Big Data Expected to Have Big Impact on Diagnostic Imaging

May 17, 2014

Radiology can see significant value from big data, according to experts at SIIM 2014.

LONG BEACH, Calif. - It will be absolutely critical in diagnostic imaging to be able to personalize the specific recommendations for your patient and to tailor the recommendations to his or her characteristics based on specific parameters, Eliot Siegel, MD, professor of diagnostic radiology at the University of Maryland School of Medicine said at SIIM 2014.

This was the underlying sentiment about the trend that is big data at SIIM, where medical imaging is seen to play a substantial role in the era of big data and personalized medicine, an era that poses a challenge in implementation.

Big data’s definition tends to vary from person to person, but essentially it refers to the analysis of large volumes of data. The power of big data is the ability to take into account multiple datasets and identify patterns, which can lead to more accurate diagnoses, a significant value in the medical field.

Big data is characterized by variety (structure to unstructured data, textual or multimedia, data graphs), velocity (how fast data comes in) and volume (massive amounts of data points, new and historical data), explained Kinson Ho, MSc, a product architect at Agfa Healthcare.

Not all of this is new, the increase in velocity is not outrageous for most of us, and the increase in volume is offset by the increase in storage and processing, Ho said. The real game changer of big data is starting something that’s specialized: special hardware with special processing, and now big data technology develops an opportunity to use common hardware. Accessibility is no longer an issue, the true power of big data is its ability to transform high end processing to a commodity, Ho said.

In many ways, we are in a continuum of Flintstone’s medicine, transitioning to a Google era, from purely anecdotal and textbook, to search engine, and finally, to actually date-driven practice, Siegel said.

Big data was practiced long before the term existed, said Henri "Rik" Primo, director of strategic partnerships at Siemens. Data mining already exists in databases like Google and Yahoo.

“When I think about big data, I think about data that is very difficult for people to consume and extract from…so let’s talk about how you extract data that’s not easily accessible,” said Khan Siddiqui, MD, CEO/CTO of higi llc.

There are three pillars of data, Ho explains, data management, which refers to managing the data in a scalable way; search, which should be built around a search engine with fast searching capabilities; and data processing, which allows for the advanced processing framework. The game changer in these pillars, Ho said, is that they are all open source.

Only a tiny fraction of information on a patient is used in practical treatment of the patient, Primo said. The advantage to big data is the ability to access existing patient records, which are valuable resources for an automated outcome analysis, but the current records are typically unstructured and poorly organized, and therefore require a time-consuming interpretation.

A strategic plan for big data in medical imaging, Primo explains, is to dynamically integrate medical images, in vitro diagnostic information, genetic information, electronic health records and clinical notes into a patient’s profile. This provides the ability for personalized decision support by the analysis of data from large numbers of patients with similar conditions.

Big data has potential to be a valuable tool, but implementation can pose a challenge. Build a medical report with context-specific and target group-specific information that requires access and analysis of big data, Primo suggests. The report can be created with the help of semantic technology, an umbrella term used to describe natural language processing, data mining, artificial intelligence, tagging and searching by concept instead of by key word.

“As time goes on, we are going to be having greater challenges in imaging informatics, even than anyone else in big data, our data is far more complex, far more high dimensional…but I think we’ll be able to take imaging and take medicine to the next generation,” Siegel said.

Radiology can add value to the era of big data by supporting implementation of structured reports, Primo said. “Structured reports are needed to improve quality and collaboration between radiology and other disciplines…this will facilitate indexing and data mining and create semantic interoperability with all departments.”