Diagnostic imaging runs the risk of becoming invisible in medicine unless it changes with the times, health IT expert Dr. Eliot L. Siegel warned conference-goers at the New York Imaging and Informatics Symposium on Monday.

In the era of evidence-based care, procedures will need to be tied to metrics around performance, efficacy, safety and cost-efficiency driven by the collection of big data.

But this poses a problem for radiology, since it's difficult to mine much information from a radiology study compared to, say, a blood test.

"Many people think of radiology as artwork," Siegel said, adding that radiology runs the risk of becoming commoditized unless it can deliver personalized care using detailed information about the patient. "The problem is, many of the rules we have are one size fits all."

Getting personal

Personalized medicine generally refers to the ability of what Siegel calls the "omics" — genomics, metabolomics, and proteomics — to provide tailored medical care to patients based on their unique genetic and physiological makeup.

Indeed, personalized, genetics-based medicine has received high-profile attention lately. Movie actress Angelina Jolie boosted awareness on the topic when she made headlines this May for undergoing a mastectomy based on a genetic test.

This trend is only expected to accelerate further as patients' genetic data become easier and less costly to access. The cost of mapping human DNA has already plummeted drastically in a short span of time, from $10 million in 2007 to $100 in 2012.

To a certain extent, the imaging community has already tried to seize on the term "personalized medicine" when referring to nuclear imaging modalities such as PET/CT and SPECT/CT, which can actually provide information about the patient on a molecular level.

But could traditional imaging also provide personalized care? According to Siegel, the answer is yes.

Diagnosis beyond imaging

To stay medically relevant, imaging studies must go beyond answering immediate clinical questions by collecting other, routine information about the patient. "I would know whether this patient had osteoporosis or coronary artery calcification," Siegel said.

For example, while a patient was undergoing a scan for an unrelated part of the body, the image report might register that the patient's aorta had reached a size threshold that put the patient at higher risk for developing an aneurysm and alert the doctor.

Eventually, Siegel envisions every patient's clinical course and data being saved and incorporated by decision-supporting algorithms to help physicians make better choices when treating patients. "Every patient's medical care would become a clinical trial," he said.

Perhaps ironically, providing personalized, individual care to patients requires that doctors collect data from as many people as possible to gain information on similar types of patients. That way, they can have a better sense of how to treat patients based on factors like their age, sex, geographical location and medical history.

Mining your own business

But there are some challenges to overcome before informatics-based imaging can become a reality.

For one, patient records are not currently set up in a way to allow for a seamless exchange of information, since the majority of patient data is taken down in an unstructured free text format. It's also not possible to search for terms within records, and information is often entered with nonstandard abbreviations and misspellings.

Restricted access to research data from big clinical trials is another obstacle to informatics-based diagnostic imaging, Siegel said.

Making use of data from the National Lung Screening Trial could potentially be a huge boon to practitioners, allowing other factors such as geographical location, age, or nodule shape to help refine current Fleischner Society diagnostic guidelines, which are based on nodule size.

On the whole, however, research data from these trials remain inaccessible. Even when this data is technically accessible, researchers must jump through cumbersome bureaucratic hoops to request permission for access with long periods of wait, often with no guarantee of success.

Playing tag

To see how much further diagnostic imaging's role in medicine could be expanded, Siegel is currently spearheading the Vasari Rembrandt Project, which explores correlations between patient genetic expression and tissue features in radiological images.

Researchers took MRI scans of brain tumors and categorized and tagged them according to over 30 different parameters. They were able to cross correlate imaging features with specific genes to predict genetic data. Siegel sees this as the way of the future in imaging.

"We need to create a mechanism to tag content of images and correlate it with genetic data," he said. Ideally, features in imaging exams and their changes over time would be rigorously quantified, tagged and made searchable, allowing algorithms to integrate this information alongside other data sets to help predict things like patient survival or disease expression.

Genomics meets imaging

Ultimately, Siegel sees a real opportunity for radiologists to seize on the current trend of genetics-based research. Gene analysis on its own presents researchers with a challenge because it's so complicated — gene expression is affected by a host of other phenomena, including epigenetics and the presence of other genes.

Automated medical images can play a role as a supplemental biomarker to help clarify genetics-based medicine, but only if radiologists are prepared to lobby for studies to share data, standardize and rigorously quantify imaging exams, and work to standardize EMR systems for better data-mining, according to Siegel.

"Personalized radiology brings tremendous added value — we should reinvent ourselves in an ever-changing world of decision-supported data and lead the way into the next generation of information systems for major improvements in health care," he said.