Fenglong Ma1, Jing Gao1, Qiuling Suo1

Slides:



Advertisements
Similar presentations
Allison Dunning, M.S. Research Biostatistician
Advertisements

David P. Taylor, MS 1,2, Nathan C. Hulse, PhD 1,2, Grant M. Wood 2, Peter J. Haug, MD 1,2, Marc S. Williams, MD 1,2 1 University of Utah, Salt Lake City,
Providing Control, Autonomy and Profitability for the Healthcare Provider PHYSICIANS PROVISO.
PREDICTIVE MODELING IN E-HEALTH USING ARTIFICIAL INTELLIGENCE MARK HOOGENDOORN (AND MANY OTHERS INCLUDING MICHEL KLEIN) MARK HOOGENDOORN (AND MANY OTHERS.
Supporting clinical professionals in the decision-making for patients with chronic diseases Mitja Luštrek 1, Božidara Cvetković 1, Maurizio Bordone 2,
University of Minho School of Engineering Algoritmi Centre Uma Escola a Reinventar o Futuro – Semana da Escola de Engenharia - 24 a 27 de Outubro de 2011.
Chen Cheng1, Haiqin Yang1, Irwin King1,2 and Michael R. Lyu1
A Classification Approach for Effective Noninvasive Diagnosis of Coronary Artery Disease Advisor: 黃三益 教授 Student: 李建祥 D 楊宗憲 D 張珀銀 D
Data Mining: A Closer Look
APPLICATION : DIAGNOSTIC CODING 1 SIEMENS  Coding is the translation of diagnosis terms describing patients diagnosis or treatment into a coded number.
Stroke Quality Measures Kathy Wonderly RN, BSPA, CPHQ Performance Improvement Coordinator Developed: May, 2012 Most recently updated: October,
CS 478 – Introduction1 Introduction to Machine Learning CS 478 Professor Tony Martinez.
Exercise Management Cancer. Pathophysiology Cancer is not a single disease; it is a collection of hundreds of diseases that share the common feature of.
Midterm Review Rao Vemuri 16 Oct Posing a Machine Learning Problem Experience Table – Each row is an instance – Each column is an attribute/feature.
A Framework for Mining Signatures from Event Sequences and Its Applications in Healthcare Data.
Computers in Healthcare Jinbo Bi Department of Computer Science and Engineering Connecticut Institute for Clinical and Translational Research University.
The potential to bring huge benefits to Patients..
HRVFrame: Java-Based Framework for Feature Extraction from Cardiac Rhythm Alan Jovic and Nikola Bogunovic Faculty of Electrical Engineering and Computing,
1 Impact of Implementing Designed Nursing Intervention Protocol on Clinical Outcome of Patient with Peptic Ulcer By Amal Mohamed Ahmad Assistant Professor,
Skeleton Based Action Recognition with Convolutional Neural Network
يادگيري ماشين Machine Learning Lecturer: A. Rabiee
“The degree to which individuals have the capacity to obtain, process, understand basic health information and services needed to make appropriate health.
Cancer 101: A Cancer Education and Training Program for [Target Population] Date Location Presented by: Presenter 1 Presenter 2.
Combining Evolutionary Information Extracted From Frequency Profiles With Sequence-based Kernels For Protein Remote Homology Detection Name: ZhuFangzhi.
Experience Report: System Log Analysis for Anomaly Detection
When deep learning meets object detection: Introduction to two technologies: SSD and YOLO Wenchi Ma.
Reduction Of Readmissions To Hospitals Based on Actionable Knowledge Discovery and Personalization Zbigniew W.  Ras Sponsored by.
How to keep active with cancer?
He Xiangnan Research Fellow National University of Singapore
Hybrid Deep Learning for Reflectance Confocal Microscopy Skin Images
Showcasing work by Jonnageddala, Liaw, Ray, Kumar, Chang, and Dai on
Diagnosing Diabetes and Predicting Complications
Helping you prevent heart disease, stroke, diabetes and kidney disease
Golden Rules to Keep Kidneys Healthy
At the end of this talk, the resident will be able to:
KDD CUP 2001 Task 1: Thrombin Jie Cheng (
An Artificial Intelligence Approach to Precision Oncology
School of Computer Science & Engineering
Family Health History Health project.
How does teamwork improve value. Dr Nils E
International Workshop
MR images analysis of glioma
Regularizing Face Verification Nets To Discrete-Valued Pain Regression
Walden University Carrie Vanzant February 7, 2010
Prevention Cardiovascular disease
Population Information Integration, Analysis and Modeling
Cheng-Ming Huang, Wen-Hung Liao Department of Computer Science
Learning with information of features
Time to CARE: A collaborative engine for practical disease prediction
An Initial Study of Survival Analysis using Deep Learning
Objectives (IPS chapter 1.1)
Predicting Pneumonia & MRSA in Hospital Patients
iSRD Spam Review Detection with Imbalanced Data Distributions
MEgo2Vec: Embedding Matched Ego Networks for User Alignment Across Social Networks Jing Zhang+, Bo Chen+, Xianming Wang+, Fengmei Jin+, Hong Chen+, Cuiping.
Atherosclerosis Insights
External Validation of Existing Stroke Risk Models
Selecting the Right Predictors
Binghui Wang, Le Zhang, Neil Zhenqiang Gong
WI-BEEP (WIreless technology and Behavioral Economics to Engage Patients with type 2 diabetes or hypertension) Angellotti E1, Pierce A2, Hescott B3, Wong.
RCNN, Fast-RCNN, Faster-RCNN
View Inter-Prediction GAN: Unsupervised Representation Learning for 3D Shapes by Learning Global Shape Memories to Support Local View Predictions 1,2 1.
Scalable and accurate deep learning with electronic health
Graph Neural Networks Amog Kamsetty January 30, 2019.
Model Enhanced Classification of Serious Adverse Events
Simplifying Healthcare
Automatic Handwriting Generation
NON-NEGATIVE COMPONENT PARTS OF SOUND FOR CLASSIFICATION Yong-Choon Cho, Seungjin Choi, Sung-Yang Bang Wen-Yi Chu Department of Computer Science &
Bug Localization with Combination of Deep Learning and Information Retrieval A. N. Lam et al. International Conference on Program Comprehension 2017.
Family Health History Health project.
Jianbo Chen*, Le Song†✦, Martin J. Wainwright*◇ , Michael I. Jordan*
Presentation transcript:

Risk Prediction on Electronic Health Records with Prior Medical Knowledge Fenglong Ma1, Jing Gao1, Qiuling Suo1 Quanzeng You2, Jing Zhou3, Aidong Zhang1 1 SUNY at Buffalo, 2 Microsoft AI & Research, 3 eHealth Inc. KDD 2018

Electronic Health Records Background Electronic Health Records Personalized Medicine “An electronic health record (EHR), or electronic medical record (EMR), is the systematized collection of patient and population electronically-stored health information in a digital format.” --- Wikipedia

Electronic Health Records (EHR) Background Electronic Health Records (EHR) A comprehensive EHR dataset that contains everything happened to a patient at the hospital. Structured Codes Spectrograms Lab Measures Images Free Text

Challenges of Mining EHR Data EHR Data with Structured Codes Temporal High dimensional Noisy An example of a patient’s visit information.

Disease Risk Prediction Task Disease Risk Prediction Utilizing historical EHR data of individuals to predict whether the patient will suffer a certain disease in the future.  An example for heart failure risk prediction.

Ignore the importance of prior medical knowledge! Existing Work Deep Learning based Risk Prediction Convolutional Neural Networks (CNN) Recurrent Neural Networks (RNN)  Cheng et al. Risk Prediction with Electronic Health Records: A Deep Learning Approach. In SDM’16. Drawback Ignore the importance of prior medical knowledge! Choi et al. RETAIN: An Interpretable Predictive model for Healthcare Using Reverse Time Attention Mechanism. In NIPS’16.

Doctor Diagnosis Process Motivation Doctor Diagnosis Process Heart Failure? Medical Knowledge

Challenge of Using Medical Knowledge Almost all the medical knowledge is represented by arbitrary rules. Tobacco use. Using tobacco can increase your risk of heart failure. (Categorical) Rule: Tobacco use Heart failure High blood pressure. Your heart works harder than it has to if your blood pressure is high. (Continuous) Rule: High blood pressure Heart failure

Risk Prediction with Prior Medical Knowledge Posterior Regularization An effective technique to convert the discrete knowledge into continuous real-valued features by modeling the posterior distribution as a constrained posterior feature set. Ganchev, et al., Posterior Regularization for Structured Latent Variable Models. JMLR, 2010.

Risk Prediction with Prior Medical Knowledge Posterior Regularization Ground Truth Our Final Goal Rules A Function Given Value Drawback: Hard to manually set reasonable bounds for constraint features.

Risk Prediction with Prior Medical Knowledge Solution Represent the desired distribution as a log-linear model. Desired distribution The Proposed Model PRIME. Any Existing Model

Risk Prediction with Prior Medical Knowledge Constraint Feature Design Patient Characteristics Ethnicity Age Underlying Diseases Disease Duration Genetics Family History ℇ denotes the set of races related to the prediction. 𝐮 is the frequency vector of underlying diseases. 𝐝 is the duration vector of underlying diseases. 𝒞 is the set of all the diagnosis codes in 𝐗. 𝒢 denotes the set of genetic disorders. ℋ represents the set of family history disorders.

Example of Designing Constraint Features Underlying Diseases and Durations Underlying Disease 401.9 278.0 305.02 Frequency 2 1 Duration (month) 23 17 9 𝐮 𝐝

Risk Prediction with Prior Medical Knowledge An easy way to understand PRIME Prediction Training Deep Learning Feature Engineering Risk Model Prediction Prior Medical Knowledge Prediction

Patient Characteristic Experiments Datasets Designing Constraint Features Feature Patient Characteristic Underlying Diseases Disease Duration Genetics Family History Ethnicity Age Heart Failure   √ √   √ COPD Kidney Disease

Performance Evaluation Measures F1 Score, Accuracy and AUROC The higher the better Results on three datasets

Constraint Feature Analysis The advantage of the proposed PRIME is to automatically learn the weights for different risk factors and constraint feature categories. Confidence of Feature Categories Confidence Matrix Learned by PRIME on the Heart Failure Dataset.

Constraint Feature Analysis Weights of Constraint Features Case Group Heart Failure ID Underlying Diseases 1 High blood pressure 2 Coronary artery disease 3 Diabetes 4 Congenital heart defects 5 Valvular heart disease 6 Alcohol use 7 Smoking 8 Obesity Control Group

Our new work is coming soon… Discussions The proposed framework PRIME is only effective for common diseases. Our new work “Fake is the New Real: Predicting Rare Diseases with Deep Generative Networks and Reinforcement Learning” is coming soon… 350 million people globally are fighting rare diseases. Only 5% of rare diseases have FDA approved therapies. Rare diseases affect more people than HIV and Cancer combined. https://blog.cirm.ca.gov/2016/03/17/rare-disease-underdogs-come-out-on-top-at-cirm-board-meeting/

Conclusions This work is the first attempt to take prior medical knowledge into account for risk prediction task. We propose a novel framework PRIME, which models prior medical knowledge as posterior regularization and learns the desired posterior distribution with a log-linear model. The proposed PRIME is a general model, which can be easily applied to any predictive models in healthcare. PRIME is able to distinguish the importance of different prior knowledge contributed to the risk prediction.

Thank You! Questions? Source code, slides and poster are publicly available at http://www.acsu.buffalo.edu/~fenglong.

Backup Directly use constraint features to predict the labels of patients? 86.3%