Algorithms for classifying various daily activities using mobility data will be introduced and used to build the characterizing models of mobility signature. Such characterizing patterns can be used to accurately measure the level of mobility associated with each individual.
3. Big data analytics using correlation networks – New techniques for building correlation networks from sensor data collected from groups will be presented. Big Data analysis tools will be introduced to analyze the developed correlation networks and predict health levels of various cases with a focus on using such tools in predicting potential health problems.
4. How to use Mobility to assess healthcare – Correlation Networks for modeling integrating various types of data will be presented. The integration model represents potential next steps in healthcare in which various types of data will be used to establish an accurate picture associated with each person’s health.
Tutorial on
Graph Neural Networks and Their Applications in Bioinformatics
Instructor
|
Dong Xu
University of Missouri
United States
|
|
|
|
Brief Bio
Dong Xu is Curators’ Distinguished Professor in the Department of Electrical Engineering and Computer Science, with appointments in the Christopher S. Bond Life Sciences Center and the Informatics Institute at the University of Missouri-Columbia. He obtained his Ph.D. from the University of Illinois, Urbana-Champaign in 1995 and did two years of postdoctoral work at the US National Cancer Institute. He was a Staff Scientist at Oak Ridge National Laboratory until 2003 before joining the University of Missouri, where he served as Department Chair of Computer Science during 2007-2016 and Director of Information Technology Program during 2017-2020. Over the past 30+ years, he has conducted research in many areas of computational biology and bioinformatics, including single-cell data analysis, protein structure prediction and modeling, protein post-translational modifications, protein localization prediction, computational systems biology, biological information systems, and bioinformatics applications in human, microbes, and plants. His research since 2012 has focused on the interface between bioinformatics and deep learning. He has published more than 400 papers with more than 22,000 citations and an H-index of 76 according to Google Scholar. He was elected to the rank of American Association for the Advancement of Science (AAAS) Fellow in 2015 and American Institute for Medical and Biological Engineering (AIMBE) Fellow in 2020.
|
Abstract
Abstract
Graph neural network (GNN) has achieved great success in bioinformatics recently. GNN learns representations of the topological structures and distribution patterns within data in addition to the features of individual data points, which makes it applicable to diverse bioinformatics problems with excellent performance. In this talk, I will first introduce some basic shallow graph embedding methods, such as locally linear embedding, random walk embedding, DeepWalk, Node2Vec, and LINE. I will then explain Graph Convolutional Network (GCN) by covering Laplacian embedding, Chebyshev polynomial approximation (ChebNet), and spectral convolution. I will also discuss several GCN variants, including GraphSAGE, gated graph neural networks, message-passing neural networks, graph attention networks, and heterogeneous graph transformer. Finally, I will discuss various GNN tasks, such as node classification, edge inference, and graph property prediction. I will use some of our own GNN applications in bioinformatics as examples, including property prediction of small molecules and drugs, electronic medical record study, and single-cell data annotation.
Keywords
deep learning, graph neural network, single-cell data analysis, electronic medical record
Aims and Learning Objectives
Help bioinformatics researchers to adopt Graph Neural Networks in their research
Target Audience
Graduate students, postdocs, and other bioinformatics researchers in academia and industry
Prerequisite Knowledge of Audience
some background in machine learning knowledge and basic bioinformatics skills
Detailed Outline
1. Concepts
a. Geometric Deep Learning
b. What is Graph Neural Network
c. Why Graph Neural Network?
d. Brief History of Graph Neural Network
2. Shallow Graph Embedding
a. Locally Linear Embedding
b. Random Walk Embedding
c. DeepWalk
d. Node2Vec
e. LINE
3. Graph Convolutional Network (GCN)
a. Laplacian Embedding
b. Chebyshev Polynomial Approximation (ChebNet)
c. Spectral Convolution
4. GCN Variants
a. GraphSAGE
b. Gated Graph Neural Networks
c. Message-Passing Neural Networks
d. Graph Attention Networks
e. Heterogeneous Graph Transformer
5. GNN Tasks and Implementations
a. Node Classification
b. Edge Inference
c. Graph Property Prediction
6. Bioinformatics Application Examples
a. Property Prediction of Small Molecules and Drugs
b. Electronic Medical Record Study
c. Single-cell Data Annotation
7. Summary and Discussion
Tutorial on
Open Reproducible Respiratory Electromyogram Analysis
Instructors
|
Candace Moore
Netherlands eScience Center
Netherlands
|
|
|
|
Brief Bio
Dr. Candace Makeda H. Moore, MD, earned her B.A, from Columbia University in the USA and her medical degree from Technion – Israel Institute of Technology in Israel. She has worked as a physician and also worked in computation for medical research. She has developed scientific software programs for various applications, including programs related to diverse aspects of radiology data, and programs for infectious disease modeling and prediction. She has a deep interest in the application of various computational techniques to both medical image data and electrophysiological data. Candace Makeda joined the Netherlands eScience Center as a Research Software Engineer in January 2022. Since 2022 she has led several research projects related to patient electrophysiological data including the ReSurfEMG project which examines respiratory electromyographic signals. As a physician she has a background that made her uniquely prepared for creating explainable, clinically relevant algorithms for research.
|
|
Walter Baccinelli
Netherlands eScience Center
Netherlands
|
|
|
|
Brief Bio
Dr. Walter Baccinelli holds a PhD in Biomedical Engineering from Politecnico di Milano, where he developed a new system for the administration of home-based physical rehabilitation for post-stroke patients, using the properties of radio frequency transmission. During the PhD, he also worked at Ab.Acus, where he developed new eHealth and telehealth solutions leveraging innovative technologies and research outputs. As a research software engineer at the Netherlands eScience Center he has worked on multiple projects including ReSurfEMG where has applied his expertise in signal processing and user-interface creation.
|
Abstract
Abstract
Respiratory electromyogram analysis has in the past relied upon closed analyses that were not reproducible. This tutorial introduces tools and software for open reproducible work in the pre-processing and analysis of respiratory EMG signals.
Keywords
EMG, respiratory, open science
Aims and Learning Objectives
Participants will be able to create synthetic EMG data and work with in code notebooks. Participants will learn to use version control, and contribute to a shared repository.
Target Audience
Researchers in respiratory physiology, pulmonology or cardiology who use EMG data
Prerequisite Knowledge of Audience
Basic python or proficiency in R.
Detailed Outline
1. Version control, theory and practicum
Participants will be taught about Git and Github
2. Jupyter notebooks, theory and practicum
Participants will use prepared Jupyter binders such as https://mybinder.org/v2/gh/ReSurfEMG/learning/main
3. Creating synthetic EMG data
Participants will use different prepared Jupyter binders for synthetic data creation
4. ReSurfEMG tour
Participants will get an overview of ReSurfEMG package, and tips on usage
5. Alternative packages
Participants will get an over view of alternative packages e.g. neurokit2