Andong Zhan 展安东

Short Bio

I’m a Ph.D. candidate in Computer Science whose mission is to build novel computer systems using machine learning algorithms to solve challenging healthcare problems. I began my research at Johns Hopkins University and I have designed and deployed several computing systems for healthcare which directly influence thousands of inpatients and outpatients as well. Thus far, I have published more than 20 peer-reviewed articles in the areas of computer science and clinical research, including IEEE Communications, ACM SenSys, JAMA Neurology, Parkinsonism & Related Disorders, and Critical Care Medicine. My work has been cited in excess of 200 times by researchers from all over the world. Meanwhile, our machine learning based clinical decision system keeps monitoring inpatients in Johns Hopkins Hospitals and generating real-time alerts for those who may develop sepsis. I’m going to graduate in May 2018 and am looking for my next adventure.

My advisers are Prof. Andreas Terzis and Prof. Suchi Saria. Also, Prof. Yanif Ahmad gave me huge guidance on my research.

I received my M.E from the Dept. of Computer Science at Nanjing University, Nanjing, China, in June 2010. I was also a member of GPS Group (Group of Grid, Peer-to-peer and wireless Sensor networks) advised by Prof. Guihai Chen.

I received my B.S from the Dept. of Computer Science at Nanjing University, Nanjing, China, in June 2007.



Current projects:

System Design and Deployment of Diagnostic Decision Support System using EHR and Machine Learning

Over the last decade, electronic health records (EHRs) have been broadly adopted by most of the American hospitals. In the next decade, incorporating EHRs with CDSS (clinical decision support system) together into the process of medicine has the potential to change the way medicine has been practiced and advance the quality of patient care. The complexity of some diseases, especially in clinical care, provides a unique opportunity for machine learning to provide new insights and has stimulated research into novel methods for this purpose. However, applying ML-based CDS has to face steep system and data level challenges. In this project, CDS-Stack, an open cloud-based platform for machine learning clinical decision support, is introduced to help researchers and developers to deploy ML-based CDS into healthcare practice. CDS-Stack integrates various components into an infrastructure for ML development and CDS delivery. Specifically, an ETL engine is proposed to transform heterogenous EHRs, either historical or online, into a common data model (CDM) in parallel so that ML algorithms can start to run training or prediction directly. A push-based online CDS pipeline is introduced and evaluated for the first time in CDSS to deliver CDS in real-time. Finally, CDS-Stack has been deployed in JHMI to deliver TREWS sepsis alerts since Nov 2017 and start to show promising outcomes.

HopkinsPD and mPDS: Smartphone-based Parkinson Disease Severity Score

Using a Parkinson disease (PD)-specific application (HopkinsPD) to: (1) assess the feasibility of remote, online recruitment and completion of app installations, (2) objectively measure and quantify five factors of PD (voice, balance, dexterity, gait, and reaction time), (3) measure daily variability of these and other factors including mobility and socialization, and (4) correlate Android app sensor data and clinical assessments from the Unified Parkinson Disease Rating Scale (UPDRS) in a subset of participants. mPDS – the mobile Parkinson Disease Score – uses a novel machine learning technology, called Disease Severity Score Learning, to generate severity score from clinical comparison. The mPDS achieved high correlation with conventional clinical Parkinson severity assessment and meanwhile, it can be performed by the patients anytime anywhere. This work has been accepted by JAMA Neurology in 2018 and reported by Nature magazine 2016.

Previous projects:

MobiBed: An Open Network Protocol Testbed For Mobile Phones In The Wild

Fig.1 MobiBed utilizes the UDP packet interface to emulate a raw socket. By moving the network stack to user space, protocol evaluation and development can be performed without root privileges.

Smartphones are becoming the dominant devices people use to access the Internet. Following this trend, a deep investigation of network protocol performance on cellular networks is becoming increasingly important. However, most of the current protocol testing on smartphones are based on laboratory scope testbeds which require root privileges or even customized firmwares. These constraints limit such tests within a laboratory scope. Unfortunately, small scale testing cannot provide results with high confidence due to the high spatial and temporal variability of cellular networks.

To address this challenge of testing network protocols at scale, we design and implement MobiBed, an open network protocol testbed for mobile phones in the wild. Each phone participating in MobiBed runs a network protocol emulator in user space that implements sandboxed network stacks. Protocol packets are encapsulated in UDP tunnels that require no kernel changes or superuser privileges (Fig. 1 Left). MobiBed's draft research paper (under submission), implementation (server on Google App Engine and Android client), and supplementary information are available here. This is a draft web site and we are working to improve it. We welcome contributions to the code and your feedback.

Accurate Calorie Expenditure of Bicyclists using Cellphones

Biking is one of the most efficient and environmentally friendly ways to control weight and commute. To precisely estimate caloric expenditure, bikers have to install a bike computer or use a smartphone connected to additional sensors such as heart rate monitors worn on their chest, or cadence sensors mounted on their bikes. However, these peripherals are still expensive and inconvenient for daily use. This work poses the following question: is it possible to use just a smartphone to reliably estimate cycling activity? We answer this question positively through a pocket sensing approach that can reliably measure cadence using the phone’s on-board accelerometer with less than 2% error. Our method estimates caloric expenditure through a model that takes as inputs GPS traces, the USGS elevation service, and the detailed road database from OpenStreetMap. The overall caloric estimation error is 60% smaller than other smartphone-based approaches. Finally, the smartphone can aggressively duty-cycle its GPS receiver, reducing energy consumption by 57%, without any degradation in the accuracy of caloric expenditure estimates. This is possible because we can recover the bike’s route, even with fewer GPS location samples, using map information from the USGS and OpenStreetMap databases.

iBrush: Writing Chinese Characters with a Flashlight and a Wireless Sensor Grid



Last modified at Mar 9, 2018 by Andong Zhan