Philips’ MR software learns as it goes

Philips Medical Solutions has developed a learning program that adjusts the planning, scanning, and processing of MR exams to match the preferences of operators and physicians. The SmartExam promises to decrease overall exam time, while increasing consistency and reliability.

SmartExam eventually will be applied to all types of exams, including cardiac MR. For the present, however, it has been optimized only for brain, spine, and orthopedic applications. It runs on Philips high-field scanners, including its 3T and 1.5T platforms, as well as the new 1T open.

The company is setting up a number of reference sites where prospective customers can see SmartExam in action. The technology is already operating at about 10 institutions that have served as beta sites for its development.

"In the beginning, people (at these sites) wondered should they give control to the system," said Jacques Coumans, Ph.D., vice president of MR global marketing for Philips Medical Systems. "But now they have seen there is complete consistency (from patient to patient)."

Patient exams can be compared more easily, because exactly the same slice acquired during the first visit is acquired in subsequent ones, he said. Exams in general are also easier to interpret because once SmartExam has learned the preferences for specific procedures, each is done the same from patient to patient.

"Exams will come up the same way every time," he said.

In this way, SmartExam promises to enhance productivity, but it has much broader significance. The learning program begins with exam settings stored on so-called SmartCards. These settings were developed by expert luminaries for specific exams. The learning algorithms refine the settings to return results that match those of the staff at individual institutions.

This reduces the learning needed to run a new machine and reduces the risk of problems in transitioning to a Philips MR scanner, for example, from equipment made by a different vendor. It also raises the possibility that SmartExams could be designed for applications outside radiology, such as cardiac MR.

The implications of this approach go even further. Software with the ability to learn the preferences of operators could benefit users of virtually any modality, as well as PACS/IT.

SmartExam is designed to adjust for the natural anatomic variability from patient to patient. It also adjusts for differences due to the way a patient is positioned in the scanner.

"SmartExam understands that the (operator) wants to see, for example, the midcoronal slice starting after the ventricular system (of the brain)," he said. "So, for each of these patients, SmartExam makes certain that this is what is seen, even though there is variation from patient to patient."

SmartExam refines the exam further for individual scans, which typically involve a half dozen or so acquisitions. Adjustments might be needed between each acquisition. These are noted by the SmartExam algorithm, and preferences are recorded.

The process begins with settings stored in Philips' "ExamCards," software files that contain settings optimized by users of Philips MR scanners for different types of scans at different field strengths. On a cue from the operator, SmartExam (referring to the preset parameters in the ExamCard) starts a survey scan, detecting anatomic landmarks and planning the diagnostic scan related to the patient's position in the magnet.

"We have designs and algorithms that allow the structure (being examined) to be found. Consequently, the mapping of the slice positions and the anatomic landmarks are correlated," Coumans said.

The operator then defines the spatial resolution and launches the diagnostic scan. SmartExam makes adjustments automatically to maintain the set resolution and image contrast regardless of variations in the field-of-view.

Changes made during the exams - for example, slice thickness, field-of-view, orientation, and angulation - are recorded and built into a SmartExam specific for the type of examination and the technologist running it. The learning continues with each successive patient undergoing that specific examination until the SmartExam requires minimal or no adjustments.

"At some point, the tech and physician say, 'This is how I would like my scans done,'" Coumans said. "Then the learning phase is over. The ExamCard has a 'star,' which makes it a SmartExam card. From that moment onward, every time you drag the SmartExam card out of the library into the scan area (of the console), the imaging is executed - the planning, the scanning, the whole process is done with a single mouse click - and you are free to do other things."

In the end, SmartExam saves time - up to 30%, according to Coumans. By doing so, SmartExam improves patient comfort, because it shortens the time of the exam. It also boosts the "comfort level" of administrators who decide to buy Philips high-field MRs. Decision makers can select Philips MR systems, he said, even if their staff are familiar with products made by another vendor, because SmartExam flattens the learning curve for new equipment.

"Bringing the SmartExam capability to Philips MR systems brings every tech to the expert level," he said. "That means replacing other systems with Philips' systems is not a nightmare or burden for the administrator or staff."

Potential benefits in new arenas eclipse advantages coming from routine use. MR is stretching into new clinical applications, notably cardiac imaging. Getting MR into cardiology is the ultimate goal for Coumans and his colleagues, as it moves the modality beyond the largely replacement-oriented market of radiology into a relatively untouched sphere.

"Currently, we see an enormous boom in CT doing the coronaries, but we see MR with no (ionizing) radiation having cardiac capabilities for function and structure, as well as the coronaries," he said. "SmartExam applied to cardiac sequences could bring simplicity to MR technology, such that cardiologists could easily adapt it."

SmartExam is not there yet. but it is on the way. Philips engineers at the company's research lab in Hamburg, Germany, are now applying the pattern recognition algorithms optimized for neuro and orthopedic applications to "time varying anatomies," Coumans said, of which the heart is the most challenging.

"We have started with a version (of SmartExam) that addresses 70% of all exams, but we have set our eyes on the holy grail, which is cardiac imaging," he said.