Predicting muscle-invasive bladder cancer with dual-layer detector spectral CT-derived extracellular volume fraction

Abstract

Objectives

To evaluate the diagnostic performance of dual-layer detector spectral CT (DLCT)-derived extracellular volume fraction (ECV) for pre-operatively distinguishing muscle-invasive bladder cancer (MIBC) from non-muscle-invasive bladder cancer (NMIBC).

Materials and methods

This retrospective study included 116 patients with pathologically confirmed urothelial carcinoma who underwent preoperative DLCT. Pathological results from transurethral resection or cystectomy served as the reference standard, classifying patients into MIBC (n = 57) and NMIBC (n = 59) groups. Two radiologists independently measured morphological features and spectral CT parameters, including iodine density and ECV. Statistical analyses involved univariate comparisons, multivariable logistic regression, and receiver operating characteristic (ROC) analysis.

Results

Multivariable analysis identified the longest tumor contact length (CL) ≥ 3 cm (Odds Ratio [OR] = 5.827; p = 0.006) and ECV (OR = 1.378 per 1% increment; p < 0.001) as independent predictors of MIBC. The area under the ROC curve (AUC) for ECV (0.886) was significantly superior to that of CL ≥ 3 cm (0.726) (DeLong test, p = 0.040). A combined model (ECV + CL) achieved an AUC of 0.869, which was not significantly better than ECV alone (p = 0.146). At the optimal cut-off of 72.3%, ECV predicted MIBC with 84.2% sensitivity and 88.1% specificity. Excellent inter-reader agreement was observed for all quantitative measurements (ICC ≥ 0.86).

Conclusion

DLCT-derived ECV is a robust, non-invasive, and reproducible quantitative biomarker that outperforms conventional morphological assessment for the pre-operative prediction of MIBC, offering significant potential for improving clinical decision-making and personalized treatment planning.

Introduction

Bladder cancer (BCa) ranks among the most prevalent malignancies worldwide, with urothelial carcinoma being the predominant histological type [1]. A critical determinant of prognosis and therapeutic strategy is whether the tumor has invaded the muscularis propria. Muscle-invasive bladder cancer (MIBC) carries a significant risk of progression and metastasis, necessitating aggressive treatments such as radical cystectomy and systemic chemotherapy [2]. In contrast, non-muscle-invasive bladder cancer (NMIBC) is typically managed with organ-preserving transurethral resection, often followed by intravesical instillations [3]. This stark dichotomy underscores the paramount importance of accurate pre-operative staging.

Currently, cystoscopy with biopsy remains the diagnostic reference standard for bladder cancer, yet it is invasive, subject to sampling error, and may not capture the deepest invasion [4]. Consequently, the pursuit of non-invasive predictors has intensified. Conventional CT, reliant on subjective morphological features, exhibits variable performance and a tendency for under-staging [5]. While mpMRI with VI-RADS shows promise by probing tissue cellularity [6, 7], its variable diagnostic performance and challenges in standardization limit widespread utility [8, 9]. This landscape underscores the need for a robust, quantitative, and accessible biomarker. The aggressive nature of muscle-invasive bladder cancer (MIBC) is closely linked to extensive stromal remodeling within the tumor microenvironment. This process is primarily driven by activated cancer-associated fibroblasts that promote aberrant collagen deposition and a fibrotic reaction, leading to extracellular matrix (ECM) stiffening and architectural reorganization [10, 11]. These changes not only provide structural support for tumor invasion but also generate pro-tumorigenic signaling. Quantifying this stromal expansion noninvasively is therefore critical for accurate preoperative assessment.

The emergence of spectral CT, particularly dual-layer detector technology (DLCT), has facilitated quantitative characterization of tissue composition beyond conventional Hounsfield units. This technique enables precise quantification of iodine density (ID), reflecting tissue vascularity and blood volume [12]. More importantly, the extracellular volume (ECV) fraction—derived from equilibrium-phase iodine concentration and hematocrit—has emerged as a promising oncologic biomarker that directly quantifies the expansion of the extracellular matrix(ECM) space [13, 14]. Previous investigations in highly desmoplastic malignancies such as hepatic and pancreatic tumors have successfully employed ECV for tumor grading, staging, and fibrosis assessment, demonstrating its unique capability to capture ECM expansion during tumor progression and stromal activation [15, 16].

We hypothesize that this principle translates to bladder cancer: the distinct alterations in the tumor microenvironment of MIBC would be reflected as a significantly elevated ECV compared to NMIBC. Although ECV has been validated in other malignancies using DECT for characterizing the tumor microenvironment [17, 18], its application in bladder cancer, and specifically using the convenient DLCT technology remains largely unexplored. Based on this hypothesis, we designed this study to rigorously evaluate the diagnostic value of DLCT-derived ECV in predicting muscle invasion. Our primary objectives were: (1) to determine whether ECV can serve as an independent, non-invasive biomarker for distinguishing MIBC from NMIBC, and (2) to benchmark its performance against a conventional morphological parameter. Furthermore, we investigated whether a combined morpho-functional model could yield incremental diagnostic benefit. We present a retrospective analysis of a consecutively enrolled patient cohort who underwent pre-operative DLCT, with the aim of validating ECV as a powerful tool for pre-operative risk stratification and personalized treatment planning in bladder cancer.

Materials and methods

Patient selection

This single-institutional, retrospective study was approved by Ethics Committee of Nanxi Mountain Hospital, Guangxi Zhuang Autonomous Region [NO.2023 (KY-E-006)], which waived the requirement for written informed consent. We consecutively enrolled patients with histopathologically confirmed urothelial carcinoma of the bladder who underwent preoperative DLCT between January 2023 and April 2025. The inclusion and exclusion criteria for this study were detailed in Fig. 1. The inclusion criteria were: (1) availability of a complete triphasic DLCT scan (unenhanced, arterial, venous, and delayed phases) performed within 20 days prior to surgery; and (2) subsequent surgical treatment via transurethral resection or radical cystectomy at our institution. Patients were excluded if the CT image quality was compromised by significant artifacts or if the pathological report was unavailable. A total of 116 patients (92 men and 24 women; median age, 69 years) constituted the final study cohort. Based on the final pathological findings from surgical specimens, patients were stratified into two groups: the MIBC group (n = 57) and the NMIBC group (n = 59).

Fig. 1.

Flow chart of patients included. By Figdraw. DLCT: dual-layer detector spectral CT; SBI: spectral-based imaging MIBC: Muscle-Invasive Bladder Cancer; NMIBC: Non-Muscle-Invasive Bladder Cancer; BCa: Bladder Cancer

CT image acquisition and analysis

All CT examinations were performed using a DLCT scanner (Philips Healthcare). A standardized triphasic contrast-enhanced protocol was used, encompassing unenhanced, arterial (30–35 s), venous (60–70 s), and delayed (180 s) phases. Iodinated contrast medium (370 mg I/mL) was administered via a power injector at a flow rate of 3.0-3.5 mL/s. Two fellowship-trained genitourinary radiologists, blinded to the pathological results, independently analyzed all images. They recorded conventional morphological features, including the longest tumor contact length (CL) with the bladder wall, tumor border (well-defined or ill-defined), and the presence of intratumoral or surface calcification/necrosis. For quantitative analysis, circular or oval regions of interest (ROIs) were carefully placed on the most solid, enhancing portion of the tumor, avoiding areas of necrosis, calcification, and artifacts, on all phases (Figs. 2 and 3). The readers measured ID and the effective atomic number (Zeff) in the arterial, venous, and delayed phases. Net enhancement (δHu) was calculated as the difference between post-contrast and non-contrast attenuation values:δHu = Hu(Contrast-enhanced Phases)-Hu(non-contrast phase). The ECV was calculated using the delayed-phase iodine concentration, adjusted for the patient’s hematocrit level, with the formula: (1-haematocrit) × (iodine density in the BCa/iodine density in the common iliac artery blood) × 100%.

Fig. 2.

A case with high-grade urothelial carcinoma demonstrating full-thickness muscular invasion. Panels from left to right: non-contrast scan, 40-keV monoenergetic image, iodine density map, monoenergetic fused effective atomic number (Zeff) map, and histopathological image. A-D: Arterial phase; E-H: Venous phase; I-J: Delayed phase. The mean iodine density (ID) values across the three phases were 1.5 mg/mL (arterial), 1.9 mg/mL (venous), and 1.6 mg/mL (equilibrium), respectively. M1-M2: Histopathological sections reveal tumor cells with hyperchromatic nuclei, infiltrating stroma and muscular layer (H&E, ×100/×400)

Fig. 3.

A case with non-invasive low-grade urothelial carcinoma. Panels from left to right: non-contrast scan, 40-keV monoenergetic image, iodine density (ID) map, monoenergetic fused effective atomic number (Zeff) map, and histopathological image. a-d: Arterial phase; e-h: Venous phase; i-l: Delayed phase. The mean ID values across phases were 0.11 mg/mL (arterial), 1.35 mg/mL (venous), and 1.17 mg/mL (equilibrium), respectively. m1-m2: Histopathological sections demonstrate increased cellular stratification with mitotic figures, without muscular invasion (H&E, ×100)

Pathological evaluation

The gold standard for diagnosis and staging was the histopathological analysis of the surgical specimens obtained from transurethral resection (n = 68) or radical cystectomy (n = 48). All specimens were processed according to standard pathological protocols, fixed in formalin, embedded in paraffin, and stained with hematoxylin and eosin. The diagnosis of urothelial carcinoma and the assessment of muscle invasion were made by experienced genitourinary pathologists in accordance with the 2016 World Health Organization classification. Muscle invasion was definitively diagnosed when tumor cells were observed infiltrating into or beyond the muscularis propria (stage T2 or higher).

Statistical analysis

Statistical analyses were performed using R software (version 4.1.0). The normality of the distribution for continuous variables was assessed using the Shapiro-Wilk test. Continuous variables with a normal distribution were expressed as mean ± standard deviation and compared between the MIBC and NMIBC groups using independent-sample t-tests. Non-normally distributed variables were presented as median (interquartile range) and compared using the Mann-Whitney U test. Categorical variables, expressed as frequencies and percentages, were compared using the Chi-squared test or Fisher’s exact test, as appropriate. Inter-reader agreement for quantitative measurements was evaluated using intraclass correlation coefficients (ICCs), with values ≥ 0.75 indicating excellent agreement. Variables showing significant differences in univariate analyses (p<0.05) were included in a multivariable binary logistic regression analysis using a backward stepwise method to identify independent predictors of muscle invasion. The diagnostic performance of the independent predictors and their combination was evaluated using receiver operating characteristic (ROC) curve analysis, and the areas under the curve (AUCs) were compared using the DeLong test.

Results

Patient demographics, clinical and morphological characteristics

A total of 116 patients with pathologically confirmed urothelial carcinoma were included in the final analysis and stratified into the MIBC group (n = 57) and the NMIBC group (n = 59) based on surgical histopathology. The comparison of baseline characteristics between the two groups is detailed in Table 1.

Table 1.

Comparison of parameters between two groups

Variables	NMIBC group,59	MIBC group, 57	χ2/t/Z	p-value
Clinical characteristics
Gender(n, %)			3.720	0.054
Male	51(86.44)	41(71.93)
Female	8(13.56)	16(28.07)
Age, years, M(P25, P75)	68 (59,73)	70 (58.5,74.5)	-0.766	0.444
Long-axis diameter, cm M(P25, P75)	2.80(2.1,4.4)	3.71(2.7,6.4)	-2.593	0.010
CL (n, %)			24.976	<0.001
<3 cm	36(61.02)	9(15.79)
≥3 cm	23(38.98)	48(84.21)
Ill-defined borders (n, %)	14(23.73)	28(49.12)	8.094	0.004
Intratumoral cystic change, necrosis and calcification (n, %)	9(15.25)	28(49.12)	15.309	<0.001
Surface calcification / Calculi (n, %)	11(18.64)	26(45.61)	9.708	0.002
Gross hematuria (n, %)	40(67.80)	45(78.95)	1.841	0.175
Smoking history (n, %)	18(30.51)	11(19.30)	1.943	0.163
Ureteral orifice involvement (n, %)			2.064	0.559
None	25(42.37)	24(42.11)
Left	15(25.42)	12(21.05)
Right	12(20.34)	9(15.79)
Bilateral	7(11.86)	12(21.05)
Pathologic grading(n, %)			27.776	<0.001
High-grade	24(40.68)	50(87.72)
Low-grade	35(59.32)	7(12.28)
Net enhancement (δHu, mean ± SD, Hu)
Arterial phase	17.83 ± 10.78	21.59 ± 13.55	-1.657	0.100
Venous phase	39.90 ± 13.66	45.87 ± 15.56	-2.198	0.030
Delayed phase	29.28 ± 9.47	34.72 ± 10.24	-2.969	0.004
DLCT characteristics ID value (mean ± SD, mg/mL)
Arterial phase	0.83 ± 0.39	1.10 ± 0.53	-3.117	0.002
Venous phase	1.47 ± 0.47	1.61 ± 0.51	-1.483	0.041
Delayed phase	1.24 ± 0.37	1.48 ± 0.35	-3.590	<0.001
Zeff(mean ± SD)
Arterial phase	7.98 ± 2.40	8.71 ± 6.68	-0.783	0.435
Venous phase	7.57 ± 0.67	7.96 ± 0.54	-3.392	0.001
Delayed phase	7.49 ± 0.60	8.10 ± 1.25	-3.355	0.001
ECV(mean ± SD, %)	26.96 ± 6.01	34.91 ± 5.51	-7.426	<0.001

MIBC: Muscle-Invasive Bladder Cancer; NMIBC: Non-Muscle-Invasive Bladder Cancer; CL: The longest tumor contact length; δHu: Delta Hounsfield Units; DLCT: Dual-Layer Computed Tomography; ID: Iodine Density; SD: Standard Deviation; Zeff: Effective Atomic Number; ECV: Extracellular Volume

No significant differences were observed between the MIBC and NMIBC groups in terms of gender (p = 0.054), age (p = 0.444), history of gross hematuria (p = 0.175), smoking history (p = 0.163), or ureteral orifice involvement (p = 0.559).

Analysis of morphological features on CT revealed several significant associations with muscle invasion. The MIBC group had a significantly larger tumor long-axis diameter compared to the NMIBC group (median 3.71 cm vs. 2.80 cm, p = 0.010). A longest tumor CL with the bladder wall ≥ 3 cm was significantly more frequent in the MIBC group (84.21% vs. 38.98%, p < 0.001). Furthermore, an ill-defined tumor border (49.12% vs. 23.73%, p = 0.004), the presence of intratumoral cystic change, necrosis, and/or calcification (49.12% vs. 15.25%, p < 0.001), and the presence of surface calcification/calculi (45.61% vs. 18.64%, p = 0.002) were all significantly more prevalent in MIBC. As expected, high-grade pathology was significantly more common in the MIBC group (87.72% vs. 40.68%, p < 0.001).

Univariate analysis of quantitative CT parameters

Univariate analysis of quantitative parameters demonstrated significant differences between the MIBC and NMIBC groups (Table 1; Fig. 4).

Fig. 4.

Univariate analysis of quantitative parameters. CL: The longest tumor contact length; δHu: Delta Hounsfield Units; ID: Iodine Density; Zeff: Effective Atomic Number; ECV: Extracellular Volume

Net enhancement (δHu), the MIBC group showed significantly higher enhancement in the venous phase (45.87 ± 15.56 HU vs. 39.90 ± 13.66 HU, p = 0.030) and delayed phase (34.72 ± 10.24 HU vs. 29.28 ± 9.47 HU, p = 0.004), but not in the arterial phase (p = 0.100).

Analysis of DLCT-derived parameters revealed that ID was significantly higher in the MIBC group across all contrast phases: arterial phase (1.10 ± 0.53 mg/mL vs. 0.83 ± 0.39 mg/mL, p = 0.002), venous phase (1.61 ± 0.51 mg/mL vs. 1.47 ± 0.47 mg/mL, p = 0.041), and delayed phase (1.48 ± 0.35 mg/mL vs. 1.24 ± 0.37 mg/mL, p < 0.001). The effective atomic number (Zeff) was also significantly higher in the MIBC group during the venous phase (7.96 ± 0.54 vs. 7.57 ± 0.67, p = 0.001) and delayed phase (8.10 ± 1.25 vs. 7.49 ± 0.60, p = 0.001), but not in the arterial phase (p = 0.435).

Most notably, the ECV was significantly elevated in the MIBC group compared to the NMIBC group (34.91 ± 5.51% vs. 26.96 ± 6.01%, p < 0.001).

Multivariable logistic regression analysis

Variables showing significant differences in the univariate analysis were entered into a multivariable binary logistic regression model to identify independent predictors of muscle invasion (Table 2). The analysis confirmed that only a longest tumor CL ≥ 3 cm (Odds Ratio [OR] = 5.827; 95% Confidence Interval [CI]: 1.639–20.715; p = 0.006) and the ECV (OR = 1.378 per 1% increment; 95% CI: 1.159–1.638; p < 0.001) remained as significant independent predictors.

Table 2.

The association between MIBC Group and NMIBC group

Variables	OR(95% CI)		p-value
Long-axis Diameter	0.981(0.834–1.153)		0.815
CL	5.827(1.639–20.715)		0.006
Intratumoral cystic change, necrosis and calcification	2.538(0.521–12.377)		0.249
Ill-defined borders	0.997(0.254–3.912)		0.997
Venous phase δHu	1.021(0.972–1.073)		0.409
Delayed phase δHu	0.996(0.936–1.060)		0.899
Arterial phase ID value	2.324(0.525–10.295)		0.267
Venous phase ID value	1.034(0.213–5.026)		0.967
Delayed phase ID value	0.483(0.051–4.606)		0.527
Arterial phase -Zeff	0.964(0.866–1.073)		0.501
Venous phase -Zeff	0.812(0.251–2.629)		0.729
ECV	1.378(1.159–1.638)		<0.001

CL: The longest tumor contact length; δHu: Delta Hounsfield Units; DLCT: Dual-Layer Computed Tomography; ID: Iodine Density; SD: Standard Deviation; Zeff: Effective Atomic Number; ECV: Extracellular Volume; OR: Odds Ratio; CI: Confidence Interval

Diagnostic performance

The diagnostic performance of the independent predictors and a combined model was evaluated using receiver operating characteristic (ROC) curve analysis (Table 3; Fig. 5).

Table 3.

The diagnostic efficacy of spectral CT parameters predicting MIBC

Variables / model	AUC	Cut-off values	sensitivity	specificity	P- values	95% CI
CL	0.726	0.452	0.842	0.610	<0.001	0.632–0.820
ECV	0.886	0.723	0.842	0.881	<0.001	0.822–0.949
Combined model	0.869	NA	0.912	0.763	<0.001	0.837–0.955

AUC: Area Under the Curve; ECV: Extracellular Volume; CI: Confidence Interval; Combined model: CL + ECV

Fig. 5.

The diagnostic performance of the independent predictors. CL: The longest tumor contact length; ECV: Extracellular Volume; CL+ECV: Combined model

The AUC for ECV was 0.886 (95% CI: 0.822–0.949), which was superior to the AUC of 0.726 (95% CI: 0.632–0.820) for the wall CL criterion (DeLong test, Z = 1.842, p = 0.066). A combined model integrating both ECV and wall CL achieved an AUC of 0.869 (95% CI: 0.837–0.955). The diagnostic performance of this combined model was significantly better than that of wall CL alone (DeLong test, Z = 2.058, p = 0.040), but it was not significantly superior to ECV alone (DeLong test, Z = 1.453, p = 0.146) (Table 4).

Table 4.

Delong test of the diagnostic efffcacy between parameters and the combined mode

Parameters	Z statistic	p value
CL vs.ECV	1.842	0.066
CL vs. Combined model	2.058	0.040
ECV vs. Combined model	1.453	0.146

CL: The longest tumor contact length; ECV: Extracellular Volume; Combined model:CL+ECV

At the optimal cut-off value of 0.723, ECV predicted muscle invasion with a sensitivity of 84.2% and a specificity of 88.1%. The morphologic criterion of wall CL (using a cut-off probability of 0.452 from the ROC analysis) had a sensitivity of 84.2% and a specificity of 61.0%. The combined model provided a sensitivity of 91.2% and a specificity of 76.3%.

Inter-reader agreement

The inter-reader agreement for all quantitative measurements, including the spectral CT parameters (ID, Zeff, ECV) and the longest tumor CL, was excellent, with ICCs all exceeding 0.86.

Discussion

This study was predicated on the hypothesis that the ECV, derived from DLCT, reflects underlying pathophysiological changes in the tumor microenvironment associated with muscle invasion in bladder cancer. We aimed to evaluate the diagnostic performance of ECV and benchmark it against conventional morphological CT features. Our principal findings can be summarized as follows: First, several morphological features, particularly a longest mucosal CL ≥ 3 cm, were significantly associated with MIBC. Second, among a comprehensive panel of spectral CT quantitative parameters, the ECV was the most potent independent predictor, and the ECV demonstrated excellent diagnostic performance (AUC = 0.886), which was statistically superior to the morphological criterion of CL ≥ 3 cm and not significantly improved by a combined model.

The value of morphological features on cross-sectional imaging for pre-operative MIBC assessment has been extensively investigated. Our findings align with prior literature, confirming that tumor size and CL with the bladder wall are critical indicators [19, 20]. We identified CL ≥ 3 cm as a significant independent predictor (OR = 5.827), consistent with studies that have established a strong correlation between larger tumor CL and deeper invasion [21, 22]. This is in line with the work of Selvaraju A et al., who similarly identified tumor CL as a key predictor, underscoring the persistent value of morphological assessment despite its limitations [20]. Other features like an ill-defined border and the presence of intratumoral necrosis/calcification were also significantly more common in MIBC in our univariate analysis, underscoring the role of disruptive growth patterns. However, the subjective nature and variable diagnostic performance of these morphological assessments limit their standalone reliability for definitive pre-operative staging. Our data corroborate this, as CL ≥ 3 cm alone provided a moderate AUC of 0.726, with a specificity of only 61%.

The pursuit of objective, quantitative biomarkers has led to the exploration of functional imaging parameters. Previous studies have demonstrated the value of spectral CT quantitative parameters in predicting muscle invasion in bladder cancer. Mengting Hu et al. [23] reported that an optimal multi-image radiomics model exhibited good diagnostic performance in predicting BCa muscle invasion (AUC 0.867), outperforming NIC alone (AUC 0.704). Similarly, Changyu Du et al. developed an ensemble model combining DECT quantitative parameters, habitat features, and 2.5D deep learning features to improve the accuracy of preoperative prediction of BCa muscle invasion [24]. In contrast to these findings, our multivariable logistic regression analysis did not reveal statistically significant associations for iodine density (ID) or effective atomic number (Zeff), which may be attributable to our limited sample size. Instead, we focused on extracellular volume fraction (ECV) as a quantitative parameter, which demonstrated significant predictive value for distinguishing muscle-invasive status in our cohort. Our study demonstrates the superior predictive power of DLCT-derived extracellular volume fraction (ECV) over both morphological and other spectral CT metrics. The significantly higher ECV observed in muscle-invasive bladder cancer (MIBC) is biologically plausible and can be directly attributed to the tumor’s invasive phenotype. The process of muscle invasion is consistently accompanied by a robust stromal reaction, characterized by increased collagen deposition, fibrosis, remodeling of the extracellular matrix (ECM), and altered microvascular permeability [10, 11, 25]. By quantifying the fractional volume of the extracellular interstitial space, ECV serves as a non-invasive imaging surrogate for this pathological desmoplasia. This expansion of the extracellular compartment represents a hallmark of the invasive phenotype, often driven by processes such as epithelial-mesenchymal transition and an activated tumor microenvironment [25, 26]. The capacity of ECV to quantify stromal expansion and reflect tumor aggressiveness has been validated in other desmoplastic malignancies. For instance, in colorectal adenocarcinoma, ECV derived from dual-layer spectral-detector CT has been shown to effectively distinguish between early and advanced pathological T stages preoperatively, highlighting its sensitivity to the profound stromal remodeling associated with deeper tumor invasion [27]. Our results strongly suggest that ECV non-invasively captures this critical pathophysiological shift in bladder cancer, thereby extending its established utility from other oncologic domains into urological imaging [13].

While other studies have utilized multi-parametric MRI, specifically diffusion-weighted imaging (DWI) with the apparent diffusion coefficient (ADC), for MIBC prediction, the application of ECV from spectral CT is novel [7, 28]. The diagnostic performance of ECV (AUC 0.886) in our study is comparable to, and potentially superior than, that reported for diffusion-weighted imaging (DWI) in systematic reviews. For instance, a meta-analysis by Cornelissen et al. reported a pooled sensitivity of 0.84 and specificity of 0.82 for DWI in the local staging of bladder cancer, a benchmark which our ECV results appear to exceed [29]. Furthermore, DLCT-derived ECV presents several practical advantages over DWI. It offers superior spatial resolution, exhibits greater resistance to artifacts, and can be seamlessly incorporated into standard pre-operative CT urography protocols. This one-stop-shop advantage of DLCT, providing both anatomical and functional data in a single acquisition, addresses a key limitation of multi-modality workflows and enhances clinical efficiency [17]. Additionally, the implementation of DLCT can further optimize the examination by enabling a significant reduction in contrast medium volume, which enhances patient safety and reduces costs [30].

Interestingly, the combined model integrating ECV and CL did not yield a statistically significant improvement over ECV alone. This finding is insightful. It suggests that ECV, as a functional parameter, already encapsulates the deep biological information related to tumor aggressiveness—including the stromal changes that coincide with and are caused by tumor size and infiltration depth. The prognostic power of ECV, rooted in its ability to quantify the fibrotic tumor microenvironment, has been demonstrated as an independent predictor of survival in other highly desmoplastic cancers, such as pancreatic ductal adenocarcinoma [31, 32].This phenomenon, where a fundamental functional biomarker supersedes the predictive value of morphology, has been observed in other cancers; for instance, in colorectal cancer liver metastases, the ECV fraction has been established as an independent predictor of both treatment response and overall survival [33].In contrast, CL provides purely morphological information that largely overlaps with the biological essence reflected by ECV. Therefore, ECV may represent a more comprehensive biomarker in diagnostic performance task.

Another important result of this study is worth discussing: smoking history is included in the analysis but does not reach statistical significance. However, smoking is a major etiologic factor in bladder cancer and is associated with more aggressive disease biology, altered tumor microenvironment, and worse outcomes [34, 35]. Smoking is known to induce chronic inflammation and extracellular matrix remodeling, which may theoretically affect ECV. Although smoking history was not an independent predictor in our analysis, the lack of significance may be due to limited sample size or incomplete smoking data (e.g., pack-years not recorded). Larger studies are needed to investigate whether smoking modifies the ECV–invasion relationship.

The clinical implications of our findings are significant. Based on our results, we propose that ECV can serve as a powerful imaging indicator for identifying high-risk MIBC. This quantitative metric directly influence clinical decision-making. For instance, in patients with a high pre-operative ECV, clinicians might have greater confidence in opting for upfront radical cystectomy rather than a staged transurethral resection, thereby potentially avoiding the risks of under-staging, repeat procedures, and delayed definitive treatment. The critical need for accurate pre-operative staging to guide these consequential treatment pathways has been emphasized in urologic oncology guidelines [2]. DLCT-derived ECV offers a non-invasive, quantitative, and highly accurate tool that can significantly refine pre-operative risk stratification and patient counseling.

Our study has several strengths. It is the first to specifically validate DLCT-derived ECV for MIBC prediction. The rigorous statistical analysis, including multivariable regression and DeLong test comparison, strengthens the validity of our conclusions. Furthermore, the excellent inter-reader agreement (ICC ≥ 0.86) for all quantitative measurements underscores the reproducibility and clinical practicality of this technique. Such high reproducibility is crucial for clinical translation and is a noted advantage of quantitative CT parameters over subjective morphological assessment.

However, several limitations must be acknowledged. First, the retrospective, single-center design may introduce selection bias and limits the generalizability of our findings. Second, while the cohort size is reasonable, a larger, multi-center validation is warranted to confirm our results and solidify the proposed ECV cutoff. Third, the manual placement of regions of interest, though performed consistently by experienced radiologists, is subject to minor variability; future research could explore automated or semi-automated segmentation to enhance standardization. Fourth, the absence of lymphovascular invasion data limits our ability to assess whether ECV correlates with microscopic vascular invasion or serves as a surrogate for this established adverse pathological feature; future studies incorporating lymphovascular invasion status are warranted to clarify this relationship. Fifth, the histopathological composition of our cohorts was imbalanced, with the NMIBC group comprising 60% low-grade tumors while the MIBC group was predominantly high-grade. This disparity represents a potential confounding bias, as the observed AUC of 0.886 may partly reflect the ability to distinguish between high-grade MIBC and low-grade NMIBC rather than serving as a general discriminator between MIBC and NMIBC across all histological grades. Additionally, the majority of patients were staged by transurethral resection rather than radical cystectomy, which may have introduced discrepancies between tumor grade and definitive muscle-invasive status. The absence of detailed histologic subtyping and variant histology data further limits our ability to determine whether the elevated ECV is driven solely by muscle invasion or is confounded by aggressive tumor biology and intratumoral heterogeneity beyond simple grading. Future studies incorporating comprehensive pathological characterization are warranted to validate the robustness of ECV across different tumor phenotypes and within clinically challenging subgroups, such as high-grade T1 tumors versus MIBC, which will be a key focus of our subsequent investigations.

Conclusion

In conclusion, this study establishes the ECV as a robust and superior independent biomarker for the pre-operative prediction of MIBC. It significantly outperforms conventional morphological assessment by providing a non-invasive window into the tumor microenvironment.

The path forward involves external validation in a multi-institutional prospective cohort. Furthermore, the integration of ECV with other data streams, such as radiomic features extracted from the spectral CT datasets or genomic markers, using advanced machine learning algorithms, could potentially yield an even more powerful and comprehensive predictive model, further personalizing the management of bladder cancer. The synergy between hand-crafted biomarkers like ECV and high-dimensional radiomic data represents the next frontier in precision imaging, as preliminary studies in bladder cancer have begun to demonstrate.

Abbreviations

DLCT

Dual layer detector spectral CT

ECV

Extracellular volume fraction

BCa

Bladder cancer

MIBC

Muscle-invasive bladder cancer

NMIBC

Non-muscle-invasive bladder cancer

ID

Iodine density

ECM

Extracellular matrix

CL

Lontact length

Zeff

Effective atomic number

ICCs

Intraclass correlation coefficients

ROC

Operating characteristic

AUC

Areas under the curve

MRI

Magnetic resonance imaging

DWI

Diffusion-weighted imaging

ADC

Apparent diffusion coefficient

Author contributions

JL, PPY contributed to the conception of the study and manuscript preparation. WZ contributed signiffcantly to analysis and manuscript preparation. DH, YF, SD, and PL were responsible for raw data acquisition. RM and PY designed the figures and tables. XL optimized the experimental parameters. XQ and XZ contributed to the conception of the study and helped perform the analysis with constructive guidance. All authors contributed signiffcantly to data processing. All authors have approved the final version of the manuscript.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). The study was approved by the Ethics Committee of Nanxishan Hospital of Guangxi Zhuang Autonomous Region [NO.2023 (KY-E-006)], which waived the requirement for written informed consent.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.