Association between 3 D CT-based volumetric fat infiltration of lumbar paraspinal muscles and bone mineral density: a clinical study

Abstract

Objective

This study investigated the association between the fat infiltration rate (FIR) of lumbar paraspinal muscles—specifically the multifidus, erector spinae, and psoas major—and bone mineral density (BMD) in postmenopausal women. By examining muscular and segmental (L2–L4) adiposity variations, we aimed to identify imaging biomarkers for early osteoporosis detection and to inform integrated “bone-muscle” therapeutic strategies.

Methods

A retrospective analysis was performed on 120 postmenopausal women who underwent dual-energy X-ray absorptiometry (DXA) and lumbar computed tomography (CT) between 2024 and 2025. Participants were categorized into three cohorts—normal bone mass, osteopenia, and osteoporosis—based on DXA T-scores. Demographic data, including BMI, were recorded and subsequently utilized as a covariate in the regression models to ensure robust statistical adjustment. The L2–L4 paraspinal musculature was automatically segmented in 3D using 3D Slicer software with the MuscleMap plug-in to determine the FIR for each muscle group. Data normality was assessed using Shapiro-Wilk tests, while inter-group differences were compared via one-way ANOVA or Kruskal-Wallis H tests. Correlations were quantified using Pearson or Spearman analyses.

Results

Age was comparable across groups ( P = 0. 160 ), but BMI and average BMD differed significantly. Multifidus FIR showed a significant inverse correlation with BMD ( P < 0. 001 ). After adjusting for BMI, multifidus fat infiltration remained an independent predictor of BMD. At the segmental level, the L3 multifidus displayed the greatest inter-group variance and the strongest negative correlation with BMD ( r = -0. 689, P < 0. 001 ).

Conclusions

Lumbar paraspinal muscle adiposity, particularly in the L3 multifidus, is inversely associated with BMD in postmenopausal women. Consequently, L3 multifidus FIR is a statistically significant auxiliary imaging metric for proactive osteoporosis screening and the assessment of osteosarcopenia risk.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12880-026-02556-0.

Introduction

Osteoporosis is a systemic skeletal disorder characterized by progressive bone mass reduction and microarchitectural deterioration, which collectively increase bone fragility and fracture susceptibility [1]. Given the rapidly aging global population, postmenopausal osteoporosis has become a critical public health challenge worldwide. Evidence indicates that the postmenopausal period is marked by rapid mineral density loss—with annual declines of up to 2. 0% in the lumbar spine and 1. 4% in the hip—frequently leading to vertebral compression fractures [2]. Beyond the immediate physical impact, such complications diminish patient quality of life and impose a substantial economic burden on healthcare systems [3]. Although dual-energy X-ray absorptiometry (DXA) remains the gold standard for diagnosis via bone mineral density (BMD) measurement, its clinical application faces limitations [4]. For instance, osteophytosis may artificially elevate BMD readings, thereby masking significant bone loss; furthermore, fractures frequently occur in patients with clinically “normal” bone mass. This clinical reality underscores the necessity of identifying synergistic risk factors beyond BMD alone.

Recently, the “bone–muscle axis” has gained prominence in clinical discourse [5–6], highlighting the intrinsic embryonic, biomechanical, and endocrine interconnections between these tissues. In postmenopausal women, aging and declining estrogen levels trigger simultaneous bone resorption and degenerative muscular changes. A primary feature of this degradation is the accumulation of intramuscular fat, a hallmark of sarcopenia. Previous research has established that such fat infiltration is a primary determinant of muscle weakness and correlates with various spinal pathologies [7–8].

While researchers have investigated the link between paraspinal muscle adiposity and BMD, significant methodological hurdles persist. Most existing research relies on single 2 D axial slices—typically at the L3 or L4 level—as a proxy for entire muscle status [9–10]. This approach is inherently limited because paraspinal groups, such as the multifidus, erector spinae, and psoas major, follow longitudinal trajectories and exhibit distinct segmental variations in morphology and fat distribution. Consequently, single-level assessments rarely capture the comprehensive anatomical profile of the lumbar segment. Furthermore, conventional manual or semi-automatic segmentation methods are labor-intensive and susceptible to measurement error.

Advances in medical imaging, specifically the open-source platform 3 D Slicer, have introduced tools for precise muscle analysis through automated, multi-dimensional quantification [11]. By employing volumetric modeling, researchers can now isolate total muscle and fat volumes, facilitating an objective calculation of the fat infiltration rate (FIR). This approach provides superior reproducibility and anatomical accuracy compared to conventional single-level area measurements.

Accordingly, the present study utilized 3 D segmentation and fat quantification for the psoas major, erector spinae, and multifidus muscles across the L2–L4 segments. This study aimed to clarify the correlation between paraspinal muscle fat infiltration and bone mass categories, thereby providing imaging evidence to support early osteoporosis screening and integrated “bone-muscle” therapeutic strategies.

Materials and methods

Study population

A retrospective analysis was performed on the clinical data of 120 postmenopausal women who were admitted to our hospital and underwent both DXA and lumbar spine CT examinations between January 2024 and December 2025, all enrolled patients underwent both examinations within a maximum interval of 2 weeks, ensuring the cross-sectional validity of the muscular and bone density data. According to the World Health Organization (WHO) diagnostic criteria [12–13], participants were classified into three groups based on lumbar T-scores derived from DXA (compared with gender-matched healthy young adults): normal bone mass group (T-score ≥ -1. 0), osteopenia group (-2. 5 < T-score < -1. 0), and osteoporosis group (T-score ≤ -2. 5), with 40 cases in each group.

Inclusion criteria

① Female gender; ② Age 55–75 years (postmenopausal period); ③ No lumbar deformity, scoliosis, or history of lumbar surgery; ④ Excellent quality of DXA and CT imaging data.

Exclusion criteria

① Age ≤ 55 years or ≥ 75 years; ② Patients with a history of lumbar spine surgery, lumbar deformity, or scoliosis; ③ Patients with incorrect body positioning during imaging examination; ④ Poor quality of imaging data affecting measurement results;⑤ Patients receiving long-term corticosteroid therapy, hormone replacement therapy (HRT)༛⑥Patients with diagnosed severe systemic metabolic/endocrine disorders such as uncontrolled Diabetes mellitus;⑦ Patients with significant motor impairment or physical disabilities that severely restrict mobility.

Data Collection

DXA data acquisition

DXA data were acquired using a GE Prodigy Advance X-ray bone densitometer. The instrument automatically calculated and recorded the BMD values of the L2-L4 vertebrae and the average BMD T-scores. All data were verified by two professional technicians to ensure accuracy.

CT data acquisition

Lumbar spine CT data were acquired using a 64-slice Siemens Definition AS 128 spiral CT scanner. The scanning range was from L1 to L5, with the following parameters: tube voltage of 120 kV, tube current of 250 mA, pitch of 1, slice thickness of 1 mm, and interslice spacing of 1 mm. The reconstructed bone window data were imported into the open-source software 3 D Slicer (https://www.slicer.org/) [14] for subsequent segmentation and analysis.

Muscle segmentation and fat quantification

The psoas major, erector spinae, and multifidus muscles at the L2-L4 segments were automatically segmented using the MuscleMap plug-in of the 3 D Slicer software [15] (Figure 1. A). To eliminate potential observer bias, all image segmentation and subsequent measurements were performed by two researchers who were blinded to the patients’ clinical data, including their DXA T-scores and group assignments. Multi-planar interactive adjustments were performed to align the sagittal axis with the vertebral long axis and the coronal axis with the lamina plane. The scissors tool in the Segment Editor module was used to crop the target muscle tissue horizontally (Figure 1. B), and the volume data of each muscle group at the L2-L4 segments were obtained (Figure 1. C-D).

Fig. 1.

A: Automated three-dimensional segmentation of lumbar paraspinal muscles (multifidus, erector spinae, and psoas major) at L2–L4 levels using the MuscleMap plug-in in 3 D Slicer. B: Multi-planar reformatting and segmental definition with interactive adjustment to align the sagittal axis with the vertebral long axis and the coronal axis with the lamina plane. C: Volumetric extraction of a single-level lumbar paraspinal muscle. D: Three-dimensional visualization of a single-level lumbar paraspinal muscle

Using the automated MuscleMap plug-in, the average processing time per patient—encompassing muscle segmentation and fat quantification—was approximately 5.2 ± 0.6 min. This represents a significant reduction in labor compared to traditional manual segmentation methods, which often require 30–45 min per case.

Threshold segmentation was performed using the Segment Editor module of 3 D Slicer. The CT value range for adipose tissue was set to -120 to -40 Hounsfield units (HU) to exclude muscle parenchyma and bone components, thereby obtaining pure fat volume data within the volume data (Figure 2. A-B). Subsequently, the single-segment paraspinal muscle volume data and pure fat volume data were subjected to the intersect logical operation using the Logical Operators of the Segment Editor module to obtain intramuscular fat voxels (Figure 2. C-D). The exact volume of fat components in this region was then automatically calculated and output using the Segment Statistics module. The paraspinal muscle fat infiltration rate (%) was calculated as follows: Intramuscular fat volume / Total volume of paraspinal muscles in the same segment × 100%.To assess workflow reproducibility, intra- and inter-observer reliability were evaluated in 30 randomly selected patients using the Intraclass Correlation Coefficient (ICC), where an ICC > 0.90 denoted excellent agreement denoted.

Fig. 2.

A: Preliminary labeling of muscle and fat on single-slice CT images. B: Three‑dimensional extraction of total volumetric fat voxels. C: Precise filtering of intramuscular fat. D: Three-dimensional fused visualization of intramuscular fat and muscle tissue

Statistical methods

Statistical analyses were performed using SPSS software (version 26. 0). The normality of continuous variables was assessed using the Shapiro-Wilk test. Normally distributed data were expressed as mean ± standard deviation; inter-group differences were analyzed via one-way ANOVA, followed by LSD-t tests for pairwise comparisons. Conversely, non-normally distributed variables were expressed as medians (interquartile ranges) and compared using the Kruskal-Wallis H test. Categorical variables were presented as frequencies (percentages) and analyzed using the chi-square test.

Correlations between bone mineral density (BMD) and paraspinal muscle fat infiltration rates (FIR) were evaluated using Pearson or Spearman analysis. Multivariable linear regression models were subsequently constructed to identify independent predictors of BMD. To provide a transparent and comprehensive evaluation of the model, we reported unstandardized regression coefficients (B), standard errors (SE), standardized regression coefficients (β), t-statistics, and corresponding P-values, alongside 95% confidence intervals (95% CI). Overall model fit was assessed using the F-statistic, the coefficient of determination (R²), the adjusted R², and the Durbin-Watson (DW) statistic to detect residual autocorrelation. Multicollinearity among independent variables was assessed using the Variance Inflation Factor (VIF), where VIF > 10 indicated significant collinearity. For all analyses, statistical significance was defined as P < 0.05.

To rigorously evaluate the diagnostic efficacy of the L3 multifidus fat infiltration rate (FIR) as a potential imaging biomarker for postmenopausal osteoporosis, a Receiver Operating Characteristic (ROC) curve analysis was performed. The Area Under the Curve (AUC) and its 95% confidence interval (CI) were calculated to quantify its discriminative ability.

Results

Baseline characteristics and paraspinal muscle metrics

This study included 120 postmenopausal women, divided equally into normal bone mass, osteopenia, and osteoporosis groups ( n = 40 per group). Age was comparable across the three groups ( P = 0. 160 ), facilitating a robust comparison of other physiological markers. In contrast, body mass index (BMI) and mean bone mineral density (BMD) differed significantly among the cohorts (all P < 0. 05 ). Specifically, BMI was significantly higher in the normal bone mass group ( 27. 09 ± 3. 14 kg/m² )than in the osteopenia ( 23. 80 ± 4. 06 kg/m² ) and osteoporosis ( 22. 73 ± 3. 26 kg/m² ) groups ( P < 0. 001 ). Mean BMD exhibited a clear, stepwise decline as skeletal health decreased, with significant differences identified in all pairwise comparisons ( P < 0. 05 ).

Regarding overall paraspinal muscle health, the multifidus fat infiltration rate (FIR) showed significant inter-group heterogeneity ( P < 0. 001 ). FIR was highest in the osteoporosis cohort, followed by the osteopenia and normal bone mass groups. For the erector spinae, FIR was significantly higher in the osteoporosis group compared to the other two cohorts ( P = 0. 022 ). In contrast, the psoas major FIR showed no significant variance across the groups ( P = 0. 071 ) (Table 1). The 3D automated segmentation demonstrated excellent reliability, with intra- and inter-observer ICC values ranging from 0.91 to 0.96 for the fat infiltration rates of all muscle groups, confirming the high stability of the quantification platform.

Table 1.

Comparison of baseline characteristics and overall paraspinal muscle indicators among three groups

Item	Normal bone mass (n = 40)	Osteopenia(n = 40)	Osteoporosis(n = 40)	F/H value	P-value
Age (years)	63. 33 ± 5. 76	65. 25 ± 5. 49	65. 52 ± 5. 41	1. 86	0. 160
BMI (kg/m²)	27. 09 ± 3. 14 ab	23. 80 ± 4. 06 a	22. 73 ± 3. 26 b	16. 70	< 0. 001
Average BMD (g/cm²)	1. 24 ± 0. 14ab	0. 96 ± 0. 04 ac	0. 80 ± 0. 10 bc	195. 50	< 0. 001
Fat infiltration rate of multifidus (%)	10. 06 ± 5. 47a	9. 50 ± 8. 62 c	13. 34 ± 8. 22 ac	9. 04	< 0. 001
Fat infiltration rate of erector spinae (%)	10. 07 ± 4. 88a	9. 84 ± 6. 84 c	11. 89 ± 6. 84 ac	3. 860	0. 022
Fat infiltration rate of psoas major (%)	6. 19 ± 4. 12	5. 53 ± 3. 52	5. 15 ± 2. 85	2. 662	0. 071

a: Significant difference between normal bone mass and osteopenia groups. b:Significant difference between normal bone mass and osteoporosis groups.c: Significant difference between osteopenia and osteoporosis groups

Segment-specific characteristics of muscle adiposity

Segment-specific analyses at the L2–L4 levels indicated that multifidus FIR was significantly associated with BMD at all measured segments ( all P < 0. 05 ). The L3 segment exhibited the most pronounced inter-group variance ( P = 0. 018 ); at this level, FIR peaked in the osteoporosis group and was lowest in the normal bone mass group. At the L2 level, multifidus FIR was significantly higher in the osteoporosis group than in the osteopenia group ( P = 0. 044). However, differences at the L4 level did not reach statistical significance ( P = 0. 054 ). In contrast to the multifidus, the erector spinae and psoas major muscles displayed no significant inter-group differences in FIR across any individual lumbar levels (all P > 0. 05 ) (Table 2).

Table 2.

Intergroup comparison of muscle fat infiltration rate at different lumbar segments

Segment	Muscle	Normal	Osteopenia	Osteoporosis	P-value
	Multifidus	10. 02 ± 5. 23	8. 56 ± 5. 64 b	11. 91 ± 6. 82 b	0. 044
L2	Erector spinae	8. 53 ± 4. 41	8. 42 ± 6. 37	9. 68 ± 5. 54	0. 525
	Psoas major	6. 38 ± 4. 28	6. 52 ± 4. 23	5. 95 ± 3. 30	0. 800
	Multifidus	9. 52 ± 5. 82 a	8. 50 ± 5. 61 b	12. 64 ± 8. 28 ab	0. 018
L3	Erector spinae	9. 72 ± 4. 33	9. 54 ± 6. 32	11. 45 ± 6. 70	0. 279
	Psoas major	6. 52 ± 4. 63	5. 35 ± 3. 48	4. 76 ± 2. 57	0. 095
	Multifidus	10. 64 ± 5. 42	11. 43 ± 12. 55	15. 48 ± 9. 16	0. 054
L4	Erector spinae	11. 96 ± 5. 30	11. 54 ± 7. 55	14. 52 ± 7. 41	0. 111
	Psoas major	5. 67 ± 3. 41	4. 71 ± 2. 36	4. 74 ± 2. 52	0. 222

a: Significant difference in muscle fat infiltration rate between the normal bone mass group and the osteoporosis group. b༚Significant difference in muscle fat infiltration rate between the osteopenia group and the osteoporosis group

Correlation and regression analyses of BMD and muscular FIR

Correlation analysis confirmed that multifidus FIR was significantly and inversely associated with BMD ( r = -0. 623, P < 0. 001 ). The L3 segment demonstrated the strongest negative correlation ( r = -0. 689, P < 0. 001 ; Fig. 3). Erector spinae FIR also correlated negatively with BMD ( r = -0. 412, P < 0. 001 ), though this correlation was substantially weaker than that of the multifidus. Conversely, psoas major FIR showed no significant correlation with BMD ( r = -0. 176, P = 0. 058 ). To determine whether paraspinal muscle adiposity independently predicts bone mass, a multivariable linear regression model was constructed with average lumbar BMD as the dependent variable, adjusting for age and BMI. The regression model proved to be statistically significant (F(3, 116) = 12.077, P < 0.001), explaining approximately 21.8% of the variance in BMD (Adjusted R² = 0.218). Crucially, after controlling for age and BMI, L3 multifidus FIR remained a significant independent negative predictor of lumbar BMD (B = -0.006, SE = 0.003, Standardized (β)= -0.209, t = -2.532, P = 0.013, 95\% CI: [-0.012, -0.000]). Among the covariates, BMI exhibited a strong positive association with BMD (β = 0.431, t = 5.287, P < 0.001), whereas age did not reach statistical significance (β = -0.125, P = 0.130). The model variance inflation factors (VIF) were consistently below 3, confirming the absence of significant multicollinearity(Table 3).

Fig. 3.

A. Comparison of fat infiltration rates among three bone mass groups; the multifidus shows a significant increase as BMD declines ( *P < 0. 001). (B) Heatmap showing the correlation between muscle fat infiltration (L2–L4) and BMD; L3 multifidus exhibits the strongest correlation ( r = -0. 689 ). (C) Linear regression between overall multifidus fat infiltration and average BMD ( r = -0. 623, P < 0. 001 )

Table 3.

Multivariable linear regression analysis for factors associated with BMD

Variable	Unstandardized(B)	StandardError(SE)	Standardized(β)	t-value	P-value	95%CI
Constant	0.804	0.226	-	3.553	0.001	[0.356, 1.252]
Age	-0.005	0.003	-0.125	-1.525	0.13	[-0.011, 0.001]
BMI	0.023	0.004	0.431	5.287	0.000	[0.015, 0.031]
L3 Multi Fat	-0.006	0.003	-0.209	-2.532	0.013	[-0.012, -0.000]

Note: Model fit: R² = 0.238; Adjusted R² = 0.218; F(3, 116) = 12.077; P < 0.001; Durbin-Watson statistic = 0.888

Diagnostic performance of L3 Multifidus FIR for osteoporosis screening ROC curve analysis

demonstrated that the L3 multifidus FIR possessed statistically significant diagnostic value in distinguishing postmenopausal women with osteoporosis from non-osteoporotic individuals (AUC = 0.648, 95% CI: 0.541–0.756, P = 0.0068; Fig. 4). Based on the maximum Youden’s index, the optimal cut-off value was determined to be 8.25%, yielding a sensitivity of 72.5% and a specificity of 55.0%.

Fig. 4.

Receiver operating characteristic (ROC) curve analysis

Discussion

A key methodological advantage of this study is the implementation of 3 D automated segmentation using 3 D Slicer software integrated with the MuscleMap plug-in. Compared with conventional manual tracing or single-slice axial metrics, this volumetric approach provides a more comprehensive and precise characterization of paraspinal muscle adiposity across the entire lumbar region. By minimizing manual measurement errors and addressing segmental variability, this 3 D-derived fat infiltration rate (FIR) achieves greater anatomical representativeness and reproducibility. Furthermore, the quantitative evaluation of adipose content was refined by restricting CT attenuation values to a specific range ( -120 to -40 HU) to extract pure intramuscular fat components. This methodology establishes a more objective, reliable foundation for investigating the association between muscle fat infiltration and bone mineral density (BMD).

The pathogenesis of postmenopausal osteoporosis is driven primarily by an imbalance in bone metabolism secondary to estrogen depletion. This metabolic disruption is closely linked to paraspinal fat infiltration, reflecting synergistic components of postmenopausal osteomyoaxis dysfunction. Estrogen deficiency simultaneously stimulates osteoclast activity and suppresses osteoblasts, leading to rapid bone loss [16]. It also induces degenerative muscle changes, including atrophy and ectopic fat accumulation [17]. Moreover, intramuscular adipose tissue acts as an endocrine source of pro-inflammatory cytokines, such as interleukin-6 (IL-6) [18]. This pathway potentially accelerates muscle degradation and bone resorption while impairing bone formation, thereby driving a pathological “muscle degeneration–bone loss” cycle.

Our findings indicate a significant correlation between multifidus FIR and BMD, with the L3 segment exhibiting the strongest association. This segmental sensitivity is likely attributable to the anatomical proximity of the multifidus to the lumbar vertebrae and its primary role in spinal stabilization.Our multivariable linear regression model successfully quantified this relationship, demonstrating that for every 1% increase in the L3 multifidus fat infiltration rate, the average lumbar BMD decreases by 0.006 g/cm² (B = -0.006, P = 0.013), independent of the compounding influences of chronological aging and systemic BMI. Although the overall model explains a moderate proportion of bone density variance (Adjusted R² = 0.218), which implies that bone loss is a highly multifactorial process driven by complex endocrine and genetic networks, the adiposity status of the deep stabilizing musculature at the lordotic apex (L3) remains a clinically meaningful and distinct marker of axial skeletal degradation. Our newly added ROC analysis revealed a significant, albeit moderate, diagnostic value of the L3 multifidus FIR (AUC = 0.648, P = 0.0068). This moderate discriminative performance is highly consistent with the multi-factorial pathogenesis of postmenopausal osteoporosis. Skeletal degradation is deeply modulated by systemic endocrine pathways, genetic predisposition, and nutritional status, meaning no single localized musculoskeletal metric can dictate the entirety of bone density variations. Nevertheless, as an independent structural parameter embedded in the ‘bone-muscle’ axis, the L3 multifidus FIR offers valuable complementary screening insights, particularly as a localized imaging marker.Conversely, the psoas major functions as a long-axis muscle and demonstrated a weaker association with localized lumbar metabolic shifts. This may explain the absence of significant inter-group FIR variance among the bone mass cohorts. This observation aligns with previous research; for instance, Carvalho et al. identified multifidus fat infiltration as a predictor of suboptimal outcomes following lumbar surgery [19]. Similarly, Maharjan et al. reported that preserving the multifidus cross-sectional area via minimally invasive techniques correlates with improved functional scores [20].

In this study, the L3 multifidus FIR showed the most marked differentiation among groups, consistent with previous literature. Fortin et al. [21] suggested that the L3–L4 level is highly representative of paraspinal health, while Hu et al. [22]. established the high reliability of measurements at these specific levels. Biomechanically, the L3 vertebra is positioned at the apex of lumbar lordosis, serving as a mechanical nexus for trunk loading. This anatomical position potentially renders the adjacent multifidus more sensitive to the metabolic and biomechanical stressors of estrogen deficiency.

Nevertheless, several limitations warrant consideration. Due to the cross-sectional design, this study cannot establish a causal relationship between muscle adiposity and bone loss. The relatively small sample size from a single institution may introduce selection bias, necessitating validation in future multi-center studies. Furthermore, although we established strict exclusion criteria for factors affecting the bone-muscle axis, such as corticosteroid use and hormone replacement therapy, our study did not quantitatively assess participants’ physical activity levels. Physical activity is a crucial modulator of both muscle and bone health; its omission in our retrospective analysis represents a potential limitation. Additionally, the Durbin-Watson value of our regression model (DW = 0.888) suggested potential residual autocorrelation. This might be attributed to the inherent nature of a retrospective study, which could not fully incorporate certain unmeasured confounding lifestyle or behavioral variables—such as quantified physical activity levels or longitudinal dietary habits—that significantly modulate both skeletal and muscular metabolism.Future prospective studies incorporating standardized physical activity questionnaires or objective monitoring (e.g., accelerometry) are warranted to further refine our understanding of these interactions. Additionally, CT-based fat quantification can be influenced by scanning parameters or individual anatomical variance. Furthermore, the absence of muscle strength data or serum biomarkers precludes a comprehensive evaluation of correlations with clinical symptoms, such as low back pain. Future research will focus on large-scale prospective cohort studies and multi-omics analyses to further elucidate the mechanisms underlying the association between paraspinal muscles and BMD. Integrating clinical strength metrics and hormonal profiles will be essential to support integrated “bone-muscle” strategies for osteoporosis management.

Conclusions

This study demonstrates a significant inverse relationship between lumbar paraspinal muscle fat infiltration and bone mineral density (BMD) in postmenopausal women, characterized by muscle group specificity and segmental heterogeneity. Specifically, the multifidus—particularly at the L3 level—emerged as the most sensitive metric for tracking variations in bone mass. Consequently, the L3 multifidus fat infiltration rate (FIR) is a promising imaging biomarker for the proactive screening of postmenopausal osteoporosis and the assessment of osteosarcopenia risk. Furthermore, automated 3 D segmentation using 3 D Slicer provides an objective, reliable methodology for quantifying muscular adiposity. This technological advancement is expected to facilitate the clinical implementation of integrated “bone-muscle” strategies for the prevention and management of osteoporosis in this population.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

RH.F and D.W co-authored this article. BB.Y collected relevant data, BQ.P prepared Figs. 1, 2 and 3, and LS.Y performed statistical analysis on the data. All authors have reviewed the final manuscript.

Funding

This study was supported by the Ningbo Clinical Research Center for Sports and Rehabilitation (2024 L004),the 2026 Zhejiang Provincial “Sword Pointing & Leading Wild Goose + X” Science and Technology Program (2026 C02 A1261(SD2))and2024 Yinzhou District Science and Technology Project (2024AS035).

Data availability

Data Availability Statement: The clinical demographics, dual-energy X-ray absorptiometry (DXA) bone mineral density values, and 3D CT-based paraspinal muscle fat infiltration metrics used to support the findings of this study are included within the article (Tables 1, 2 and 3). The raw retrospective imaging datasets (CT scans and DXA files) analyzed during the current study are not publicly deposited to protect patient confidentiality in accordance with the Ethics Committee of Ningbo No. 6 Hospital. However, anonymized statistical datasets are available from the corresponding author, Dr. Dan Wang, at 18906627553@163.com on reasonable request.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the principles of the Declaration of Helsinki and approved by the Ethics Committee of Ningbo No. 6 Hospital (Approval No. 2025–098). Due to the retrospective nature of this study and the use of anonymized clinical imaging data, the requirement for informed consent was waived by the Ethics Committee of Ningbo No. 6 Hospital. All patient data were fully anonymized prior to analysis to ensure participant privacy.

Consent for publication

Not applicable. This study is a retrospective analysis utilizing fully anonymized clinical imaging data and demographic characteristics. No individual participant’s hidden identity, personal information, or identifying images are presented in this manuscript. Therefore, the requirement for explicit patient consent for publication was not applicable.

Competing interests

The authors declare no competing interests.