Chairs: May Wang and Shankar Subramaniam, UCSD and IEEE EMBS BHI Technical Committee
Advanced Health Informatics is one of 14 Grand Challenges of 21st Century identified by US National Academy of Engineering.
Introduction to the Session:
This session will focus on the common underlying problem in biomedical and health informatics. Whether it is an electrocardiogram or an electronic medical record or the genome of a patient, the most important basic problem is obtaining knowledge that can be useful to practicing clinical researcher. In the first talk, Dr. Winslow will highlight the inaccessibility to one of the most basic physiological measurement, namely, an electrocardiogram for clinicians to interoperate with other phenotypes and obtain useful information that can be used in diagnosis and treatment of a patient. The new methods developed in his laboratory, not only permit interoperable access to this important data, but also permits tools to operate on the data to obtain practical knowledge. In the second talk, Dr. Livingston addresses one of the most challenging problems faced by physicians. The EMR turns out to be one of the most useful yet highly frustrating entity, where a clinician literally has to search for a needle in a haystack to make sense. Dr. Livingston offers the myriad challenges faced by the physician and talks about the imminent challenges this offers to a data scientist and engineer. In the third talk, Dr. Avi Ma’ayan addresses one of the most important issues that is the foundation for personalized medicine. We can sequence human genome and the transcriptome inexpensively and effectively, but what does this mean? How can we infer from this exploding data, signals that specific normal and pathological function in humans? Dr. Ma’ayan addresses how the resources he has developed enable transformation of data to knowledge which will have an impact for personalized medicine.
WaveformECG: A Web-Based Platform for Managing, Visualizing, Annotating, and Analyzing ECG Data
Raimond L. Winslow, Ph.D.
The Institute for Computational Medicine and Department of Biomedical Engineering, The Johns Hopkins University School of Medicine and Whiting School of Engineering, Baltimore MD 21218
The electrocardiogram (ECG) is a measurement of time-varying changes of body surface potentials produced by the underlying electrical activity of the heart. It is the most commonly collected data in heart disease research. This is because changes of the ECG waveform reflect underlying aspects of heart disease such as intraventricular conduction, depolarization, and repolarization disturbances, coronary artery disease, and structural remodeling. Many studies have investigated the use of different ECG features for predicting risk of coronary events such as arrhythmia and sudden cardiac death, however it remains an open challenge to identify markers that are both sensitive and specific. Commecial vendors have developed information systems that accept, store, and analyze ECGs acquired using local monitors. The challenge in applying these systems in clinical research is that they are closed systems that do not provide APIs by which other software systems can query and access their stored digital ECG waveforms for further analyses, or means for adding and testing novel data processing algorithms. They are designed for use in patient-care, rather than for clinical research. We have developed WaveformECG to address this un-met need. WaveformECG is a web-based tool for managing and analyzing ECG data developed as part of the CardioVascular Research Grid (CVRG) Project funded by the NIH National Heart, Lung and Blood Institute. With WaveformECG, users can browse their file system to upload ECG data encoded in a variety of vendor formats for storage. WaveformECG automatically extracts and stores the ECGs as a time-series. Once data is uploaded, a browser can be used to select, view and scroll through individual digital ECG lead signals. Points and time intervals in ECG waveforms can be annotated using ontology from the Bioportal ontology server operated by the National Center for Biomedical Ontology, and annotations are stored with the waveforms so they can be retrieved. This enables features of interest to be marked and saved for others to refer to. Users can select groups of ECGs for computational analysis using multiple algorithms. At the click of a button, analyses can be distributed across multiple CPUs to decrease processing time. Analysis results can be viewed and downloaded. WaveformECG has also been integrated with the I2B2 clinical data warehouse system. This bi-directional coupling allows users to define study cohorts within I2B2, analyze ECGs within WaveformECG, and then store analysis results within I2B2. This talk will present key features of WaveformECG, the ways in which it has been used to address clinical research problems, and extensions that are planned for the future.
The EMR: What Went Wrong and How Can We Fix It?
Although electronic medical records (EMRs) were supposed to greatly improve the quality of healthcare, they not only have not accomplished that goal, but they are considered to be the most significant detriment to the practice of medicine today as expressed by most clinicians. When trying to assess what went wrong with the computerization of medical record systems it is important to consider history. The first consideration involves how medical records evolved and the other is how federal policy influenced medical record documentation. In the days before there were medical textbooks, clinicians kept notes about their patients so that they could try to understand various disease patterns. Those notes evolved into medical records enabling physicians to keep track of their patient’s progress and communicate important information about patients to other clinicians. Billing for services eventually got linked to a physician’s involvement in patient care requiring documentation of minute details regarding a physician’s involvement in every patient encounter. Computers were brought in to assist with this tracking and these billing systems became the foundation for today’s EMRs. As a consequence the modern EMR is mostly oriented towards being a billing system rather than a medical record-keeping system. In recent years, the Office of the National Coordinator established policies requiring the use of EHRs to fulfill a variety of functions. They pursued a very aggressive course, essentially forcing the entire medical community to adopt EMRs. This happened too quickly, resulting in the EMR systems having limited functionality. Most EMRs in use today fall short of the physician’s expectations and physicians view their interface with EMRs as one of the most detrimental aspects of their occupation. There is a pressing need for engineers and clinicians to come together to develop new EMRs that assist the physicians daily activities rather than get in their way. There is also a need for policymakers to reconsider their requirements for EMRs.
Collecting and Exploring Gene Signatures
Avi Ma’ayan, Ph.D.
Omics technologies produce genome-wide snapshots of the genome and proteome. To understand the results from such studies differentially expressed genes and proteins are placed in the context of prior knowledge. Such prior knowledge can be organized as gene-set labeled by a common functional term. We have assembled over 180,000 such annotated gene sets from over 70 resources. Besides developing tools to query this gene set knowledge-base, I will discuss how this organized knowledge can accelerate discovery by learning the patterns within it.