Approach Comparative Effectiveness Research (CER) hypothesis prediction in personal health • Unstructured Message filtering, extraction messages • Professional labeling by medical students • Sentiment Analysis (attitude: pos/neg/neu) Yunliang Jiang (Computer Science, UIUC) Advisor: Bruce Schatz (Medical Information Science, UIUC) • Direct comparison by same patient (e.g. I With: Vera Qingzi Liao (UIUC) and Qiaozhu Mei (Umich) prefer T1 to T2) • Indirect comparison by same patient (e.g. Sponsor: USDA-NRI, UIUC-IGB, State Farm Scholarship positive to T1, negative to T2) • Indirect comparison in overall case • Demographic Analysis (sex, age, region) • Statistical test Problem Conclusion and Plans • How to compare the effectiveness of different • Our approach can predict patients’ drugs/treatments by personal health messages? preference consistently (e.g. people prefer • Patient’s opinion on treatments (attitude, preference) meditation to SSRI) can interpret the effectiveness • Different groups may have specific • Data source: MedHelp (100+ forums, 1M+ messages) preferences (e.g. Age 65+ people prefer • Representative Dataset: radiation than chemo; South people prefer • Breast Cancer (70K messages, 18K patients, Hormonal therapy more than Northeast Chemo v.s. Radiation v.s. Hormonal) people.) • Depression (186K messages, 38K patients, • Predicted hypothesis => need clinic proof Meditation v.s. Drug treatment)
Approach Text Mining for Health Informatics • Parikshit Sondhi (UIUC) Construct symptom relationship graphs • Similarity via Random Walk Restart Advisor: ChengXiang Zhai (UIUC) with Jimeng Sun (IBM), Hanghang Tong (IBM) Heart Heart Wheezing Wheezing Attack Attack w 2 Fever Fever Cough Cough Ankle Ankle w 1 Edema Edema w 5 Rales Rales w 3 Chest Chest w 4 Pain Pain Problem Results and Conclusion • Mining symptom relationships from unstructured patient records • Dataset: • CHF Cases: 500K (4.6K Patients) • Controls: 500K (8.4K Patients) • Find symptom mentions related to Successfully identified related diseases and known CHF symptoms symptoms, confirmed by clinical experts
Detecting Anomalous Accesses in EHR Audit Logs Approach Siddharth Gupta (UIUC) • Experience Based Access Management (EBAM) • Cluster the users and patients in a high Advisor: Carl A. Gunter (UIUC) dimensional space based on features such as with Mario Frank (UC Berkeley) and SHARPS (diagnosis, procedures, medications) • Type all users based on interaction with the patient clusters. Dataset: 7 million access logs, 8000 Users and 25000 Patients Progress and Plans Problem • • Insider threats Completed modeling (LDA) to form clusters of - curiosity and personal information users and patients. • - financial fraud Selected 15-25 Topics for each feature, using - medical identity theft. perplexity measure. • • Complex workflows and difficult to handle Analyzed probability of a user belonging to a access in emergency situations. patient topic and typing them. • • Big Data with limited semantic knowledge Plan to evaluate the model using “random about user-patient interactions. user model”.
Approach Feature extraction Processing of raw data generated by -Sampling frequency: 60Hz phone sensors to extract walking motion -time domain features: mean, standard deviation and mean crossing rate and build gait detection models -frequency domain: peak location, entropy, sub-band energy and sub-band energy ratio Qian Cheng (Computer Science, UIUC) Training data Advisor: Bruce Schatz -5min walk for each type of motions; carry With: Joshua Juen (Electrical and Computer phone in different place; user feedback to Engineering, UIUC ), Yanen Li (Computer Science, label each set UIUC), Yunliang Jiang ((Computer Science, UIUC) Server operation -Receive and storage all useful data Sponsor: USDA-NRI, UIUC-IGB -Run models and return detecting decisions Progress and Plans Problem • Time-series data process Raw data to gait features • -Combine the acceleration and orientation Personalized model for detecting abnormal information. Interpret acceleration in body health based vertical-horizontal coordinates • Integrated system with phone clients and -Iteratively split data to same pieces(2000pts) servers Frequency domain data process -Use FFT to interpret time-series data to frequency domain. -Filters to remove noise generated by phone Model building -Decision tree and SVM for classification -Semi-supervised strategy to improve models
Title: Selective Consumption of Online Approach Health Information Empirical studies Case studies on to analyze user real health info Q. Vera Liao (HCI Group, CS Dept.) behavior patterns websites Advisor: Prof. Wai-Tat Fu Design “informed” system and interface Problem: Progress and Plans • Internet provides overwhelmingly large amount, • Experiment 1: credibility judgment diverse, and sometimes contradicting information. to • Experiment 2: selective exposure bias health information seekers • In progress: analyze data on Medhelp.com to see o Varying quality: malicious source, unreliable how users interact with different opinions. information • Plan to conduct more user studies to understand o Varying opinions: Information bias how to nudge users to make “better” selection • I am interested in understanding and designing • Plan to develop system or interface that supports system/interface in support of health information users to navigate through the diverse opinion space seekers’ of online health information when facing complex o Selection of high quality informatio n health related decisions o Deliberation on diverse opinions
Building 3D Tele-immersive (3DTI) Approach Applications for Remote Physical -- Real-time 3D data acquisition and model Therapy reconstruction using Kinect camera -- 3D semantic-based compression and transmission framework -- Synchronization framework for multi-modal 3DTI streams Henry Pengye Xia -- Intuitive UI solution using mobile phones and Advisor: Klara Nahrstedt Kinect camera TEEVE@UIUC Problem Progress and Plans -- Remote physical therapy requirements -- Current: -- Definition of 3DTI system 1. A working 3DTI testbed 1. High Bandwidth usage 2. 3D semantic-based middleware implementation 2. Low delay for interactivity 3. TEEVE-Remote UI -- Deficiency of current technology 1. 3D video acquisition, compression, transmission -- Future: 2. Multi-stream synchronization 1. Explore the multi-modality from the perspective 3. Intuitive user-interaction (UI) solution of synchronization
Safe Reconfiguration of Acute Care Approach Monitoring Systems • An abstract model of dynamic clinical environment • A highly modular system architecture • Encapsulation of MDPnP complexity Maryam Rahmaniheris (UIUC) • Consistent propagation of necessary updates throughout the monitoring system Advisor: Lui Sha(UIUC) Collaborator: Richard Berlin (UIUC) Progress and Plans Problem • An abstract model of monitoring system is proposed • Medical Device ”Plug -and- Play” (MDPnP) can • An MDPnP architecture with proven safety and mitigate the preventable medical errors consistency is developed • Dynamic clinical environment poses new design • Performing more thorough analysis of the architecture challenges • Collecting more use cases to evaluate the efficiency of our proposed model and architecture to enforce • The following questions must be answered safety and consistency in different acute care • Is the current configuration safe and consistent? scenarios • Did the system reconfiguration cause changes in the measurements interpretation?
Biosensor for Point of Care HIV Diagnostics for Global Health Approach Umer Hassan (UIUC), Advisor: Rashid Bashir (UIUC) Problem The Differential counting of CD4 T lymphocytes using the whole blood The biggest challenge the country faces today is diagnosing all of its HIV-infected people and helping them take full advantage of the existing treatments (Science, p.167, Vol. 337, 13 July 2012). Progress & Results Device & Future Directions 1200 y = 0.9814x 1100 Chip CD4 T Counts /uL R² = 0.8772 1000 900 800 700 600 500 400 Pillars in Capture Chamber CD4 Counter Device with Flat 400 500 600 700 800 900 1000 1100 1200 Chambers Proposed Plans Carle Control CD4 T Counts/uL The CD4 Counts from our device shows a good correlation with the control counts • Start testing the current device on the patient samples obtained from the Carle Hospital • CD4/CD8 Counting on the same device • Integrating flow metering in the device
Recommend
More recommend
