IBM Research-China Showcases 8 Papers at Medinfo 2017

Share this post:

Using cognitive computing to improve healthcare in China has been one of the main focus areas for the IBM Research-China organization. With 7 billion hospital visits per year, the country’s healthcare systems have the opportunity to tap into that data and uncover new insights for disease management and population health.  Collaborations with 15 leading Chinese hospitals in addition to government agencies, universities, pharmaceuticals, and medical device companies, helped establish a broad and trustful ecosystem for data-informed medical decision-making. We’ve detailed some of our work in applying cognitive analytics to healthcare in a blog post earlier this year.

The team continues to publish innovative research in this space. At the 2017 MedInfo Conference in Hangzhou, China IBM Research-China is presenting 8 papers covering a suite of technological advances designed to assist clinicians in the day-to-day management of chronic diseases from four perspectives: prevention, assistance in diagnosis, treatment, and patient engagement. Altogether, the team published 18 co-authored papers with local healthcare partners over the span of last year.



Major technical innovations presented at the conference are captured below:

WINNER: “Best Paper Award” – Second Place

Shijing Guo, Xiang Li, Xin Du, Haifeng Liu, Guotong Xie. Group-based Trajectory Analysis for Long-Term Use of Warfarin Therapy in Atrial Fibrillation Patients.      

Thursday Morning, August 24, MEDINFO 2017. Full paper, Oral presentation.

Patient use patterns for long-term medications may affect the clinical outcomes. Warfarin, which is a long-term oral medication, has been shown its effectiveness in reducing the risk of ischemic stroke and other thromboembolism (TE) for Atrial fibrillation (AF) patients. This paper discovers the trajectory patterns of warfarin use in AF patients, and uncovers how different trajectories are associated with different TE outcomes. Furthermore, the factors affecting the future warfarin use are identified. Predicting the warfarin use trajectory for new patients can help to efficiently target the specific patient groups and propose relevant interventions for warfarin use.

Enliang Xu, Xiang Li, Jing Mei, Shiwan Zhao, Gang Hu, Eryu Xia, Haifeng Liu, Guotong Xie, Meilin Xu, Xuejun Li. Applying Risk Models on Patients with Unknown Predictor Values: An Incremental Learning Approach.

Wednesday, August 23, MEDINFO 2017. Full paper, Oral presentation.

Risk prediction models are used in clinical decision making to help patients make an informed choice about their treatments. In clinical practice, developed risk models cannot be directly applied on patients with unknown or missing predictor values. In this work, we propose an incremental learning approach to impute a patient’s unknown predictor values based on his/her k-nearest neighbors (k-NN) from the incremental population. The experimental results show that k-nearest neighbors based incremental learning for data imputation can gradually increase the prediction performance when applying a developed risk model on new patients.

Eryu Xia, Haifeng Liu, Jing Li, Jing Mei, Enliang Xu, Xiang Li, Gang Hu, Guotong Xie, Xuejun Li, Meilin Xu. Gathering Real World Evidence with Cluster Analysis for Clinical Decision Support.

Thursday, August 24, MEDINFO 2017. Full paper, Oral presentation.

Clinical decision support systems have been shown to enhance clinical performances in many applications. In this paper, we propose the workflow of using cluster analysis in clinical decision support systems, providing data evidence for clinical decisions and compiles a wide range of clinical practices to inform the performance of individual clinicians. An example use of the system has been demonstrated under the scenario of blood lipid management in type 2 diabetes. This paper represents a step toward promoting patient-centered care and enabling precision medicine.

Guoyu Tang, Yuan Ni, Guotong Xie, Xinli Fan, Yanling Shi. A Deep Learning based Method for Similar Patient Question Retrieval in Chinese. MEDINFO 2017.

 Wednesday, August 23, MEDINFO 2017. Full paper, Oral presentation.

The online patient question and answering (Q&A) system attracts an increasing amount of users in China. Patients will post their questions and the registered doctors would provide the corresponding answers. Instead of waiting for the response from a doctor, the newly posted question could be quickly answered by finding a semantically equivalent question from the Q&A achieve. In this study, we investigate a novel deep learning based method, i.e. the supervised neural attention model (SNA), to retrieve the similar patient question in Chinese. The experimental results show that our SNA method achieves better performance than all other compared methods.

Jing Mei, Shiwan Zhao, Feng Jin, Lingxiao Zhang, Haifeng Liu, Xiang Li, Guotong Xie, Xuejun Li, Meilin Xu. Deep Diabetologist: Learning to Prescribe Hypoglycemic Medications with Recurrent Neural Networks.

Wednesday, August 23, MEDINFO 2017. Full paper, Poster.

In healthcare, applying deep learning models to electronic health records (EHRs) has drawn considerable attention. EHR data consist of a sequence of medical visits, i.e. a multivariate time series of diagnosis, medications, physical examinations, lab tests, etc. This sequential nature makes EHR well matched to the power of Recurrent Neural Network (RNN). In this paper, we propose “Deep Diabetologist” – using RNNs for EHR sequential data modeling, to provide the personalized hypoglycemic medication prediction for diabetic patients. Our experimental results demonstrate the improved performance, compared with a baseline classifier using logistic regression.


Yiqin Yu, Xiang Li, Gang Hu, Haifeng Liu, Jing Mei, Yuan Ni, Guoyu Tang, Guotong Xie, Weiming Xu. Personal Health Self-Management in a Data Perspective

Wednesday August 23, 2017 Full paper, Poster.

Along with the growth in numbers of patients with chronic diseases, personal health self-management is critical. In this poster we introduce a mechanism for personalized health self-management for chronic diseases based on data. Specifically, we address the self-management problem by introducing three modules into our tool Personal Health Advisor (PHA): personal health risk assessment, similar patients profiling, and health question answering. We built a personal health data flow mechanism to allow the automatic data extraction from printed medical records, and flowing the data among different modules.

Xiang Li, Haifeng Liu, Jingang Yang, Guotong Xie, Meilin Xu, Yuejin Yang. Using Machine Learning Models to Predict In-hospital Mortality for ST Elevation Myocardial Infarction Patients. MEDINFO 2017.

Wednesday, August 23, MEDINFO 2017. Full paper, Oral presentation.

ST-elevation myocardial infarction (STEMI) is a major cause of hospitalization and mortality in China. Accurate prediction of in-hospital mortality is critical for clinical decision making to STEMI patients. We used machine learning approaches to build in-hospital mortality prediction models for STEMI patients from Chinese Acute Myocardial Infarction (CAMI) registry data. We first performed feature engineering on CAMI data to identify potential predictors. Then supervised learning methods were applied to build the prediction models. The experimental results show that our models achieve higher prediction performance than the previous in-hospital mortality prediction models for STEMI.

Haifeng Liu, Xiang Li, Guotong Xie, Xin Du, Ping Zhang, Chengming Gu, Jingyi Hu. Precision Cohort Finding with Outcome-Driven Similarity Analytics: A Case Study of Patients with Atrial Fibrillation.  MEDINFO 2017.

Wednesday, August 23, MEDINFO 2017. Full paper, Oral presentation.

Dividing patients into similar groups plays a significant role in disease phenotyping, risk stratification, and personalized medicine. We propose an outcome-driven approach to identify clinically similar patients which are grouped together as a precision cohort. The approach quantitatively measures the similarity between patients in terms of a particular clinical outcome of interest. We demonstrate the effectiveness of the approach in a real-world case study: from an atrial fibrillation patient cohort which is usually considered to be with high risk of ischemic stroke (IS), our approach successfully identified a precision cohort of patients with truly low risk of IS.

More Healthcare stories

Advancing the Potential of AI in Medical Imaging at MICCAI 2020

I believe one of the most promising areas for AI to make an impact is in the field of medical imaging. Through advancements in AI that allow for more intelligent and accurate analysis of video and still images, there is hope that clinicians will soon be able to widely augment the data and information they […]

Continue reading

AI Data Tracker Encourages Scientific Research into COVID-19 Non-Pharmaceutical Interventions

What impact do measures such as shelter-in-place, mask wearing, and social distancing have on the number of COVID-19 cases? How do the COVID-19 quarantine measures that have been implemented by North American countries compare to South American countries? These are just a few questions about the wide range of non-pharmaceutical interventions (NPIs) that have been applied by governments, globally.

Continue reading

IBM Research and the Broad Institute Seek to Unravel the True Risks of Genetic Diseases

In 2019, IBM and the Broad Institute of MIT and Harvard started a multi-year collaborative research program to develop powerful predictive models that can potentially enable clinicians to identify patients at serious risk for cardiovascular disease (1, 2). At the start of our collaboration, we proposed an approach to develop AI-based models that combine and […]

Continue reading