Cascon 2012, Eliot Siegel, Opportunities in Healthcare for Big Data

At #Cascon 2012, Eliot Siegel, U. Of Maryland, Big Opportunities in Healthcare for Big Data and Real-Time Analytics.

In radiology, sophisticated medical devices, but primitive IT like the 1970s.

Typical doctor only reads 2 to 3 hours per month, while amount of data doubles over 5 years.

Last year spoke on Watson Q/A project, will speak quickly on opportunities.

Hottest application is on fraud detection.

Population analytics, particularly at payer level, then drill down e.g. diagnostic codes for at risk segments.

Evaluation of patient cohorts, e.g. with protheses.

Predictive / opportuntity analysis, possible readmission.

Diagnotic imaging, from terabytes now to petabytes. Cardiac study could have 1500 scans. Functional MRI with iterative reconstruction. How to scale up, send out.

Images have rich metadata. Automatic anatomy identification. Tumour characteristics, can tailor to patient and type of cancer.

Based on genomic, image and clinical data, could generate survival rate probability graph.

Patient signal information, e.g. pulse, oxygen. In future, will monitor specific patients, try to predict when they will come in.

Personalized medicine, DNA sequencing is under $1000, will continue to fall. Stephen Gold says analytical market will be $76 billion by 2015.

Watson: natural language, hypothesis generation, then evidence-based learning.

IBM patient care and insights, not just from electronic medical record, but also socioeconomic, where they’ve travelled, etc.

Pre-personalized medicine. In 2008, Kennedy had seizure, found malignant glioma. But no history, information systems not connected to compare similar cases.

Want to transcend clinical trials.

Compare Google search to question asking on probable diagnoses. Want analytic decision-support.

How to consume Electronic Medical Record? Unstructured.

Brought in Jeopardy team. Case of rash. No way to get links, will copy and paste. No person in charge of patient problem list.

Multiple clinical trials, with tens of thousands of patients. After published, no access to data.

Scott Duvall will talk about VA’s corporate data warehouse Vinci. Can’t export data, so need analytics in house, to answer questions.

Everyone guards their own databases, valuable for grant proposals

Million veteran project.

Want synthesis / display of complex information in EMR, would like clever graphical way, then drill down.

Would like to work in emergency room, before EMR is written.

So many amazing big data challenges in healthcare and medical imaging. Ten to fifteen years behind.