Elizabeth Hofheinz, M.P.H., M.Ed.
Indicating the time and potential error involved in manually measuring saggital alignment, researchers from the Departments of Orthopaedic Surgery and Radiology at the Icahn School of Medicine at Mount Sinai in New York tested a new fully automated measurement of spinopelvic parameters from lateral lumbar radiographs.
Their work, “Deep Learning Automates Measurement of Spinopelvic Parameters on Lateral Lumbar Radiographs,” appears in the December 1, 2020 edition of Spine.
Samuel K. Cho M.D. is an orthopedic surgeon at Mount Sinai and was a co-author on this work. He told OSN, “Our lab is dedicated to improving patient outcomes following spine surgery. Recently, AI has been shown to be a powerful tool in almost all sectors of society, including medicine. We decided to apply this technology to spine surgery and naturally led to our work.”
Previously, Dr. Cho and colleagues published work on automating the measurement of the L1-S1 Cobb angle with standard weight-bearing lateral spine radiographs.
Here, they sought to build on that research by improving spine segmentation results and label the thoracolumbar vertebral bodies and the sacrum and femoral heads. They also set out to measure pelvic incidence, pelvic tilt, and sacral slope—and validate their accuracy and precision against surgeon/manual measurements. Finally, they aimed to create an algorithm that can obtain measurements with instrumentation present.
The researchers used data from 816 sequential patients who received lateral lumbar radiographs between June 27, 2011 and July 23, 2012. They wrote, “Images from 816 patients receiving lateral lumbar radiographs were collected sequentially and used to develop a convolutional neural network (CNN) segmentation algorithm. 653 (80%) of these radiographs were used to train and validate the CNN. This CNN was combined with a computer vision algorithm to create a pipeline for the fully- automated measurement of spinopelvic parameters from lateral lumbar radiographs. The remaining 163 (20%) of radiographs were used to test this pipeline. 40 radiographs were selected from the test set and manually measured by three surgeons for comparison.”
“Algorithm measurements of L1-S1 cobb angle, pelvic incidence, pelvic tilt, and sacral slope were not significantly different from surgeon measurement. In comparison to gold standard measurement, the algorithm performed with a similar mean absolute difference to spine surgeons for L1-S1 Cobb angle (4.30 ± 4.14° vs 4.99 ± 5.34 ̊), pelvic tilt (2.14 ± 6.29 ̊ vs 1.58 ± 5.97 ̊), pelvic incidence (4.56 ± 5.40 ̊ vs 3.74 ± 2.89 ̊), and sacral slope (4.76 ± 6.93 ̊ vs 4.75 ± 5.71 ̊).”
“We were surprised to learn that the AI model that we developed worked as well as it did, meaning it yielded results on par with human surgeons,” said Dr. Cho to OSN. “AI is a powerful tool that can help improve the care that we provide to our patients. Our study is hopefully one example among many in terms of how it can actually do that.”