Deep learning shows promise for cutting overutilization of lumbar imaging, surgery

Researchers in Germany have piloted the use of AI to distinguish patients likely to benefit by surgery for low-back pain from those who should do better skipping the OR in favor of conservative therapies.

If the team’s approach proves out in larger studies, this ability alone might justify the technology’s adoption. That’s because lumbar imaging and interventions are often named among healthcare’s iffiest expenses due to the hit-or-miss predictability of the outcomes.

However, there’s more. The researchers believe their approach is translatable to avenues of medicine beyond orthopedic spine care.

Lead author of the study is André Wirries, MD, of the Hessing Foundation in Augsburg. Senior author is Samir Jabari of Erlangen University Hospital in Erlangen. European Spine Journal published the work Oct. 13.

The team used data from 60 patients with herniated lumbar discs to train and test a deep learning algorithm. The goal was to get the model to accurately predict scores on an established disability indicator, the Oswestry Disability Index (ODI), as recorded six months after surgery or the start of conservative therapies. These consisted of inpatient and outpatient physical therapy combined with oral painkillers and/or spinal injections.

“A 100% accurate prediction of ODI range could be achieved by dividing the ODI scale into 12% sections,” the authors report. “A maximum absolute difference of only 3.4% between individually predicted and actual ODI after six months of a given therapy was achieved with our most powerful model.”

Wirries and colleagues further found the AI application allowed clinical decision-makers to compare the actual patient values after six months with the prediction for the alternative therapy, showing deviations up to 18.8%.

“We believe that the approach of a supervised artificial intelligence will improve the predictability of a therapy outcome and thus help to individualize therapy recommendations for patients such as those with a disc herniation,” the authors conclude. “This approach … can serve as a basis for further developments of AI, not only in the field of spinal therapy but also in many other areas of medicine where randomization or inclusion of high patient numbers is not feasible.”

The study is available in full for free.