TL;DR
This study demonstrates that adding a coronal Short Tau Inversion Recovery (STIR) sequence to standard lumbar spine MRI protocols significantly improves the detection of conjoined lumbosacral nerve roots (CLNRs). This enhancement leads to higher sensitivity, better inter-reader agreement, and a more accurate diagnosis, which is crucial for surgical planning and preventing iatrogenic nerve injury.
What Problem Does This Paper Address?
Conjoined lumbosacral nerve roots (CLNRs) are a relatively common anatomical variation, yet they are often overlooked on standard lumbar spine MRI scans. This oversight is problematic because CLNRs can be mistaken for herniated discs during spinal surgery, potentially leading to nerve injury and failed back surgery syndrome. While CT myelography and some MRI sequences have shown promise in detecting CLNRs, there's a need for a standardized, non-invasive method that reliably identifies these anomalies. This paper investigates whether incorporating a coronal STIR sequence into a routine lumbar MRI protocol can enhance CLNR detection accuracy compared to standard protocols.
Methods
This retrospective study involved two radiologists assessing MRI scans from 751 patients. The assessment was conducted in three readouts:
- Consensus Readout (Reference Standard): Two experienced radiologists, using the expanded protocol (including coronal STIR), established a consensus diagnosis for CLNR. In ambiguous cases, a third senior radiologist was consulted. This consensus served as the ground truth.
- Standard Protocol Readout: The same two radiologists independently reviewed the scans again, this time using only the standard lumbar MRI protocol (without coronal STIR).
- Expanded Protocol Readout: After a 6-week interval to prevent recall bias, the radiologists re-evaluated the scans using the expanded protocol (including coronal STIR).
The study also assessed lumbosacral transitional vertebrae (LSTV) during the third readout. Diagnostic metrics like sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were calculated for both the standard and expanded protocols. Inter-reader agreement was measured using Cohen's kappa statistic.
Datasets
The study utilized retrospective MRI data from 751 patients who underwent lumbar spine MRI within a 3-year period. These patients were selected from an initial pool of 976 potentially eligible patients based on specific inclusion and exclusion criteria (e.g., age > 18, no severe scoliosis, no prior lumbar surgery). The study did not introduce a new benchmark dataset.
Key Insights and Ideas
- Coronal STIR is a Game Changer for CLNR Detection: The most significant finding is the dramatic improvement in sensitivity and overall detection rate when the coronal STIR sequence is included. This suggests that STIR's ability to highlight fluid and edema, which can be associated with nerve root pathology or variations, is particularly effective for visualizing CLNRs.
- Improved Inter-Reader Agreement: The addition of coronal STIR not only helped individual readers but also significantly improved the agreement between them. This indicates that the sequence provides clearer, more consistent visual cues for CLNR identification, reducing subjective interpretation differences.
- LSTV as a Potential Red Flag: The study found a significant association between CLNR and lumbosacral transitional vertebrae (LSTV). This suggests that the presence of an LSTV on an MRI might warrant a closer look for CLNR, as these anatomical variations can co-occur.
- Axial T2-weighted is Good, but Coronal STIR is Better: While axial T2-weighted sequences were the primary means of initial detection in the standard protocol, the coronal STIR sequence proved to be the most helpful for initial detection in the expanded protocol, highlighting its crucial role.
Key Results
- Sensitivity Boost: Sensitivity for CLNR detection increased from 34.2% without coronal STIR to 81.6% with it.
- Accuracy Improvement: Overall accuracy rose from 96.3% to 98.9%.
- Specificity and PPV: Specificity slightly improved from 99.6% to 99.9%, and PPV increased from 81.3% to 96.9%.
- Inter-Reader Agreement: Agreement improved from moderate (κ=0.592) without coronal STIR to almost perfect (κ=0.915) with it.
- CLNR Prevalence: A CLNR prevalence of 5.1% (38 out of 751 patients) was found in the consensus readout.
- LSTV Association: LSTV prevalence was 13.3% in patients without CLNR and 26.3% in patients with CLNR, showing a significant correlation.
Limitations and Future Work
- Reference Standard: The study used MRI-based consensus as the reference standard, rather than surgical correlation, which is acknowledged as a limitation. However, surgical findings are not always more reliable than imaging for CLNR detection.
- Slice Thickness: The coronal STIR sequence used a 4 mm slice thickness, which is at the upper limit of recommendations. Thinner slices (≤ 3 mm) might further improve detection by reducing partial volume effects.
- STIR Contrast: The coronal sequence was always acquired with STIR contrast. Future studies could compare coronal T2-weighted non-fat-suppressed imaging with coronal STIR for CLNR detection.
- Exclusion of Preoperative Patients: The exclusion of patients with prior surgery (e.g., spondylodesis) introduces a selection bias, as metal artifacts can affect MRI quality.
- LSTV Classification: The classification of LSTV was based on the assumption of 12 thoracic ribs, which might affect accuracy.
- Clinical Implications: The study did not delve into the clinical implications of the additionally detected CLNRs.
Future work could involve comparative studies with surgical correlation, investigations using thinner slice thicknesses, and exploring the clinical impact of CLNR detection with coronal STIR in follow-up studies.
Summary updated