Detecting clinically significant prostate cancer with a distributed parameter model based on quantitative dynamic contrast-enhanced magnetic resonance imaging

Introduction

Prostate cancer (PCa) is the sixth most commonly diagnosed malignancy and the seventh leading cause of cancer-related mortality in China (1). Due to its considerable morbidity and mortality, accurate diagnosis is crucial for guiding appropriate treatment. Currently, the diagnosis of PCa primarily relies on serological testing, histopathological biopsy, and imaging evaluation. Among these, prostate-specific antigen (PSA) screening remains the most widely used approach. However, its low specificity and high false-positive rate limit its ability to distinguish clinically significant PCa (csPCa) from indolent disease or benign prostatic hyperplasia. This limitation, compounded by the heterogeneous biological behavior of PCa, frequently leads to both overtreatment and undertreatment (2). In addition to serological testing, histopathological biopsy remains the reference standard for PCa diagnosis, providing tissue for Gleason score (GS) grading (3), which is the standard system for assessing tumor aggressiveness. However, conventional systematic biopsy, typically performed under transrectal ultrasound guidance, is invasive and associated with discomfort, bleeding, infection, and other procedure-related risks. In addition, it is susceptible to sampling errors, which may lead to the underdiagnosis of csPCa and the overdiagnosis of clinically insignificant PCa (ciPCa) (4,5). Given these limitations, biopsy is typically reserved for cases with equivocal imaging findings or persistently high clinical suspicion and is not used as a routine diagnostic tool for all patients.

Imaging has become a commonly used tool before prostate biopsy is applied in the diagnostic pathway and serves as a noninvasive means to improving patient selection and reducing unnecessary procedures. Two principal strategies are currently employed: biparametric magnetic resonance imaging (bp-MRI) is based on T2-weighted imaging (T2WI) and diffusion-weighted imaging (DWI), while multiparametric MRI (mp-MRI) additionally incorporates dynamic contrast-enhanced MRI (DCE-MRI). Of these approaches, mp-MRI has emerged as the preferred technique for the detection of PCa, local staging, and biopsy guidance (6). To standardize image acquisition and interpretation, the Prostate Imaging Reporting and Data System (PI-RADS) was introduced, providing a structured scoring framework (7). The most recent revision, PI-RADS v2.1, further refines the acquisition protocols and clarifies scoring criteria, thereby improving specificity and reducing overdiagnosis of ciPCa, and also supports MRI-guided biopsy to increase the detection rates of csPCa (8-10). According to PI-RADS v2.1, DCE-MRI should be mainly used as a binary classifier for peripheral zone (PZ) lesions scored as 3 on DWI, which should be upgraded to PI-RADS 4 when early enhancement is observed. The current assessment relies predominantly on qualitative visual interpretation of enhancement and washout patterns. However, this qualitative approach oversimplifies the complex pharmacokinetic processes underlying contrast dynamics and fails to harness the quantitative potential of DCE-MRI. Moreover, substantial overlap in enhancement characteristics between benign and malignant tissues—particularly within the transition zone (TZ)—further limits its diagnostic reliability. For example, benign prostatic hyperplasia nodules may demonstrate strong early enhancement, whereas certain malignant lesions may lack early enhancement or washout, leading to false-positive or false-negative results (11). Due to these limitations, there is a need to better exploit the diagnostic potential of quantitative DCE-MRI and expand its role within the PI-RADS framework.

Indeed, several studies have investigated the use of quantitative DCE-MRI-derived parameters for the assessment of PCa (12,13). The Tofts model (14) adopts a two-compartment framework with the assumption of instantaneous equilibrium to estimate tracer kinetics. However, with these simplified and incomplete compartmental assumptions, this model cannot fully characterize tissue perfusion and vascular permeability dynamics, which limits its diagnostic accuracy and results in suboptimal performance in tumor diagnosis (15). To overcome these limitations, the extended Tofts model (ETM) (16) was developed and included plasma volume (V_P) fraction, which can improve the separation of vascular and extravascular contributions and enhance the physiological interpretability of the estimated parameters. Park et al. applied the ETM for differentiating csPCa from ciPCa and found that extravascular extracellular volume fraction (V_e) exhibited certain discriminatory potential, although its area under the curve (AUC) was just 0.643 due to the quality and size of the dataset (17). Recently, quantitative DCE-MRI image analysis has progressed markedly, and more advanced tracer kinetic models have been developed, including the adiabatic tissue homogeneity (ATH) model (18,19) and the distributed parameter (DP) model (20,21). Compared with the Tofts model or ETM, the DP model explicitly accounts for the bidirectional exchange of the tracer between the vascular and extravascular-extracellular compartments, as well as intravascular transport effects. By incorporating these exchange dynamics, the model is a more physiologically complete representation of tissue microcirculation and allows for the separate estimation of blood flow, vascular permeability-surface area product (PS), mean transit time (MTT), and V_P, thereby providing a more comprehensive description of microvascular physiology and tracer transport. Thus, the DP model is particularly well suited for capturing the heterogeneity of tumor perfusion and permeability, which are critical for distinguishing csPCa from ciPCa. On this basis, we evaluated the ability of DP model-derived quantitative parameters to differentiate ciPCa from csPCa and compared their diagnostic performance with that of ETM and the ATH model to determine their relative clinical utility. This study represents a continuation and extension of our previous research (22). We present this article in accordance with the STARD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2717/rc).

Methods

Patient

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments and was approved by the Institutional Research Ethics Board of The First People’s Hospital of Yunnan Province (approval No. KHLL 2021-137). Written informed consent was obtained from all participants. Consecutive patients with clinical suspicion of PCa who underwent DCE-MRI examination between June 2022 and May 2025 were considered eligible for the study. The inclusion criteria were as follows: (I) an elevated PSA; (II) clinical symptoms; (III) no history of androgen castration treatment, chemoradiotherapy, or biopsy before MRI examination; (IV) acceptable quality of MR images; and (V) pathological results obtained by a combination of standard transrectal ultrasound (TRUS)-guided 12-core systematic biopsy and MRI-TRUS cognitive fusion biopsy (n=33) or radical prostatectomy (n=17) within a week after the MR examination. Patients were excluded for the following reasons: (I) unsatisfactory image quality of DCE-MRI, such as significant motion artifacts (n=3) and difficulty in delineating artery input (n=10); (II) prostate biopsy performed within 6 weeks before the MRI examination (n=2); and (III) absence of histopathologic reports due to no biopsy or no surgery (n=7). Although a subset (data collected between June 2022 and February 2024) of the data has been reported previously (22), this overlap in data sample did not affect the results of the present study. Specifically, the earlier work focused on differentiating PCa from normal prostate tissue through use of DCE-MRI kinetic parameters derived from the Tofts model, ETM, Brix’s conventional two-compartment model (Brix), ATH model, DP model, and the initial area under the signal-time curve (IAUC), whereas the present study investigated the diagnostic value of quantitative DCE-MRI coupled with the ETM, ATH model, and DP model in distinguishing csPCa from ciPCa within a larger dataset (data collected between June 2022 and May 2025). Consistent with a recent review (3), PCa was classified as ciPCa if GS ≤6 and as csPCa if GS ≥7. The flowchart of patient inclusion is provided in Figure 1.

Figure 1 Flowchart of patient inclusion. DCE, dynamic contrast-enhanced; MRI, magnetic resonance imaging; PCa, prostate cancer; TRUS, standard transrectal ultrasound.

Image acquisition

MRI examinations were performed on a 3.0-T scanner (MAGNETOM Prisma, Siemens Healthineers, Erlangen, Germany) at The First People’s Hospital of Yunnan Province. Each scan included axial T1-weighted imaging (T1WI) [repetition time (TR) =500 ms, echo time (TE) =9.7 ms, slice thickness =3.0 mm, gap =0.3 mm, and number of slices =24], T2WI in three planes (TR =3,000–4,000 ms, TE =107–114 ms, slice thickness =3.0 mm, and gap =0.3 mm), DWI (TR =4,000 ms, TE =57 ms, slice thickness =3.0 mm, gap =0.3 mm, and b-values =50, 1,000, and 2,000 s/mm²), and DCE-MRI. DCE-MRI was performed with a three-dimensional volumetric interpolated breath-hold examination sequence in the axial direction (TR =2.8 ms, TE =0.82 ms, slice thickness =3.0 mm, gap =0.6 mm, field of view =300 mm × 248 mm, matrix =160×99, number of slices =24, number of excitations =1, postcontrast flip angle =15°, and precontrast flip angles =5°, 10°, and 15°). Before injection of the contrast agent, 10 repetitions of the sequence were performed for each flip angle (5°, 10°, and 15°), and native (precontrast) tissue T1 values were estimated with the variable flip angle method. A total of 120 dynamic scans with a temporal resolution of 2 s were performed immediately after intravenous administration of gadopentetate dimeglumine (Magnevist, Bayer HealthCare Pharmaceuticals, Leverkusen, Germany) at a rate of 2.0 mL/s and a dose of 0.1 mmoL/kg body weight. Tissue contrast concentration-time curves were derived from the dynamic scans via estimation of the difference in post- and precontrast relaxation rate (1/T1) for kinetic modeling (22).

Image analysis

The DCE-MRI data were imported into commercially available software (FISCA Healthcare, Nanjing, China) for kinetic model analytics. For each participant, the arterial input function was sampled from the bilateral (internal and external) iliac artery in the central slices to avoid possible effects of inflow and inhomogeneity near boundaries. A single arterial input function was used in the process of computation. A total of six parameters from the DP model, four from the ETM, and seven from the ATH model were generated on a voxel-wise basis. Voxels exhibiting fitting failures or nonphysiological values (i.e., undefined numeric results in the process of numerical computation, such as division by zero) were excluded from subsequent analysis. Detailed descriptions of parameter abbreviations, definitions, and units are provided in our previous publication (22). Regions of interest (ROIs) for the DP/ATH models and ETM were manually delineated on PS and transfer constant (K^trans) maps independently by two experienced radiologists with 6 and 28 years of expertise in prostate MRI, respectively. ROI placement was cross-referenced with T2WI and DWI, and the median value within each ROI was calculated for statistical analysis. The intraclass correlation coefficient (ICC) was used to test the interobserver consistency. Table 1 shows the ICC values for the measured parameters, where most ICC values are greater than 0.8, indicating very good agreement between measurements from two observers. Thus, the parameter values as measured by two observers were averaged and used in the subsequent analyses.

Statistical analysis

Statistical analyses were performed with MATLAB 2020b (MathWorks, Natick, MA, USA). All histopathologically confirmed PCa lesions were included. Continuous variables are expressed as the median and interquartile range and were compared with the Mann-Whitney U test; meanwhile, categorical variables are expressed as frequencies with percentages and were analyzed with the Fisher’s exact test. A P<0.05 was considered statistically significant.

DCE-MRI parameters derived from the ETM, ATH model, and DP model were evaluated for their ability to differentiate ciPCa from csPCa lesions, with the GS serving as the reference standard. Analyses were performed in the overall cohort and further stratified by lesion location, including the PZ and TZ. Additional subgroup analyses were conducted according to lesion size (<1 vs. ≥1 cm) and time-intensity curve (TIC) types. The diagnostic performance of individual DCE-MRI parameters was assessed via receiver operating characteristic (ROC) curve analysis. For each parameter, sensitivity, specificity, accuracy, area under the ROC curve (AUC), cutoff value, negative predictive value (NPV), and positive predictive value (PPV) were calculated. An AUC value above 0.8 was considered to indicate effectiveness for discrimination.

To further investigate the added diagnostic value of DCE-MRI modeling, bp-MRI and mp-MRI approaches were compared within the PI-RADS framework. In this study, bp-MRI included lesion assessment based on T2WI and DWI in accordance with the PI-RADS criteria, whereas mp-MRI included additional DCE-MRI-derived parameters. The performance of bp-MRI and mp-MRI in distinguishing ciPCa from csPCa was subsequently evaluated.

Results

Patient demographics

A total of 70 patients with 88 biopsy-proven PCa lesions were included, comprising 42 ciPCas (GS 3+3=6) and 46 csPCas (GS >3+3). The median age (72 vs. 71 years) and weight (76 vs. 78 kg) did not differ significantly between groups (both P>0.05). In contrast, PSA and PSA density were markedly higher in the csPCa group than in the ciPCa group (PSA: 23.5 vs. 6.74 ng/mL; PSA density: 0.53 vs. 0.16 ng/mL/cc; both P<0.001). The distribution of ciPCa and csPCa was found to be 58.7% and 69.0% in the PZ, respectively, while it was 41.3% and 31.0% in the TZ, respectively, which did not represent a significant difference (P=0.26). Tumor size ≥1 cm was more frequent in the csPCa group (95.7% vs. 83.3%; P=0.03). Enhancement curve types also differed significantly, with type 3 curves being more common in the csPCa group than in the ciPCa group (71.7% vs. 45.2%; P=0.002). The details of the characteristics are summarized in Table 2.

Diagnostic performance of DCE parameters in differentiating csPCa from ciPCa

To evaluate the diagnostic utility of quantitative DCE parameters, the performance of kinetic parameters derived from different tracer kinetic models in distinguishing ciPCa from csPCa was analyzed. Representative cases are illustrated in Figures 2,3. In the low-grade case (GS 3+3=6), the suspected lesion in the TZ appeared equivocal, with reduced but indeterminate signal changes on T2WI and DWI. However, quantitative pharmacokinetic maps provided more definitive information: the DP-derived PS (21.0) and ATH-derived PS (23.6) were both markedly lower than the median values observed in csPCa (33.5 and 35.6, respectively), suggesting absence of high perfusion and indicating a lower-grade lesion. These findings were consistent with a diagnosis of ciPCa, which was subsequently confirmed by biopsy, thereby supporting the diagnostic value of DCE-derived parameters. In contrast, the high-grade case (GS 4+4=8) exhibited a clearly defined lesion with low T2 signal and marked hyperintensity on DWI. DCE-MRI revealed heterogeneous but intense enhancement, and quantitative analysis showed a significantly higher PS derived from the DP model (34.6), a high K^trans from the ETM (0.33), and a higher PS from the ATH model (34.0), collectively indicating substantially higher perfusion and vascular permeability. These findings illustrate the ability of DCE-MRI-derived parameters across different pharmacokinetic models (DP model, ETM, and ATH model) to capture distinct perfusion and permeability characteristics between ciPCa and csPCa.

Figure 2 A 76-year-old male (78 kg) with histopathologically confirmed prostate cancer with GS 6 (3+3) via biopsy. (A) T2WI revealed a suspected low-signal-intensity lesion (arrow) in the TZ. (B) DWI showed an indeterminate lesion in the corresponding region (arrow). (C) DCE-MRI revealed a lesion with ill-defined margins (arrow). (D) DP model-derived parametric map of the ROI (arrow) indicated a PS of 21.0. (E) ETM analysis of the ROI (arrow) indicated a K^trans of 0.28. (F) ATH model-based analysis of the ROI (arrow) indicated a PS of 23.6. ATH, adiabatic tissue homogeneity model; DCE-MRI, dynamic contrast-enhanced magnetic resonance imaging; DP, distributed parameter; DWI, diffusion weighted imaging; ETM, extended Tofts model; GS, Gleason score; K^trans, transfer constant; PS, permeability surface area product; ROI, region of interest; T2WI, T2 weighted imaging; TZ, transitional zone.

Figure 3 A 71-year-old male (79 kg) with histopathologically confirmed prostate cancer with GS 8 (4+4) located in the PZ. (A) T2WI revealed a focal low-signal-intensity lesion (arrow) within the prostate gland. (B) DWI showed corresponding high signal intensity (arrow). (C) DCE-MRI revealed heterogeneous enhancement in the same ROI (arrow). (D) The parametric map derived from the DP model demonstrated a markedly increased PS of 34.6, indicating high perfusion within the lesion (arrow). (E) The ETM parametric map of the ROI (arrow) indicated an elevated K^trans of 0.33, consistent with increased vascular permeability. (F) The ATH model-derived map of the ROI (arrow) indicated a PS of 34.0. ATH, adiabatic tissue homogeneity model; DCE-MRI, dynamic contrast-enhanced magnetic resonance imaging; DP, distributed parameter; DWI, diffusion weighted imaging; ETM, extended Tofts model; GS, Gleason score; K^trans, transfer constant; PS, permeability surface area product; PZ, peripheral zone; ROI, region of interest; T2WI, T2 weighted imaging.

Tables 3-5 summarize the quantitative DCE parameters derived from the three pharmacokinetic models, which were assessed for their ability to discriminate csPCa from ciPCa in the overall cohort and in subgroup analyses stratified by lesion location (PZ and TZ) and lesion size (≥1 cm). For the ETM (Table 3), both K^trans and rate constant of backflux (K_ep) values were significantly higher in the csPCa group than in the ciPCa group (all P<0.001), whereas V_e and V_P did not show significant differences. For the ATH model (Table 4), the blood flow (F), extraction fraction (E), elimination rate constant (Ke), and PS values were significantly elevated in the csPCa group as compared with those in the ciPCa group for overall and PZ lesions, with PS demonstrating the most consistent discriminative ability (P<0.001 for overall and PZ). In contrast, MTT and V_P did not show significant differences. For the DP model (Table 5), multiple parameters—including F, MTT, V_P, and PS—were significantly higher in the csPCa group than in the ciPCa group across most analyses. Among these, PS exhibited the strongest discriminative performance, with highly significant differences between csPCa and ciPCa in the overall, PZ, TZ, and ≥1 cm lesions (all P<0.001). These findings indicate that, although the ETM and ATH model-derived parameters can differentiate csPCa from ciPCa to an extent, the DP model provides superior and more consistent discriminative capability, particularly through the PS parameter.

Diagnostic performance of DCE parameters based on ROC analysis

To further evaluate the diagnostic characteristics of quantitative DCE-MRI parameters, ROC curve analysis was performed with GS serving as the reference standard. This allowed for a direct comparison of kinetic parameters derived from different tracer kinetic models. Figure 4 presents the ROC analyses of the three DCE-MRI kinetic models with reference to GS, showing clear variability in diagnostic performance across the parameters. Within the DP model, PS achieved the highest discriminative accuracy (AUC =0.87), followed by F (0.83) and V_P (0.81), whereas MTT, E, and V_e demonstrated limited utility (AUC ≤0.68). The ATH model similarly identified PS as the most informative parameter (AUC =0.82), with E, K_e, and F showing only moderate performance (AUC =0.69–0.70) and V_P/V_e approaching chance levels. The ETM yielded the weakest results overall, with K^trans (0.71) and K_ep (0.68) performing modestly and V_e (0.54) and V_P (0.52) demonstrating minimal discriminative capacity.

Figure 4 ROC curves of the DCE parameters derived from the three models with AUC values. (A) ETM, (B) ATH model, and (C) DP model. ATH, adiabatic tissue homogeneity mode; AUC, area under the curve; DCE, dynamic contrast-enhanced; DP, distributed parameter; ETM, extended Tofts model; K_e, elimination rate constant; K^trans, transfer constant; MTT, mean transit time; PS, permeability surface area product; ROC, receiver operating characteristic; V_e, extra-vascular extracellular volume fraction; V_P, plasma volume fraction.

Table 6 provides a detailed summary of the diagnostic accuracy metrics. Consistent with the ROC findings, the ETM-derived parameters exhibited only modest performance, while the ATH model-derived PS showed moderate diagnostic value (AUC =0.82, accuracy =0.80, and specificity =0.86). By contrast, the DP model demonstrated the most robust and consistent diagnostic characteristics, with PS yielding the highest overall performance (AUC =0.87, specificity =0.83, and accuracy =0.78), closely followed by F and V_P. Other DP model-derived measures, including MTT and V_e, provided limited diagnostic contribution. Overall, these results confirm that quantitative DCE parameters possess clinically relevant value in diagnosing PCa, with the DP model—particularly PS—demonstrating superior accuracy and reliability in differentiating csPCa from ciPCa.

Superior diagnostic performance of mp-MRI over bp-MRI with DCE

bp-MRI, which relies on T2WI and DWI within the PI-RADS framework, has been increasingly used for PCa detection due to its shorter acquisition time and lack of contrast administration. However, the omission of DCE may limit its sensitivity, particularly in equivocal cases, and diagnostic accuracy based solely on T2WI and DWI can be especially prone to error among less-experienced readers. To evaluate the incremental diagnostic value of DCE, we therefore compared the performance of bp-MRI with that of MRI (mp-MRI), which incorporates DCE-derived parameters, using PI-RADS assessments made by a junior radiologist to specifically test whether DCE improves diagnostic reliability.

As shown in Table 7, bp-MRI (T2WI + DWI) provided a baseline (AUC =0.88, sensitivity/specificity =0.85/0.81, accuracy =0.83, and NPV/PPV =0.83/0.83). Incorporating DCE into mp-MRI yielded differing effects across models. The ETM and ATH model had increased sensitivity (both 0.89) but at the cost of specificity (0.74 and 0.76, respectively), resulting in unchanged or only marginally improved discrimination (AUC =0.88–0.89) and no gain in accuracy (0.82–0.83). In contrast, the DP model-based mp-MRI achieved the best overall performance, with the highest AUC (0.92) and accuracy (0.86), alongside balanced sensitivity/specificity (0.91/0.80). Notably, the DP model improved NPV to 0.89 while maintaining PPV at 0.83—matching bp-MRI—indicating improved rule-out capability without a compromise in positive predictive performance. Overall, these findings indicate that the addition of DCE enhances the diagnostic accuracy of prostate MRI as compared with bp-MRI alone. Among the three pharmacokinetic models, the DP model demonstrated the most robust overall diagnostic performance, achieving the highest AUC, accuracy, and sensitivity-specificity balance in differentiating csPCa from ciPCa.

Discussion

In this study, we investigated the value of quantitative DCE-MRI in diagnosing PCa, focusing on three kinetic models (DP model, ATH model, and ETM). Our results demonstrated that the parameter values in csPCa were consistently higher than those in ciPCa and that PS derived from the DP model achieved the best performance (AUC =0.87). Importantly, integrating DCE parameters provided greater improvements for the diagnostic accuracy of mp-MRI (0.86) than for that of bp-MRI (0.83), suggesting that the role of DCE in prostate MRI could be augmented if quantitative parameters derived with an advanced kinetics model are employed for examining PCa.

In PI-RADS, lesion categorization primarily centers on three subjective criteria—shape, signal intensity, and lesion boundaries. Such qualitative descriptions are inherently susceptible to variability in image quality and radiologist experience, often resulting in limited inter- and intraobserver agreement. Ziayee et al. systematically compared the diagnostic performance of radiologists with varying levels of experience with and without DCE and demonstrated that DCE provided significant diagnostic benefit for less-experienced radiologists (23). Our findings align with this observation; specifically, when a less-experienced radiologist assessed lesions using only T2WI and DWI within the bp-MRI framework (in Section “Superior diagnostic performance of mp-MRI over bp-MRI with DCE”), diagnostic performance was limited (Table 7). However, when DCE-derived parameters were incorporated into mp-MRI, diagnostic performance improved. These results highlight the potential of DCE to reduce interreader variability and to enhance diagnostic reliability, particularly for less-experienced readers (23).

Despite these advantages, the PI-RADS v2.1 guidelines assign DCE-MRI only a secondary role, primarily for upgrading equivocal PZ lesions (DWI score =3) (8). Moreover, the interpretation of DCE-MRI remains qualitative, and there is a focus on the presence or absence of focal early contrast enhancement rather than on quantitative measurements. In fact, DCE-MRI provides crucial biological information on tissue perfusion that cannot be replaced by other sequences, and numerous studies have examined the added benefits of quantitative pharmacokinetic methods (24), but their reported benefits remain inconclusive. This indicates the need for further research to establish robust quantitative DCE biomarkers that can be systematically integrated into clinical practice and potentially elevate the diagnostic weight of the biomarkers within the PI-RADS framework.

In recent years, quantitative DCE has garnered heightened interest in the field of prostate MRI (25). Peng et al. reported that K^trans estimated with ETM shows no additional benefit compared with the apparent diffusion coefficient in differentiating PCa from benign tissues (26). Meanwhile, Ziayee et al. found that none of the Tofts-derived parameters could differentiate csPCa from ciPCa (11). However, other reports have shown that DCE contributes to improved lesion characterization, particularly in the assessment of recurrence or in cases in which the prostate morphology is altered. Franiel et al. reported that blood volume, MTT, and permeability derived from a sequential three-compartment model facilitate the differentiation of csPCa and ciPCa (27). Tosoian et al. found that integrating DCE into clinical practice could improve risk stratification accuracy and reduce the need for invasive biopsies, thereby minimizing the adverse effects of overtreatment (28). These contradictory findings reflect the ongoing controversy regarding the clinical utility of DCE-MRI.

Our recent study indicated that most kinetic tracer model-derived parameters—including MTT, V_P, and PS from the DP model and E and PS from the ATH model—can differentiate malignant lesions from benign tissues in the prostate (22). The performance of ETM-derived parameters is inferior to that of DP or ATH model-derived parameters. The less encouraging performance of ETM, along with aforementioned conflicting results across different PCa DCE studies, could be attributed to a variety of factors such as population differences, MRI protocols, model choice, and reference standards. Theoretically, this can be explained by the more simplified assumptions about the intravascular distribution of gadolinium contrast in the Tofts model and ETM. In tissue microenvironment, the transport of contrast agent essentially includes the transport driven by blood flow and the exchange between intravascular space and extravascular space reflecting the permeability of the vessel wall. The original Tofts model assumed that the intravascular space is very small and does not take into account this space. The limitation of this assumption in terms of tumor interrogation was soon appreciated, and a parameter of V_P was added into the Tofts model, resulting in the ETM. However, the impact of blood flow on the transport of contrast agent remains unaccounted for, which might explain the less encouraging performance of the Tofts model or ETM and the conflicting findings in PCa DCE studies.

For example, K^trans from the ETM is often used as a surrogate for PS, but it describes PS only when tracer transport across the vessel wall is limited by permeability in situations of high blood flow. Both the PS from the ATH and DP models indicate a high degree of diagnostic accuracy in detecting csPCa, with AUC values greater than 0.8, and the PS from the DP model achieved the best performance (AUC =0.87). The single-parameter PS had an AUC range of 0.85–0.89 in detecting csPCa, which is comparable to that of experienced radiologists (29). These results highlight the potential added diagnostic value of DCE and indicate that its contribution in PI-RADS might be improved if DCE data are analyzed with an advanced kinetics model.

In PI-RADS v2.1, DCE is considered only a secondary criterion and primarily used to upgrade equivocal (DWI score 3) lesions in the PZ. This restricted weighting limits the role of DCE, although several studies have demonstrated its potential value in cases of altered prostate morphology. For example, Stabile et al. demonstrated that DCE is useful for evaluating local recurrence after interventions such as transurethral resection and focal therapy, for which standard PI-RADS scoring may be inapplicable (30). Similarly, Panebianco et al. reported that DCE-MRI can be more valuable than DWI in assessing radiation therapy, as DWI is more prone to artifacts from low-dose-rate treatments (31). In our study, incorporating quantitative DCE parameters into mp-MRI led to improved diagnostic performance as compared with bp-MRI (Table 7). These findings indicate that the diagnostic value of DCE for PCa could be strengthened if an advanced kinetics model is employed to process the acquired DCE data and derive quantitative tissue microenvironment parameters interrogation. Quantitative modeling, particularly with the DP model, can provide objective biomarkers to support biopsy decisions and improve mp-MRI accuracy, with potential for future development as standalone imaging markers in the evaluation of PCa. Of note, in the DP model, the sensitivity (0.83) was higher than the specificity (0.76) for V_P, while for PS, the sensitivity (0.74) was lower than the specificity (0.83). In terms of clinical application, V_P from the DP model could be valuable in detecting csPCa, for which misdiagnosis could have serious consequences, whereas PS may be useful in detecting ciPCa, for which a false diagnosis could lead to overtreatment.

Furthermore, image quality remains a critical determinant of diagnostic accuracy in prostate MRI and is explicitly emphasized in PI-RADS v2.1. Stanzione et al. investigated discrepant prostate MRI and biopsy results and reported that one major factor was suboptimal MRI image quality and its associated diagnostic errors (32). Verma et al. concluded that a high-quality DCE-MRI examination is essential for evaluating the extent of primary and recurrent PCa, which requires a good understanding of the technical aspects and limitations of image acquisition and postprocessing techniques (13). For DCE-MRI, in addition to appropriate kinetic modeling, optimal DCE-MR scans are essential for reliable quantitative analysis. PI-RADS v2.1 recommends a temporal resolution below 15 seconds. In this study, a high-temporal-resolution protocol (2 seconds) was employed to better capture rapid enhancement dynamics. High-field 3.0-T scanners are preferred for generating high-quality images, while robust motion correction is necessary to mitigate artifacts that can otherwise compromise kinetic parameter estimation.

There are several limitations in this study that should be addressed. First, this work represents a preliminary evaluation of DP model-derived parameters in detecting csPCa and was conducted in a relatively small cohort of 70 patients with 88 lesions from a single institution. Subgroup analyses (e.g., PZ/TZ and ≥1/<1 cm) further reduced the sample size, potentially lowering statistical power. Validation in studies with a larger, multicenter sample among a diversity of populations is needed. Second, although no significant differences in DP model-derived parameters were observed between PZ and TZ lesions in this study, clinical heterogeneity across tumor locations might have influenced the outcomes. Patients with PZ tumors often experience poorer prognosis than do those with TZ tumors (33), partly due to these tumors’ anatomical proximity to vessels, nerves, the prostatic capsule, and seminal vesicles. Further work is required to determine whether quantitative DCE parameters can capture these biological and clinical differences. Finally, ROIs were manually delineated on PS maps from the DP model, K^trans maps from the ETM, and PS maps from the ATH model by two experienced radiologists. Although using median values within each ROI helps mitigate sensitivity to precise boundary placement, manual delineation may still introduce observer-related variability.

Conclusions

This study examined whether quantitative DCE-MRI could improve the differentiation of csPCa and ciPCa, addressing a key limitation of the current PI-RADS framework in its underutilization of the diagnostic potential of DCE parameters. Kinetic parameters derived from ETM, the ATH model, and the DP model were systematically analyzed in a cohort of patients with biopsy-proven lesions, with GS serving as the reference standard. The results showed that the ETM and ATH model parameters yielded only moderate diagnostic performance, whereas the DP model, particularly the PS, consistently achieved the highest discriminative accuracy. These findings highlight that quantitative DCE-MRI, especially when analyzed with the DP model, provides robust and complementary diagnostic information beyond conventional qualitative assessment. Incorporation of these parameters into mp-MRI interpretation could enhance risk stratification, reduce interreader variability, and ultimately improve clinical decision-making in the management of patients with PCa.

Acknowledgments

The authors would like to thank FISCA Healthcare for providing the DCE calculation model.

Footnote

Reporting Checklist: The authors have completed the STARD reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2717/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2717/dss

Funding: This study was supported by the Natural Science Foundation of Zhejiang Province (No. ZCLQN26F0206), the Jiaxing Public Welfare Project (No. 2023AY11052), and the Yunnan Provincial Department of Science and Technology-Kunming Medical University Applied Basic Research Joint Special General Project (No. 202201AY07 0001-254).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2717/coif). The authors have no conflicts of interest to declare.

Ethical Statement: 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 and its subsequent amendments. This study received approval from the Institutional Research Ethics Board of The First People’s Hospital of Yunnan Province (approval No. KHLL 2021-137), and written informed consent was obtained from all participants.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Han B, Zheng R, Zeng H, Wang S, Sun K, Chen R, Li L, Wei W, He J. Cancer incidence and mortality in China, 2022. J Natl Cancer Cent 2024;4:47-53. [Crossref] [PubMed]
2. Ilic D, Djulbegovic M, Jung JH, Hwang EC, Zhou Q, Cleves A, Agoritsas T, Dahm P. Prostate cancer screening with prostate-specific antigen (PSA) test: a systematic review and meta-analysis. BMJ 2018;362:k3519. [Crossref] [PubMed]
3. Matoso A, Epstein JI. Defining clinically significant prostate cancer on the basis of pathological findings. Histopathology 2019;74:135-45. [Crossref] [PubMed]
4. Epstein JI, Zelefsky MJ, Sjoberg DD, Nelson JB, Egevad L, Magi-Galluzzi C, Vickers AJ, Parwani AV, Reuter VE, Fine SW, Eastham JA, Wiklund P, Han M, Reddy CA, Ciezki JP, Nyberg T, Klein EA. A Contemporary Prostate Cancer Grading System: A Validated Alternative to the Gleason Score. Eur Urol 2016;69:428-35. [Crossref] [PubMed]
5. Pierorazio PM, Walsh PC, Partin AW, Epstein JI. Prognostic Gleason grade grouping: data based on the modified Gleason scoring system. BJU Int 2013;111:753-60. [Crossref] [PubMed]
6. Smith CP, Harmon SA, Barrett T, Bittencourt LK, Law YM, Shebel H, An JY, Czarniecki M, Mehralivand S, Coskun M, Wood BJ, Pinto PA, Shih JH, Choyke PL, Turkbey B. Intra- and interreader reproducibility of PI-RADSv2: A multireader study. J Magn Reson Imaging 2019;49:1694-703. [Crossref] [PubMed]
7. Turkbey B, Purysko AS. PI-RADS: Where Next? Radiology 2023;307:e223128. [Crossref] [PubMed]
8. Turkbey B, Rosenkrantz AB, Haider MA, Padhani AR, Villeirs G, Macura KJ, Tempany CM, Choyke PL, Cornud F, Margolis DJ, Thoeny HC, Verma S, Barentsz J, Weinreb JC. Prostate Imaging Reporting and Data System Version 2.1: 2019 Update of Prostate Imaging Reporting and Data System Version 2. Eur Urol 2019;76:340-51. [Crossref] [PubMed]
9. Moldovan PC, Van den Broeck T, Sylvester R, Marconi L, Bellmunt J, van den Bergh RCN, et al. What Is the Negative Predictive Value of Multiparametric Magnetic Resonance Imaging in Excluding Prostate Cancer at Biopsy? A Systematic Review and Meta-analysis from the European Association of Urology Prostate Cancer Guidelines Panel. Eur Urol 2017;72:250-66.
10. Panebianco V, Barchetti G, Simone G, Del Monte M, Ciardi A, Grompone MD, Campa R, Indino EL, Barchetti F, Sciarra A, Leonardo C, Gallucci M, Catalano C. Negative Multiparametric Magnetic Resonance Imaging for Prostate Cancer: What's Next? Eur Urol 2018;74:48-54. [Crossref] [PubMed]
11. Ziayee F, Ullrich T, Blondin D, Irmer H, Arsov C, Antoch G, Quentin M, Schimmöller L. Impact of qualitative, semi-quantitative, and quantitative analyses of dynamic contrast-enhanced magnet resonance imaging on prostate cancer detection. PLoS One 2021;16:e0249532. [Crossref] [PubMed]
12. Alonzi R, Padhani AR, Allen C. Dynamic contrast enhanced MRI in prostate cancer. Eur J Radiol 2007;63:335-50. [Crossref] [PubMed]
13. Verma S, Turkbey B, Muradyan N, Rajesh A, Cornud F, Haider MA, Choyke PL, Harisinghani M. Overview of dynamic contrast-enhanced MRI in prostate cancer diagnosis and management. AJR Am J Roentgenol 2012;198:1277-88. [Crossref] [PubMed]
14. Tofts PS, Brix G, Buckley DL, Evelhoch JL, Henderson E, Knopp MV, Larsson HB, Lee TY, Mayr NA, Parker GJ, Port RE, Taylor J, Weisskoff RM. Estimating kinetic parameters from dynamic contrast-enhanced T(1)-weighted MRI of a diffusable tracer: standardized quantities and symbols. J Magn Reson Imaging 1999;10:223-32. [Crossref] [PubMed]
15. Sourbron SP, Buckley DL. Tracer kinetic modelling in MRI: estimating perfusion and capillary permeability. Phys Med Biol 2012;57:R1-33. [Crossref] [PubMed]
16. Tofts PS. Modeling tracer kinetics in dynamic Gd-DTPA MR imaging. J Magn Reson Imaging 1997;7:91-101. [Crossref] [PubMed]
17. Park H, Kim SH, Kim JY. Dynamic contrast-enhanced magnetic resonance imaging for risk stratification in patients with prostate cancer. Quant Imaging Med Surg 2022;12:742-51. [Crossref] [PubMed]
18. St Lawrence KS, Lee TY. An adiabatic approximation to the tissue homogeneity model for water exchange in the brain: I. Theoretical derivation. J Cereb Blood Flow Metab 1998;18:1365-77.
19. St Lawrence KS, Lee TY. An adiabatic approximation to the tissue homogeneity model for water exchange in the brain: II. Experimental validation. J Cereb Blood Flow Metab 1998;18:1378-85.
20. Dennis Cheong LH, Tchoyoson Lim CC, Koh TS. Dynamic contrast-enhanced CT of intracranial meningioma: comparison of distributed and compartmental tracer kinetic models--initial results. Radiology 2004;232:921-30. [Crossref] [PubMed]
21. Koh TS, Cheong LH, Hou Z, Soh YC. A physiologic model of capillary-tissue exchange for dynamic contrast-enhanced imaging of tumor microcirculation. IEEE Trans Biomed Eng 2003;50:159-67. [Crossref] [PubMed]
22. Zhang H, Yang J, Wu K, Hou Z, Du J, Yan J, Zhao Y. Comparison of tracer kinetic models in differentiating malignant from normal prostate tissue using dynamic contrast-enhanced MRI. Front Oncol 2024;14:1450388.
23. Ziayee F, Schimmöller L, Boschheidgen M, Kasprowski L, Al-Monajjed R, Quentin M, Radtke JP, Albers P, Antoch G, Ullrich T. Benefit of dynamic contrast-enhanced (DCE) imaging for prostate cancer detection depending on readers experience in prostate MRI. Clin Radiol 2024;79:e468-74. [Crossref] [PubMed]
24. Pinto PA, Chung PH, Rastinehad AR, Baccala AA Jr, Kruecker J, Benjamin CJ, Xu S, Yan P, Kadoury S, Chua C, Locklin JK, Turkbey B, Shih JH, Gates SP, Buckner C, Bratslavsky G, Linehan WM, Glossop ND, Choyke PL, Wood BJ. Magnetic resonance imaging/ultrasound fusion guided prostate biopsy improves cancer detection following transrectal ultrasound biopsy and correlates with multiparametric magnetic resonance imaging. J Urol 2011;186:1281-5. [Crossref] [PubMed]
25. Tavakoli AA, Hielscher T, Badura P, Görtz M, Kuder TA, Gnirs R, Schwab C, Hohenfellner M, Schlemmer HP, Bonekamp D. Contribution of Dynamic Contrast-enhanced and Diffusion MRI to PI-RADS for Detecting Clinically Significant Prostate Cancer. Radiology 2023;306:186-99. [Crossref] [PubMed]
26. Peng Y, Jiang Y, Yang C, Brown JB, Antic T, Sethi I, Schmid-Tannwald C, Giger ML, Eggener SE, Oto A. Quantitative analysis of multiparametric prostate MR images: differentiation between prostate cancer and normal tissue and correlation with Gleason score--a computer-aided diagnosis development study. Radiology 2013;267:787-96. [Crossref] [PubMed]
27. Franiel T, Lüdemann L, Taupitz M, Rost J, Asbach P, Beyersdorff D. Pharmacokinetic MRI of the prostate: parameters for differentiating low-grade and high-grade prostate cancer. Rofo 2009;181:536-42. [Crossref] [PubMed]
28. Tosoian JJ, Carter HB, Lepor A, Loeb S. Active surveillance for prostate cancer: current evidence and contemporary state of practice. Nat Rev Urol 2016;13:205-15. [Crossref] [PubMed]
29. Cai JC, Nakai H, Kuanar S, Froemming AT, Bolan CW, Kawashima A, Takahashi H, Mynderse LA, Dora CD, Humphreys MR, Korfiatis P, Rouzrokh P, Bratt AK, Conte GM, Erickson BJ, Takahashi N. Fully Automated Deep Learning Model to Detect Clinically Significant Prostate Cancer at MRI. Radiology 2024;312:e232635. [Crossref] [PubMed]
30. Stabile A, Giganti F, Kasivisvanathan V, Giannarini G, Moore CM, Padhani AR, Panebianco V, Rosenkrantz AB, Salomon G, Turkbey B, Villeirs G, Barentsz JO. Factors Influencing Variability in the Performance of Multiparametric Magnetic Resonance Imaging in Detecting Clinically Significant Prostate Cancer: A Systematic Literature Review. Eur Urol Oncol 2020;3:145-67. [Crossref] [PubMed]
31. Panebianco V, Villeirs G, Weinreb JC, Turkbey BI, Margolis DJ, Richenberg J, Schoots IG, Moore CM, Futterer J, Macura KJ, Oto A, Bittencourt LK, Haider MA, Salomon G, Tempany CM, Padhani AR, Barentsz JO. Prostate Magnetic Resonance Imaging for Local Recurrence Reporting (PI-RR): International Consensus -based Guidelines on Multiparametric Magnetic Resonance Imaging for Prostate Cancer Recurrence after Radiation Therapy and Radical Prostatectomy. Eur Urol Oncol 2021;4:868-76. [Crossref] [PubMed]
32. Stanzione A, Lee KL, Sanmugalingam N, Rajendran I, Sushentsev N, Caglič I, Barrett T. Expect the unexpected: investigating discordant prostate MRI and biopsy results. Eur Radiol 2024;34:4810-20. [Crossref] [PubMed]
33. Ali A, Elumalai T, Venkatesulu B, Hekman L, Mistry H, Sachdeva A, Oliveira P, Clarke N, Baena E, Choudhury A, Bristow RG. Tale of two zones: investigating the clinical outcomes and research gaps in peripheral and transition zone prostate cancer through a systematic review and meta-analysis. BMJ Oncol 2024;3:e000193. [Crossref] [PubMed]