Transforming healthcare through data
In recent years, analyzing and synthesizing the unprecedented volume of digital health data has presented many challenges to computer science professionals, medical researchers, and clinicians. The successful handling of this data can help improve personal health management, clinical care, medical research, and public health. eHealth theme programs encourage collaborations between different institutions. In particular, we encourage interdisciplinary graduate students to train with computer scientists and health experts in order to address the challenges in health data mining.
Feature Stories
Asia Workshop Focus: Healthy Humans
Microsoft Research Asia focused on addressing computing trends that are expected to transform the delivery of healthcare services during an eHealth Workshop in Beijing, China, in February 2010. This event provided a forum for faculty and researchers from medicine, medical engineering, biology, computer science, and electrical engineering-related disciplines to exchange ideas on research and activities.
New Technology Gives More Mobility to Seniors and the Disabled
To increase mobility for people with disabilities, newly developed computing technologies have been applied to build an intelligent wheelchair with a brain, eyes, and ears for home and hospital environments.
Collaborative Research and Projects
Microsoft Research collaborates with researchers in advanced computing techniques such as data mining, mobile sensors, natural languages processing, signal processing, visualization, and medical imaging. Below is a list of the current eHealth-related projects that Microsoft Research funds in the Asia-Pacific region.
A Personalized eHealth Agent for Monitoring Diabetes Patients in a Mobile Computing Environment
Investigator: Byoung-Tak Zhang, School of Computer Science and Engineering, Interdisciplinary Programs in Cognitive Science, Brain Science, and Bioinformatics, Seoul National University
Abstract: This project aims to develop an intelligent eHealth agent that monitors the health status of diabetes patients. The agent collects patients' behavioral data by using the sensors on a mobile phone, learns the life patterns of the patient, and suggests personalized health plans—such as exercises and meals—on the go.
Analyzing Patient Videos in Clinical Settings
Investigator: Alex Yong-Sang Chia and How-Lung Eng, Institute for Infocomm Research, Singapore
Abstract: The focus in this project is to design a set of tools to help doctors analyze the activities and behaviors of patients. In particular, we would like to explore the use of patient video data that is collected from specific scenarios such as that from the Epilepsy Monitoring Unit. Detection of epilepsy and seizures is a fundamentally difficult research problem.
Build a Question and Answer Knowledge Base for Health Care Workers and Patients
Investigator: Gao Cong, School of Computer Engineering, Nanyang Technological University
Abstract: Our idea is to build an automatic Question and Answering (QA) service on priority health care that estracts, distills, and disseminates the existing health care knowledge on the web to help people—anytime and anywhere.
Clinical Report Search Engine
Investigator: Eiji Aranaki, The Center for Knowledge Structuring, University of Tokyo
Abstract: With the rapidly growing use of electronic health records, the possibility of large-scale clinical information extraction has drawn much attention. For example, more than 4,000 patients come to our hospital (The University of Tokyo Hospital) daily. Our project aims to develop a clinical report search engine that enables the extraction of clinical records with the specified patient status.
Creation of Research Data Warehouse for Computer Assisted Diagnosis of Colon Cancer Based on Whole Slide Histopathological Images
Investigator: Maode Lai, M.D., professor of Pathology, School of Medicine, Zhejiang University
Abstract: Cancer is one of the principal causes of death. The detection methods for cancer—such as PET, CT, MRI, and histopathology—are widely used. Only histopathology of a rich source of information in high resolution can reliably and correctly detect cancer. This project aims at creating an open research data warehouse of thousands of whole slide colon cancer biopsy images with sharing opportunities for the public as well as developing an automated analysis tool with CAD imagery of colon cancer.
Entity Linking For Health Text Information Aggregation
Investigator: Su Jian, senior scientist, group leader, Information Extraction and Text Mining (IETM), Institute for Infocomm Research
Abstract: This project investigates the knowledge integration of health text from different sources—such as medical and health related articles and electronic health records—to foster related research through developing entity-linking technology. Specifically, the project will look into the incorporation of topic modeling and attribute extraction, enhancement on acronym handling, and also machine-learning issues to advance entity-linking technology.
Identifying the Presence and Certainty of Clinical Conditions from Clinical Reports
Investigator: Jong C. Park, Computer Science Department, Korea Advanced Institute of Science and Technology (KAIST)
Abstract: The specific details of clinical conditions and their proper understanding play a critical function in clinical decision making. Such clinical conditions are usually spelled out by medical experts in terse but plain English and stored as clinical reports. In this project, we propose to develop methods to classify mentions of clinical conditions into meaningful categories and develop a method that helps medical personnel decide which conditions should be manually reviewed to determine their presence and certainty.
Information Extraction in Chinese Medical Records: Creation of Research Data Warehouse and Organizing Challenges with Microsoft Research Asia
Investigator: Yubo Fan, professor and dean, School of Biological Science and Medical Engineering, Beihang University
Abstract: This project aims at building the first open research data warehouse for information extraction of Chinese medical records and publishing of the annotated Chinese medical records for research use. We also plan to collaborate with Microsoft Research Asia to hold challenges by using the annotated Chinese medical record and implementation of the evaluation framework to compare various algorithms from the challenges.
Knowledge Discovery Based on Medical Data Mining Service in Cloud
Investigator: Feipei Lai, Department of Electrical Engineering and Department of Computer Science & Information Engineering, National Taiwan University
Abstract: In the fields of bioinformatics, a huge number of papers report the amazing results obtained by applying machine-learning methods to medical data. But we have observed that it is not common to apply these outcomes to clinical practice. The major reason is, the model produced by the training methods, and the training methods themselves, are not as convincing as the statistical results.
Ontology-based Text Mining for Unstructured eHealth Data
Investigator: Sungyoung Lee, director, East-West Neo Medicinal u-Lifecare IT Research Center, Kyung Hee University
Abstract: The limitation of the existing key phrase algorithms is that output key phrases may contain irrelevant information along with relevant information. In the health domain, the security, privacy, and accuracy of medical data is vitally important. Therefore, this project aims to improve the key phrase extraction procedure by exploiting different hierarchical levels of ontology so that we can achieve the required level of accuracy for medical domain.
Personalized Health Risk Analysis and Prediction Using Data Mining Techniques
Investigator: Vincent S. Tseng, Department of Computer Science and Information Engineering/Institute of Medical Informatics, National Cheng Kung University
Abstract: In this project, we aim to construct a novel health risk mining mechanism with the prototype system for predicting personal health risk on chronic diseases like diabetes based on the patient's health examination data, personal profile, and the life habits. A very novel part of this work is that the subject’s personal profile, health examination results, and the life habits are all taken into consideration simultaneously. Moreover, novel data mining mechanisms will be developed for discovering the health risk patterns that are based on the series of historic health examination records instead of the latest examination results only, as used in currently existing approaches.
Daily Activity Exercise Stress Evaluation Based on Plantar Pressure and Foot Motion Monitoring
Investigator: Deyu Li, Bioengineering Department, Beihang University
Abstract: As appropriate exercise stress plays a significant role in maintaining human health, this project is working on a method and data acquisition device for evaluating the exercise stress of daily activities and health status, based on continuous monitoring of the plantar pressure distribution and foot movement. Compared with existing devices to monitor patients for disease diagnosis and therapy, this project, with low cost, is designed not only for patients, the disabled, and elders to select appropriate rehabilitation measures and exercise modes to avoid serious injury, but also for the general population to promote health by helping them choose the level of exercise stress that corresponds to their health status. In this era when people focus on maintaining good health, the proposed project will partially meet the needs of different populations.
Context-Aware and Cloud Computing of Human Daily Activities Through Analyzing Tri-axial Acceleration Signals for Health Monitoring of College Students
Investigator: Lianwen Jin, School of Electronic and Information Engineering, South China University of Technology
Abstract: This project is to study the problem of context-awareness of college students' daily activities and to build a health monitoring system that is based on a cloud computing platform. The system is composed of mobile devices, tri-axial accelerometers, and the cloud computing platform. Mobile devices collect data, which is generated by an accelerometer that is placed in the user's pocket, and which transmits the data to the cloud platform. All data will be processed in the cloud platform, and results will be fed back to users through a 3G wireless network. Through signal-processing and machine-learning algorithms, the collected tri-axial acceleration signals will be classified into four to nine daily basic activities, such as walking, running, jumping, and climbing stairs. The project focuses on collecting information of college students' daily activities over a long period, and analyzing the relationship between people’s health status and their daily activities. This system is believed to be beneficial for many new potential applications, such as innovative ubiquitous mobile services and cloud health care.
Global Cancer Map: Large-scale Meta-analysis of Cancer Microarray Data
Investigator: Jaewoo Kang, Department of Computer Science and Engineering, Korea University
Abstract: This project is to exploit advances in high-performance heterogeneous computing to lay the foundations for faster and more reliable medical image analysis including registration, modeling/simulation, and segmentation. The results of the project will transform the role of technology in future health-care systems by providing low-cost ubiquitous automated/semi-automated diagnosis, planning, and visualization solutions, and also by introducing real-time and enhanced inter-operative tools for non-invasive image-guided surgery. The goal of the project is to contribute to the general understanding of the trade-offs and challenges in transitioning from traditional serial computation into the new era of parallel platforms.
