Introduction
Chronic low back pain (LBP) is a leading cause of global disability, yet its diagnosis remains notoriously difficult. For decades, the clinical utility of magnetic resonance imaging (MRI) has been a subject of intense debate. While MRIs often reveal structural issues like disc degeneration, these same features are frequently observed in individuals without any pain. This overlap has led to a prevailing clinical consensus that spinal MRIs are of limited diagnostic value, as they appear to lack the specificity required to distinguish between normal aging and pathological conditions.
However, this consensus often overlooks the nuances of age and spinal anatomy. Previous studies have been hampered by the subjectivity of human radiologists and the difficulty of analyzing large datasets consistently. This study addresses these limitations by utilizing an automated AI tool to perform an objective, large-scale comparison of symptomatic and asymptomatic populations, aiming to determine if structural defects correlate with pain when evaluated through a standardized lens.
Methodology
To ensure consistency across thousands of images, the researchers employed SpineNet, a verified automated machine-learning system for MRI analysis. SpineNet can analyze a spinal image in approximately five seconds—a task that typically takes a human radiologist 30 minutes—eliminating the variability inherent in subjective grading.
Study Design and Cohorts
The study focused on a large cross-sectional cohort of females aged 30–80, categorized into three groups:
- Asymptomatic Group (TwinsUK): 701 females with no reported back pain in the preceding three months.
- Symptomatic Group (OSCLMRIC): 763 females referred to secondary-care spinal triage for chronic low back pain (>3 months).
- Validation Group (Genodisc): An additional 756 symptomatic females used to confirm the trends observed in the primary symptomatic group.
Annotation and Grading
SpineNet re-annotated all sagittal T2-weighted MRIs using standardized metrics:
- Disc Degeneration: Graded on the Pfirrmann scale (1–5, where 5 represents severe degeneration).
- Binary Features: Presence or absence of disc herniation, central canal stenosis, marrow signs (Modic changes), and endplate defects.
- Spatial Stratification: Data were analyzed separately for the upper lumbar spine (L1–L3) and the lower lumbar spine (L4–S1).
Key Insights
The Importance of Spinal Level and Age
The study reveals that the correlation between structural degeneration and pain is highly dependent on the specific spinal level. In the upper lumbar spine (L1–L3), the prevalence of features like herniation and stenosis was low and statistically similar between pained and pain-free individuals. However, in the lower lumbar spine (L4–S1), the differences were stark. For subjects under 50, severe disc degeneration (Pfirrmann grades 4 and 5) was three to four times more prevalent in symptomatic patients than in the asymptomatic cohort.
Vindicating the Clinical Use of MRI
The findings challenge the idea that "everyone has degenerate discs." While mild degeneration is common with age, severe structural defects are significantly more frequent in those suffering from chronic pain. For example, in the 40–49 age bracket, approximately 30% of symptomatic lower lumbar discs showed herniation, compared to only 9% in the asymptomatic group. Similarly, marrow changes were present in 32% of symptomatic discs versus 11% of asymptomatic ones.
The "Clean MRI" Phenotype
One of the most significant findings was that 30% of younger symptomatic patients (<50 years) exhibited minimal to no disc degeneration and zero identified structural features. This suggests that chronic low back pain is not a monolithic condition. Instead, it likely comprises at least two distinct categories: one driven by mechanical/structural failure and another driven by non-structural factors, such as chemical inflammation, psychosocial influences, or augmented central pain processing.
Discussion and Clinical Impact
By providing definitive data on age-related degeneration, this study demonstrates that MRI can be a powerful tool for clinical stratification. Rather than dismissing scans as non-specific, clinicians can use them to differentiate between patients with clear structural pathology and those whose pain may stem from other pathways.
Limitations
The study's focus on a female-only cohort may limit its generalizability to men. Additionally, the symptomatic patients were drawn from secondary care, potentially representing a more severe subset of the population. The use of supine (lying down) MRIs also means that load-induced structural failures, which might only appear when a patient is standing, were not captured.
Future Directions
Automated MRI analysis provides a pathway for more personalized medicine in back pain treatment. Future research could incorporate axially loaded (standing) MRIs to further investigate the 30% of patients with "clean" scans. By stratifying patients based on objective imaging data, researchers can better target treatments and investigate the underlying mechanisms of pain in both structural and non-structural cases.
Summary updated