Newswise — Because the spine is made up of repeating elements that look alike, surgeons can mistakenly operate on the wrong vertebra. To avoid this, Johns Hopkins researchers have developed a software program that works seamlessly with currently available procedures to assist a surgeon’s determination of which vertebra is which. Results from its first clinical evaluation show that the LevelCheck software achieves 100 percent accuracy in just 26 seconds. Details of the study will appear in the April 15 issue of the journal Spine.
“Wrong-level spine surgery is never meant to happen,” says Jeffrey Siewerdsen, Ph.D., a professor of biomedical engineering at Johns Hopkins and a member of the Armstrong Institute for Patient Safety and Quality. “But it happens nearly four times a week in the U.S.”
Surgeons go to great lengths to get their procedures right, because mistakes are costly to patient health. They can result in pain, require follow-up surgeries, and create instability or degeneration of the spine, according to Jean-Paul Wolinsky, M.D., an associate professor of neurosurgery and oncology at Johns Hopkins and co-author of the study.
Before a standard spinal operation, patients receive a diagnostic CT or MRI scan that the surgeon uses to plan the surgery. Once the patient is on the operating table, often days later, the surgeon typically counts down from the skull or up from the tailbone to determine which vertebra to operate on, often marking the patient’s anatomy with thin metal pins. These pins are visible in an X-ray image taken in the operating room to verify the target site. But the doctor’s initial planning on the preoperative scan is not visible in the X-ray image, leaving room for error, particularly when working on challenging cases exhibiting missing or extra vertebrae, a loss of anatomical landmarks from previous surgeries, or other anomalies.
LevelCheck uses a standard desktop computer outfitted with a graphics processing unit, commonly used for video games, to align a patient’s 3-D preoperative CT image with the 2-D X-ray image taken during surgery. The result is an X-ray image showing the pins that act as landmarks for the surgeon, overlaid with the planning information from the CT scan.
“LevelCheck does not replace the surgeon’s expertise. It offers helpful guidance and decision support, like your GPS,” says Siewerdsen.
To test its accuracy, the team analyzed pre- and intraoperative images of 20 consecutive patients who had undergone spine surgery. By shifting the images, they simulated 10,000 surgeries and measured how long the software needed to correctly line up the images 100 percent of the time: only 26 seconds.
“This study is the first to demonstrate that LevelCheck works with real patient images,” says Siewerdsen. “It shows that the software can deal with challenges like changes in patient anatomy and the presence of surgical tools in the X-ray image.”
Sheng-Fu Lo, M.D., evaluated the results to find what factors can cause the software to fail. “The software doesn’t always get it right if it is stopped early,” says Lo, “but given 26 seconds or more, LevelCheck found the right level every time.”
“We can’t eliminate the possibility of wrong-level surgeries,” says Wolinsky, “but this is an additional level of security — an independent check — that works quickly within our standard surgical workflow. Although LevelCheck in its current form requires a preoperative CT scan for most patients, the benefit is well worth it.”
In its early development, LevelCheck won an award from the North American Spine Society for Spine Technology of the Year. Work now underway by the research team includes a larger evaluation with images from 200 patients and testing in real time to measure its effects on time, workflow, accuracy and safety.
Other authors of the report include Yoshito Otake, Varun Puvanesarajah, Adam Wang, Ali Uneri, Tharindu De Silva, Benjamin Elder, C. Rory Goodwin, Thomas Kosztowski, Jason Liauw, Mari Groves, Ali Bydon, Daniel Sciubba, Timothy Witham, Nafi Aygun and Ziya Gokaslan of the Johns Hopkins University School of Medicine; and Sebastian Vogt and Gerhard Kleinszig of Siemens Healthcare in Germany.
This work was supported by a grant from the National Institute of Biomedical Imaging and Bioengineering (R01 EB017226) and with partial funding from an academic-industry research partnership with Siemens Healthcare. Dr. Siewerdsen also served on a Scientific Advisory Board for Siemens Healthcare. The terms of this arrangement are being managed by the Johns Hopkins University in accordance with its conflict of interest policies.