Mayo Clinic helps radiologists pinpoint potential problems

The Need

Mayo Clinic sought to transform the way it processes and interprets medical imaging results to enable more accurate detection.

The Solution
Mayo Clinic worked with IBM to create new analytic algorithms and integrate them seamlessly into its radiology workflow.

What makes it Smarter
Powerful algorithms pinpoint potential problem areas within medical images and flag them based on the probability of abnormality.

The Result
“Our imaging and radiology work will help provide faster and better information for our physicians and improved treatments for our patients.” – Bradley Erickson, M.D., Ph.D., Head of Mayo Clinic’s Radiology Informatics Lab, Co-Director MI3C

Business Benefits

• 50 times reduction in motion-correction processing, enabling radiologists to provide results in minutes, not hours
• Estimated 25 percentage point increase in diagnostic sensitivity for detection of brain aneurysms through intelligent flagging of high-risk areas
• Improved patient outcomes through improved diagnostic capabilities
• Improved radiologist productivity and effectiveness through the integration of detection testing into workflow
• Ability to extend the analytical model to other imaging modalities by virtue of SOA-based development approach

• IBM WebSphere® Process Server
• IBM DB2® for Linux®, UNIX® and Windows®

• IBM System x®
• IBM BladeCenter® HS20
• IBM System Storage^TM DS8100

• IBM Systems and Technology Group

• Instrumented: Imaging results are captured and trigger execution of algorithms that identify and mark suspicious areas within image results.
• Interconnected: Enhanced imaging results are seamlessly integrated into radiology process workflows and accessed for interpretation.
• Intelligent: Embedded insights inserted by the algorithms improve radiologist visibility to hidden, potentially significant conditions.

For radiologists to be able to detect subtle changes over time that can indicate problems, medical images must be crisp and as close to “apples-to-apples” comparisons as possible. Getting them to this point generally requires special algorithms, some to correct image blur or distortion commonly caused by the patient’s breathing, heartbeat or movement, others to properly align images when the patient’s position or equipment differences caused a difference in image orientation. Given the complexity of these algorithms, it can take hours to run them and determine whether the original images were correctable. This means that patients would have to be rescheduled if images could not be corrected electronically.

Mayo Clinic attracts patients from around the world. By necessity, such patients tend to compress a lot of activity—including multiple tests and specialist consultations—into a very short time. This dynamic is one reason the culture of speed and efficiency prevails within Mayo Clinic’s operations. Pooling their complementary expertise, Mayo Clinic and IBM modified the image correction algorithms to take full advantage of the computational capabilities of the IBM BladeCenter QS21 Cell Blade servers at the core of the MI3C solution. Using the Cell Broadband Engine™—a powerful processing architecture developed by IBM, Sony and Mitsubishi—the MI3C executes the algorithms 50 times faster, enabling radiologists to interpret results within minutes of the patient’s test.

Weaving intelligence into the imaging workflow
In addition to shortening the image processing cycle, Mayo Clinic also saw an opportunity to use the MI3C expertise to improve the efficiency of the image comparison and analysis process executed by radiologists. To prove the concept, Mayo Clinic and IBM targeted the especially complex and manual process of identifying brain aneurysms, the abnormal bulging outward of arteries in the brain. Simply having the technical capability to perform this analysis wasn’t enough. To maximize its clinical effectiveness, Mayo Clinic and IBM needed to weave this automatic “read” seamlessly into the radiology workflows.

Using IBM WebSphere Process Server to model and orchestrate the process flow, IBM worked with Mayo Clinic radiologists to design a medical imaging workflow that enables radiologists to run detection algorithms in the course of their typical activities using the same equipment. While Mayo Clinic selected Magnetic Resonance Angiography (MRA), which produces pictures of blood vessels in the brain, the solution was designed to be easily configured to any type of imaging technology. When MRA images are acquired, they are automatically routed to MI3C servers, where algorithms align the images properly and analyze them (based on Mayo Clinic criteria) to find and visually mark potential aneurysms. The results of the detection algorithm can then be routed to Mayo Clinic’s PACS (Picture Archiving System) network, where they are viewable on the radiologist’s workstation. Throughout the process, images are stored in an IBM DB2 for Linux, UNIX and Windows data server, while workflow logic is run on IBM System x servers and IBM System Storage.

The impact on the new workflow is that radiologists are able to better verify their interpretation of MRA images. For Mayo Clinic, the benefit is neither time nor cost savings, but the improvement in diagnostic accuracy—and therefore the quality of patient care—that the new imaging workflow enables. In practice, the proof-of-concept solution generated a 95 percent accuracy rate in detecting aneurysms, compared with 70 percent for manual interpretation (“The Detection and Management of Unruptured Intracranial Aneurysms.” Authors: Wardlaw JM, White PM ... Brain 2000;123(pt2):205-21). On the heels of its brain aneurysm detection success, Mayo Clinic expects to apply the same approach for other tests with challenging radiology detection requirements such as better characterization of breast lesions with MRI.

The MRA solution was designed with this flexibility in mind. Emblematic of the SOA approach IBM followed, the solution employs a library of reusable Web Services that perform specific functions. The workflows at the heart of the solution, orchestrated by IBM WebSphere Process Server, invoke these services as needed. The solution’s reliance on reusable assets will make it much easier for Mayo Clinic to extend this service model to other radiological processes. Bradley Erickson, M.D., Ph.D., head of Mayo Clinic’s Radiology Informatics Lab and MI3C co-director, sees the clinic’s medical imaging initiatives as reinforcing its constant and single-minded emphasis on improving patient care. “Our imaging and radiology work will help provide faster and better information for our physicians and improved treatments for our patients,” says Erickson.

For more information
To learn more about how IBM can help you transform your business, contact your IBM sales representative or IBM Business Partner.

Visit us at:

ibm.com/healthcare