Sessions at OSCON 2010 with video on Thursday 22nd July

For years you’ve been leaving your computers turned on in order to process data packets for UC Berkeley’s SETI@home – that’s great! Please keep it up!

Did you ever want to get more involved?

Do you think about the ‘why’, ‘how’, ‘what if?’ of SETI and want to offer improvements? Do you like contributing to open source code development projects? Do you think SETI projects should be looking for different kinds of signals, and do you have great algorithms for finding them in noise? Are you good at seeing hidden patterns, do you have some time to examine data coming from SETI observations with the Allen Telescope Array in real time?

Do you wear headphones and listen to music while you work to sharpen your concentration? Could you imagine listening to data instead and responding to anomalies? If any of these ideas described you, then you should check out setiQuest.org – we want to make the SETI Institute’s SETI programs on the Allen Telescope Array better and more comprehensive. We have some ideas of things we can/should do, but we need resources and your help to get the job done. We have some ideas of things we’d like to do, but we don’t yet understand whether or how we can make them happen; maybe you hold the key. There are things you know about signal processing, data manipulation and distribution, and crowdsourcing that we need you to help us learn and implement, so we can improve our searches.

In short, thanks for all those years of paying for the electrons coming out of your wall socket, but now we also want your ‘thinkons’! We are eager to move ahead and conduct better searches so that together we can find any signals that are out there! We want you to become part our team and start adding ‘Earthling’ to your personal profiles when you identify yourself to the world. It’s time to change the humanity’s point of view of who we are (individually and collectively) to one that is more cosmic and inclusive.

Modern health care information is highly heterogeneous, distributed, and difficult to leverage in downstream analyses that are critical to quality of patient care, including diagnosis, treatment, and outcome prediction. There are many data types to deal with (medical flowsheets, free text notes, lab results, measurements recorded by clinicians and automatically captured by instruments, waveforms) and a variety of competing standards and formats for organizing and transmitting this data (HL7, SNOWMED, ICD-9, UMLS) – not to mention proprietary and vendor-specific stores. The information landscape is growing at a rapid pace, but medical informatics nevertheless lags far behind other domains in its ability to leverage massive amounts of data to improve service and build effective data-driven tools. One of the largest obstacles to a medical information revolution is inaccessible data locked in proprietary silos, unavailable to clinicians, researchers, and information systems alike. Any progress will require that this information be unlocked from independent systems that collect and manage it.

NASA’s Jet Propulsion Laboratory (JPL) and the Whittier Virtual Pediatric Intensive Care Unit (Whittier VPICU) group at Children’s Hospital Los Angeles have been collaborating since 2003 in the development of open source grid software for the description, organization, management, sharing, and analysis of highly granular data from pediatric intensive care units (PICUs). We are leveraging one of NASA’s flagship grid software technologies, the Object Oriented Data Technology (OODT) framework, to assist in this regard. OODT is hosted at the Apache Software Foundation (ASF) and is a podling within the Apache Incubator. OODT provides a set of loosely coupled components for data capture, discovery, access, and distribution that can be instantiated and connected via modern web protocols and data formats (REST, RDF, etc.) for a particular deployment in a domain. The OODT framework enables CHLA clinicians, researchers, and software to access large amounts of data from a variety of proprietary sources (e.g., hospital-wide EHR systems, bedside monitors, unit-specific applications, homegrown databases) in a unified manner and can be extended readily to enable sharing of data between institutions.

Besides OODT, our project plans to utilize other open source software, including search technologies from Apache Lucene (Solr, Tika, etc.), and common platforms (Ubuntu, Redhat, etc.) as a means for building reliable, value-added software at low cost. We also have begun development of a common semantic architecture for describing similar data from disparate sources, and we plan to continue expanding this into a comprehensive ontology for all PICU clinical data and to share it as a free, open standard.

Our long-term goal is to construct a national distributed data-sharing network to drive the next generation of research into data-driven decision support tools and comparative effectiveness and outcomes analysis. Whittier VPICU and JPL both have experience in building such national collaborative networks: CHLA helped to develop a network of over 80 PICUs that share limited datasets for performance evaluation; JPL has used OODT to construct a variety of scientific data sharing networks, including the National Cancer Institute’s Early Detection Research Network (EDRN).

In this talk, we will discuss the current state of OODT and its successful deployment in projects such as EDRN, the motivation for its use as a means of unlocking and unifying health data at CHLA and across other institutions, and our experiences in leveraging open source software to provide a foundation for building advanced data-driven clinical decision support systems to improve the quality of pediatric intensive care going forward.