Receive a weekly summary and discussion of the top papers of the week by leading researchers in the field.

In Spine ; h5-index 57.0

STUDY DESIGN : Retrospective, randomized, radiographic study assessing age-related changes (ARCs) on lumbar magnetic resonance imaging (MRI) using an ensemble method.

OBJECTIVES : This study proposed to develop a novel reporting method to calculate a predicted "age estimate" for the ARC seen on lumbar MRI.

SUMMARY OF BACKGROUND DATA : Lumbar MRI reports include pathological findings but usually not the prevalence data of common findings which has been shown to decrease the need for narcotics in the management of non-specific lower back pain (NSLBP). Comparing the normal age estimation for lumbar spine degenerative changes/ARC on MRI and comparing this to the patient's real age may improve patient outcome in the management of NSLBP.

METHODS : A total of 60 lumbar MRI were taken from patients aged between 0 and 100 years. Lumbar MRI features reported as associated with age on review of the literature were measured on each MRI and statistically evaluated for correlation with age. Factors found to be associated were then entered into an ensemble model consisting of several machine learning techniques. The resulting ensemble model was then tested to predict age for a further 10 random lumbar MRI scans. One further lumbar MRI was then assessed for observer variability.

RESULTS : Features that correlated with age were disc signal intensity, the appearance of paravertebral and psoas muscle, disc height, facet joint size, ligamentum flavum thickness, Schmorl nodes, Modic changes, vertebral osteophytes, and high-intensity zones. With the ensemble model, 80% of estimated spinal age were within 11 years of the subjects' physical age.

CONCLUSION : It would appear that the intervertebral discs, and many other structures that are subjected to loading in and around the lumbar spine change their lumbar MRI appearance in a predictable way with increasing age. ARC on lumbar MRI can be assessed to predict an "expected age" for the subject.Level of Evidence: 2.

Sneath Robert J S, Khan Atif, Hutchinson Charles

2021-Jul-01