Introduction
The lumbar spine (LBS) supports the upper body’s weight, accommodates the expanded abdominal cavity, and facilitates spinal movement. It also houses the spinal cord, with intervertebral foramina providing passage for spinal nerves. Physiologically, the LBS serves three primary functions: weight-bearing, neural protection, and locomotion [1]. Although motion between two adjacent vertebrae is limited, the entire spine, along with the intervertebral discs, enables substantial flexibility [1]. These physiological functions depend on the structural integrity maintained by the spinal ligamentous system [2].
The anterior ligamentous complex, including the anterior and posterior longitudinal ligaments, preserves the stability of the vertebral bodies [3]. The posterior ligamentous complex (PLC) stabilizes the posterior elements of the spine [2] and includes the ligamentum flavum, interspinous and supraspinous ligaments, zygapophyseal joint capsules [4], and the midline interlaminar ligament [4]. Damage to these structures can lead to significant spinal dysfunction.
Spinal injuries, typically resulting from high-energy trauma, are best assessed using computed tomography (CT). Magnetic resonance imaging (MRI) provides valuable information for severe cases [5], but in resource-constrained settings, MRI may be unavailable [6]. Transporting critically injured patients to facilities with MRI access can increase the risk of adverse outcomes or even mortality [7]. Therefore, it is imperative to establish a CT-based evaluation system for assessing the severity of LBS injuries.
Injuries to the LBS may also compromise the ligamentous system, potentially altering the lumbar intervertebral space (LIVS) [8]. Consequently, establishing normative reference parameters for LIVS could aid in assessing the extent of ligamentous and structural compromise. However, no such reference system currently exists. The present study introduces an LIVS Dictionary derived from measurements obtained from a cohort of 3,030 patients who underwent abdominal thin-slice enhanced CT scanning.
Materials and Methods
This study was conducted in compliance with the principles of the Declaration of Helsinki. The study protocol was approved by the Ethics Committee of the Tarim University (approval no., 2024070902).
Data collection and study population
Whole abdominal thin-slice CT scans from 3,030 patients were collected at a single center. Of these, 457 met the inclusion and exclusion criteria (Fig. 1). The required sample size was calculated using the difference in means between two groups [9], yielding an estimated cohort size of 480; therefore, 500 samples were initially selected. The patient count was verified using the “RANDBETWEEN” function in Excel, and stratified random sampling was employed during the third screening phase. Finally, 457 subjects were enrolled to develop the LIVS Dictionary (183 females and 274 males; mean age, 61 years; range, 23–91 years). To validate the LIVS Dictionary, a sample of 87 patients with five distinct types of lumbar fractures was randomly selected. Details of the LBS fractures are presented in Table 1.
CT acquisition protocol
All images were acquired using a Revolution CT scanner (GE HealthCare, Chicago, IL, USA). The region of interest extended from the apical portion of the diaphragm to the symphysis pubis. Scanning parameters were as follows: tube voltage, 120 kV; automatic current modulation (Auto-mA); collimator width, 0.625×64 mm; slice thickness, 0.625 mm; and slice spacing, 0.625 mm. After obtaining informed consent, an iodinated contrast medium (80–100 mL) was administered intravenously at a rate of 2.5–3.5 mL/sec. Additional scanning parameters were consistent with a recent study [10].
Radiological measurements
To quantify the LIVS, images were post-processed using 3D Slicer (ver. 5.6.2; https://www.slicer.org/). Measurements were performed within the Volume Rendering module, following standardized procedures illustrated in Fig. 2. In brief, seven distinct spaces were assessed in three-dimensional (3D) view using the “Line” tool in the Markups module. The same measurement protocol was applied to both the LIVS and validation groups.
To evaluate injuries related to the anterior LBS, three categories encompassing 16 parameters were utilized: (1) Lateral lumbar vertebral body space (LABS): This category included six parameters, measured sequentially from anterior to posterior, with three parameters on each side. These were designated as LABS L_1 to LABS L_3 on the left (Fig. 2A) and LABS R_1 to LABS R_3 on the right. (2) Lumbar vertebral body thickness at the mid-sagittal plane (LVBT): Five parameters labeled as LVBT_1 to LVBT_5 (Fig. 2B). (3) Endplate space at the mid-sagittal plane (EPS): This category consisted of five parameters, EPS_1 to EPS_5, measured from anterior to posterior (Fig. 2C).
Furthermore, the posterior LBS was modeled using four categories, each incorporating 10 parameters: (1) Lumbar intervertebral foramen space (LVFS): defined by four parameters, i.e., LVFS_LS (left short axis), LVFS_LL (left long axis) (Fig. 2D), LVFS_RS (right short axis), and LVFS_RL (right long axis). (2) Transverse process space (TPS): defined by four parameters, i.e., TPS_LM (medial site of the left TPS), TPS_LL (left lateral site) (Fig. 2E), TPS_RM (right medial site), and TPS_RL (right lateral site). (3) Spinous process space (SPS), characterized by two parameters: SPS_1 and SPS_2 (Fig. 2F), representing anterior-to-posterior orientation. (4) Vertebral foramen space (VFS), also defined by two parameters: VFS_L (long axis) and VFS_S (short axis) (Fig. 2G).
Statistical analysis
Line charts were used to visualize sex- and age-related variations within the LIVS Dictionary. Quantitative data for each parameter are presented as mean±standard deviation with 95% confidence intervals (CIs). Comparisons between LIVS parameters and those from patients with LBS fractures were performed using the Student t-test, with statistical significance set at p≤0.05. Sex- and fracture-related differences were analyzed using the t-test, and Cohen’s d values were calculated to estimate the magnitude of variation. The overall distribution of the 28 LIVS parameters between fracture and controls was assessed using uniform manifold approximation and projection (UMAP) analysis. Variations in these parameters among different lumbar fracture types were further evaluated using Cohen’s d. To validate the LIVS Dictionary, Elastic Net receiver operating characteristic (ROC) curve analysis with 95% CIs was performed. All statistical analyses and visualizations were performed using Python (ver. 3.11.5; https://www.python.org/).
Results
A total of 91,551 parameters were assessed, comprising 76,644 from the non-fracture group and 14,907 from the validation group. The LIVS Dictionary included 183 female subjects (mean age, 61 years; range, 24–91 years), and 274 male subjects (mean age, 61 years; range, 23–86 years). In the validation group, there were 45 female subjects (mean age, 57 years; range, 19–84 years), and 42 male subjects (mean age, 47 years; range, 19–82 years). All p-values were calculated for all subjects combined (LIVS and validation groups).
Across most parameters, male subjects exhibited significantly higher values than females (all p<0.05), except for the VFS short axis. Notably, three parameters were significantly higher in females: the short axis of the LVFS on both sides and the short axis of the VFS.
The overall LIVS Dictionary data are summarized in Table 2, while detailed parameter-specific values are shown in Fig. 3. The complete datasets are provided in Supplements 1–3.
Anterior LIVS
Fig. 4 illustrates age-related differences in LIVS between sexes. In males, a positive correlation was observed for LABS R_1 from L3 to L4 up to age 60, whereas the corresponding parameter on the left side did not show a significant trend. Across LABS parameters, females generally exhibited stronger positive or comparable correlations with age, while males tended to show negative correlations. For LVBT, both sexes demonstrated negative correlations with age; however, females showed a significant decrease in LVBT_3 and LVBT_4 beyond age 80. Additionally, in terms of EPS from L2 to S1, males experienced an increase before age 50, followed by a significant decrease, whereas the trend in females was less pronounced.
Posterior LIVS
Most parameters in the posterior LIVS demonstrated age-related reductions (Fig. 5). In the LVFS, females exhibited negative correlations, particularly in the lower LIVS segments (L2–S1), which may explain their increased susceptibility to lower limb discomfort. In contrast, males over 70 years of age showed an increase in LVFS measurements. Regarding TPS, two LIVS segments (T12 to L1 and L5 to S1) showed fluctuations with increasing age. The vertebral foramen also demonstrated an increasing trend in LIVS values from L3 to S1, whereas VFS_S remained relatively stable with age in both sexes.
Additional LIVS
Lumbarization was observed in 24 subjects (5.25%), including 10 males, while only two cases were noted in the validation group. Within the LIVS dataset, the following seven parameters showed statistically significant differences (p<0.05): LABS_L2, LABS_L3, LABS_R2, LABS_R3, EPS_4, LVFS-LS, and LVFS-RS. The overall prevalence of lumbarization was 5.25% (n=24), whereas complete sacralization, involving fusion of both the transverse process and the LBS body with the sacrum, was identified in 1.09% (n=5) of subjects.
Validation comparisons of LIVS
On further examining sex-related variations in LIVS, 27 out of the 28 parameters were found to differ significantly between females and males (Table 2, Fig. 6A). Males exhibited greater LIVS values in 25 out of 28 parameters, whereas females showed higher values in three. The most pronounced sex-based differences were observed in the anterior LIVS parameters (Fig. 6B).
Comparisons between the fracture and control groups were further stratified by sex (Fig. 7). Although significant differences were observed in most parameters between males and females (Fig. 7B, D), UMAP analysis revealed close clustering of the two sexes, indicating overall similarity in LIVS patterning. In contrast, subjects with lumbar spinal fracture exhibited fewer significantly different parameters; however, UMAP revealed distinct clustering patterns between fracture and control groups (Fig. 7A), suggesting structural deviations associated with fracture-related disability.
To identify fracture-specific influences on LIVS, variations across five fracture types were analyzed (Fig. 8). The number of parameters showing significant differences was lowest in pedicle fractures (5/28), followed by articular process (6/28), spinous process (11/28), vertebral body (12/28), and transverse process fractures (25/28). Notably, transverse process fractures resulted in increased LIVS values, while pedicle and vertebral body fractures tended to reduce them, implying their distinct mechanical impacts on spinal stability.
Validation of the LIVS Dictionary using Elastic Net ROC curve analysis demonstrated strong predictive performance (Table 3, Fig. 9). The Elastic Net model achieved an overall area under the ROC curve (AUC) of 0.90 (95% CI, 0.88–0.92) and an accuracy of 0.94. Predictive performance was slightly higher in males (AUC=0.94) than in females (AUC=0.91), confirming robust classification accuracy, particularly for males.
Discussion
The LBS plays a crucial role in supporting daily activities, and injuries to this region can lead to substantial morbidity [2]. Early diagnosis is therefore essential to minimize complications, with CT being the radiological modality of choice for initial assessment. However, even with thin-slice CT, accurate evaluation of soft-tissue damage remains challenging [11]. Although MRI has significantly advanced the detection of spinal cord and soft tissue injuries, its use is limited by longer acquisition times and higher costs. Moreover, MRI evaluation of ligamentous injuries in the spine remains technically demanding [11], and underdeveloped regions often have limited access to MRI machines [6]. These limitations highlight the need to enhance the diagnostic capability of CT for comprehensive LBS assessment.
The ossified spinous structures of the LBS serve as attachment sites for the lumbar spinal ligaments and, together with the intervertebral discs, maintain spinal stability [3,12]. These structures preserve the LIVS, which reflects the functional integrity of the ligamentous complex. Accordingly, a system capable of measuring all LIVS parameters can provide valuable insights into the structural and biomechanical status of the LBS.
Only a few studies have presented reference data for LIVS. Fyllos et al. [13] measured multiple indices in 119 adults using MRI, including several anterior LIVS parameters comparable to those analyzed in the present study. Their reported values for the L3–S1 region were consistent with our data for LVBT and EPS. However, Fyllos et al. [13] did not stratify data by sex, whereas our analyses revealed significant sex-related differences across nearly all LIVS parameters. In addition, their reference dataset did not include patients with spinal dysfunction, which limits its clinical applicability. King et al. [14] also conducted quantitative assessments of lumbar discs using MRI; while some of their findings support our EPS analysis, they evaluated additional disc-related indices. Importantly, both previous studies relied exclusively on MRI measurements, while the present work establishes normative LIVS parameters derived from CT imaging, providing broader clinical accessibility and utility.
In contrast, Karabekir et al. [15] analyzed 100 dry human vertebrae and MRI images from 21 subjects, and their findings were validated by the LIVS Dictionary (LVBT). Similarly, Korez et al. [16] measured the EPS in 20 subjects using CT images and developed comprehensive algorithms for analysis; their results were validated within a 95% CI in our LIVS Dictionary. A recent study by Zhuang et al. [17] examined the lower lumbar interlaminar space using CT-based 3D rendering to support lumbar puncture decision-making. Castro-Mateos et al. [18] also applied algorithmic analysis to thin-slice CT images from 30 patients, providing clinically relevant deflection-area data, although quantitative measurements were not performed. To date, however, no published studies have provided specific reference data for LIVS. The present study fills this gap by developing a comprehensive LIVS Dictionary.
Identifying PLC injuries on CT scans remains a major diagnostic challenge, and several previous studies have attempted to address this issue. One study reevaluated 105 patients with acute thoracic and lumbar vertebral fractures who had undergone both CT and MRI, reporting a diagnostic sensitivity of 82%, based on one positive finding among seven criteria, which were validated by three experienced radiologists [19]. Vaccaro et al. [20] proposed that the SPS could serve as an indicator of PLC injury, with a spacing of 7 mm or more suggesting potential disruption. Moliere et al. [21] introduced a method based on changes in the paraspinal fat pad to assess PLC injuries on CT. Kwon et al. [22] developed another diagnostic tool, a modified SPS-based approach comparable to ours, for detecting PLC injuries. Chen et al. [23] further expanded on this concept by applying multiple-angle CT assessments to evaluate PLC integrity. However, none of these studies reported sex-specific differences. Our results demonstrate that most LIVS parameters differ significantly between females and males, and that these variations also depend on the specific LIVS level examined. Importantly, our work provides a comprehensive dataset addressing both sex-based and regional LIVS differences.
In addition to sex-related and LIVS-specific variations, age is a pivotal factor influencing LBS spacing. De Schepper et al. [24] conducted a large cohort study employing a four-score method to evaluate the degree of lumbar disc degeneration and reported a strong correlation between low back pain and EPS, excluding the LIVS between L5 and S1. Tanaka et al. [25] introduced the Thompson grading system, which has since been widely adopted for assessing lumbar disc degeneration [26]. Although most research on LBS degeneration has focused on intervertebral disc morphology and signal changes, Kalichman et al. [27] emphasized the association between spinal stenosis and degenerative low back pain. In our study of 457 unselected subjects, we observed a prevalence of osteoporosis that appeared unrelated to low back pain. This finding likely reflects generalized, age-related degenerative changes rather than a specific pain etiology. Collectively, these results underscore the multifactorial nature of LBS degeneration and establish a foundation for future investigations exploring age-, sex-, and structure-dependent variations in LIVS.
Both lumbarization and sacralization were observed, with prevalence rates of 5.25% (n=24) and 1.09% (n=5), respectively. These findings are consistent with those reported by Mahato [28,29], who reported sacralization and lumbarization prevalences of 1.2% [28] and 3.9% [29], respectively, although the lumbarization rate in our cohort was slightly higher.
Comparison of Cohen’s d values revealed that the 10 most significant differences between females and males were all located in the anterior LIVS parameters. These variations may serve as potential diagnostic or predictive indicators [30–33]. Our results demonstrated that females differed significantly from males in most of the 28 LIVS parameters, with males generally exhibiting higher values, though not universally so. This finding suggests that analyzing female and male data separately may enhance the performance of predictive models. Given the pronounced sex-related differences, these results may also have implications in forensic medicine, particularly for sex determination based on skeletal remains [34].
The fracture-type analysis provided additional insights into the biomechanical consequences of different injury patterns. Pedicle and articular process fractures had minimal impact on LIVS, influencing only five to six parameters. In contrast, vertebral body and spinous process fractures showed similar tendencies to alter LIVS, particularly in parameters related to the spinal body and vertebral foramen. Transverse process fractures produced the most extensive changes, resulting in increased values across most LIVS parameters. Validation of the LIVS Dictionary using AUC analysis demonstrated strong predictive robustness, with AUC values exceeding 0.9. These findings indicate that the current LIVS Dictionary represents a reliable quantitative reference and provides a foundation for the development of advanced diagnostic and predictive models in LBS assessment.
The present study has several limitations. First, although a substantial amount of LVIS-related data has been provided, the full potential of these data is yet to be realized. Second, while the anterior LIVS Dictionary demonstrated robust predictive performance for injury assessment, the posterior LIVS Dictionary showed comparatively lower accuracy. Because lumbar spinal injuries often result from complex biomechanical mechanisms, further analyses incorporating the deep muscular system and multiple fracture types may help clarify the underlying pathophysiology and improve the precision of the LIVS Dictionary. Accordingly, future modifications are planned to enhance the predictive sensitivity of the posterior LIVS Dictionary. Third, the data are derived from a single center. Expanding the dataset to include participants from multiple centers and diverse racial and demographic backgrounds would greatly enhance the generalizability and robustness of the LIVS Dictionary.
Conclusions
This study developed a LIVS Dictionary using CT-based 3D rendering from 457 asymptomatic LBS subjects. The dictionary quantifies normalized intervertebral spacing from T12 to S1 and incorporates 28 distinct parameters per LIVS segment. Males generally exhibited larger LIVS values, particularly in anterior regions, highlighting the need for sex-specific reference models. Age- and fracture-related analyses revealed distinct mechanical and morphological patterns, with transverse process fractures producing the most extensive LIVS alterations. Validation using the Elastic Net model demonstrated strong predictive accuracy (AUC >0.9), confirming the robustness of the LIVS Dictionary. Overall, this framework provides a quantitative reference for evaluating LBS integrity and provides a foundation for future applications in clinical diagnosis, biomechanical modeling, and forensic identification.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement
