AI developed for diagnosing lumbar central canal stenosis with abdominal CT scans

A Seoul National University Bundang Hospital (SNUBH) research team has developed an artificial intelligence (AI) program capable of diagnosing lumbar central canal stenosis using abdominal computed tomography (CT) scans.

The program achieved a diagnostic accuracy of 84 percent, comparable to that of radiology specialists, and holds significant promise for patients who cannot undergo magnetic resonance imaging (MRI).

Lumbar central canal stenosis is a condition where the spinal canal in the lower back narrows, compressing the spinal cord or nerve roots. It affects about 30 percent of individuals aged 60 and older in Korea, with prevalence increasing with age.

Key symptoms include lower back pain and numbness in the legs or hips, and severe cases can lead to loss of nerve control over bowel and bladder functions.

Accurate diagnosis of lumbar central canal stenosis can be challenging due to its symptom overlap with other conditions like a herniated disc, which often improves with conservative treatments. While MRI is the primary diagnostic tool, it is unsuitable for patients with implanted metal devices such as spinal cord stimulators or pacemakers, necessitating the use of CT scans.

Recognizing the limitations of MRI, the SNUBH team, led by Professors Lee Joon-woo and Lee Young-joon of the Department of Radiology, sought to develop an AI-based diagnostic tool using abdominal CT scans.

Abdominal CT is a widely used and cost-effective imaging method that is unaffected by metal implants and often captures spinal structures as part of its scope.

The team trained the AI algorithm using CT data from 109 patients who underwent both abdominal and lumbar CT scans.

The program automatically analyzed the CT images and classified patients as having stenosis if the dural sac within the lumbar spine measured 100 mm² or less.

The algorithm demonstrated an 84 percent diagnostic accuracy, matching that of radiologists interpreting lumbar CT scans. For severe lumbar central canal stenosis cases, the diagnostic accuracy exceeded 85 percent, and the AI successfully identified asymptomatic and mild cases that are traditionally difficult to diagnose using CT alone.

The team expects that the AI program’s integration into clinical settings could simplify the diagnosis of lumbar central canal stenosis cases, potentially allowing for diagnosis through routine abdominal CT scans conducted for other medical purposes, such as general health check-ups.

This approach could save time and reduce the need for additional imaging.

“Abdominal CT is one of the most common imaging techniques for examining the abdomen and internal organs,” Professor Lee Joon-woo said. “Noting that the lumbar spine is often captured during these scans, we developed this program to identify lumbar spinal stenosis using existing data.”

Professor Lee Young-joon also said, “This study demonstrates that an AI program can achieve diagnostic accuracy close to that of radiology specialists.”

The team aims to develop a comprehensive diagnostic program that addresses all spinal disorders in the future, he added.

The study was published in Skeletal Radiology.