Neurology
COHORT
● Cohort
MRI axial-sagittal difference ratio predicts early recovery in cervical spinal cord injury without fracture
Frontiers in Medicine
Published March 30, 2026
DOI ↗
This retrospective cohort study analyzed 118 patients with cervical spinal cord injury without fracture and dislocation (CSCIwoFD) who underwent surgical treatment between August 2020 and June 2024. The objective was to examine the relationship between MRI-derived increased signal intensity (ISI) parameters and the recovery rate of Japanese Orthopaedic Association (JOA) scores to evaluate their predictive value for functional outcomes. MRI scans at admission showing high signal intensity changes on axial and sagittal planes were analyzed. Two attending physicians independently measured axial parameters (maximum signal intensity [Amax], minimum signal intensity [Amin], mean signal intensity [Am], total spinal cord area [At]) and sagittal parameters (maximum signal intensity [Smax], minimum signal intensity [Smin], mean signal intensity [Sm], length of high signal intensity [S1]). Good recovery was defined as a recovery rate (RR) > 50%, and poor recovery as RR ≤ 50%. The follow-up period ranged from 12 to 33 months. The mean preoperative JOA score was 8.58 ± 3.02 points. At 3 months postoperatively, the mean JOA score was 13.16 ± 2.22 points, with an average recovery rate of 49.42 ± 23.82%. At 2 years postoperatively, the mean JOA score was 15.11 ± 1.53 points, with an average recovery rate of 70.11 ± 28.48%. When using 'good recovery at 3 months postoperatively' as the grouping variable, the axial-sagittal difference ratio (ASDR) was significantly associated with good recovery (β = -0.5101, p = 0.036), whereas other variables were not statistically significant. When using 'good recovery at 2 years postoperatively' as the grouping variable, no variables reached statistical significance in the multivariate model, though ASmax showed a positive trend (regression coefficient 3.645, p = 0.170). Using a nonparametric test with 'good recovery at 3 months postoperatively' as the grouping variable, Ad showed a statistically significant difference (p = 0.011), another parameter showed a difference (p = 0.015), and another showed a difference (p = 0.046). Treating ASDR as a single predictive factor, ROC curve analysis yielded an AUC of 0.64. According to the Youden index, the optimal predictive threshold corresponded to a logistic regression probability of 0.157, which approximated an ASDR cutoff of 3.06. At this threshold, the model's sensitivity was 1.00, specificity was 0.32, and the Youden index was 0.32.
Imagine injuring your spinal cord in your neck without breaking any bones. It's a serious injury, and doctors use MRI scans to look for bright spots that show where the cord is damaged. But what do those bright spots actually mean for your recovery? This study looked back at 118 people who had this type of injury and had surgery. Doctors measured the size and brightness of the damage on their MRI scans and tracked how much function they regained using a standard score. On average, patients improved after surgery, with their scores getting better over two years. The researchers found that one specific number—a ratio comparing the damage seen from the side view versus the top-down view on the MRI—was linked to having a 'good recovery' at the three-month mark. When they tested this ratio as a predictor, it had moderate accuracy. It was very sensitive (catching everyone who would have a good recovery) but not very specific (it also flagged many who wouldn't). The study suggests this simple MRI measurement could be a useful early screening tool to help doctors and patients understand the likely pace of recovery in the crucial first few months after surgery.
What this means for you: A specific ratio from an MRI scan may help predict early recovery after a neck spinal cord injury.
View Original Abstract ↓
ObjectiveAlthough patients with cervical spinal cord injury without fracture and dislocation (CSCIwoFD) often exhibit increased signal intensity (ISI) on magnetic resonance imaging (MRI), the prognostic significance of these changes has not been clearly established. This study retrospectively analyzed 118 patients with confirmed CSCIwoFD to examine the relationship between MRI-derived ISI parameters and the recovery rate of Japanese Orthopaedic Association (JOA) scores, in order to evaluate their predictive value for functional outcomes and disease severity.MethodsA retrospective analysis was conducted on 118 patients diagnosed with cervical spinal cord injury without fracture and dislocation (CSCIwoFD) who underwent surgical treatment at our hospital between August 2020 and June 2024. MRI scans at admission showing high signal intensity changes on axial and sagittal planes were analyzed. Two attending physicians independently measured and recorded the following parameters according to a standardized protocol: Axial values: maximum signal intensity (Amax), minimum signal intensity (Amin), mean signal intensity (Am) and total spinal cord area (At). Sagittal values: maximum signal intensity (Smax), minimum signal intensity (Smin), mean signal intensity (Sm) and length of high signal intensity (S1). Based on previous literature, RR > 50% was defined as good recovery, and RR ≤ 50% as poor recovery. Patients were accordingly classified into good prognosis and poor prognosis groups. The correlation between the above data and postoperative JOA score and improvement rate were analyzed.ResultsAll 118 patients successfully underwent surgery. The follow-up period ranged from 12 to 33 months. The mean preoperative JOA score was 8.58 ± 3.02 points. At 3 months postoperatively, the mean JOA score was 13.16 ± 2.22 points, with an average recovery rate of 49.42 ± 23.82%. At 2 years postoperatively, the mean JOA score increased to 15.11 ± 1.53 points, with an average recovery rate of 70.11 ± 28.48%, indicating a definitive therapeutic effect of surgery. When “good recovery at 3 months postoperatively” was used as the grouping variable, the results showed that the ASDR was significantly associated with good recovery (β = −0.5101, p = 0.036), whereas the other variables were not statistically significant. When “good recovery at 2 years postoperatively” was used as the grouping variable, none of the variables reached statistical significance in the multivariate model. Among them, the ASmax had a regression coefficient of 3.645 (p = 0.170), indicating a positive trend that did not reach significance. Using a nonparametric test with “good recovery at 3 months postoperatively” as the grouping variable, the Ad showed a statistically significant difference (p = 0.011); ““showed a statistically significant difference (p = 0.015); and ““showed a statistically significant difference (p = 0.046). Treating the axial–sagittal difference ratio as a single predictive factor, ROC curve analysis yielded an AUC of 0.64. According to the Youden index, the optimal predictive threshold corresponded to a logistic regression probability of 0.157, which approximated an axial–sagittal difference ratio cutoff of 3.06. At this threshold, the model’s sensitivity was 1.00, specificity was 0.32, and the Youden index was 0.32.ConclusionThis study provides valuable insights into the utility of routine MRI sequences for evaluating CSCIwoFD. The ASDR demonstrated moderate predictive efficacy as a single-variable parameter and may serve as a sensitive screening indicator for early postoperative neurological improvement. Its incorporation into prognostic assessment could offer clinically relevant information and support decision-making. We therefore recommend considering this ratio as a core variable in the construction of early postoperative prognostic models.