Abstract
Objectives
Reactive stroma plays a pivotal role in the genesis, progression, and metastasis of prostate cancer (PCa). Higher reactive stromal grade (RSG) generally portends a poorer prognosis. The aim of the study is non-invasively evaluate RSG by preoperative mono-exponential model, stretch-exponent model (SEM) and diffusion kurtosis imaging (DKI), and isolate the independent predictor of high RSG (> 50% reactive stroma) in parameters of mono-exponential model, SEM and DKI.
Methods
Totally, 54 low RSG (≤ 50% reactive stroma) patients and 26 high RSG patients were prospectively enrolled in the study. Apparent diffusion coefficient (ADC), mean kurtosis (MK), mean diffusivity (MD), distributed diffusion coefficient (DDC), and heterogeneity index (α) values of all lesions were measured on GE Workstation 4.6. Spearman’s rank correlation analysis was used to analysis the correlation between RSG and parameters of SEM and DKI. Receiver-operating characteristic (ROC) curve analysis was performed to evaluate the diagnostic performance of those parameters in differentiating low RSG and high RSG. DeLong’s test was used to assess whether the differences of AUC for each parameter were statistically significant. Binary logistic regression analysis was performed to identify independent predictors of high RSG.
Results
ADC (r = − 0.352, p = 0.001), DDC (r = − 0.579, p < 0.001) and MD (r = − 0.597, p < 0.001) values showed significant negative correlations with RSG, while MK value (r = 0.658, p < 0.001) demonstrated a significant positive correlation. MK (AUC = 0.816, p < 0.001) was superior to ADC (AUC = 0.717, p < 0.001), DDC (AUC = 0.781, p < 0.001) and MD (AUC = 0.774, p < 0.001) in differentiating low and high RSG, but the differences between these AUCs were not statistically significant (all p > 0.05). Binary logistic regression analysis demonstrated a statistically significant model (χ² =43.222, p < 0.001), and showed that MK (odds ratio = 10.185; 95% CI: 2.467 ~ 21.694; p < 0.001) and MD (odds ratio = 0.014; 95% CI: 0.003 ~ 0.367; p < 0.001) were the independent predictors of high RSG.
Conclusion
Although ADC, DDC, and MD values were significantly negatively correlated with RSG, and MK was significantly positively correlated, and all three models—mono-exponential model, SEM, and DKI—demonstrated good performance in differentiating between low and high RSG, only parameters MD and MK values of DKI were identified as independent predictors of high RSG.
Keywords: Prostate cancer, Reactive stromal grade, Diffusion kurtosis imaging, Stretch-exponent model
Background
Prostate cancer (PCa) is one of the most common malignant tumors among men worldwide [1]. In recent years, the incidence of PCa has been increasing annually, partly due to the widespread use of prostate-specific antigen (PSA) screening [2, 3]. Evidence indicates that the 5-year biochemical recurrence-free survival rate following surgery is significantly higher in prognostic grade group (PGG) 2 (Gleason 3 + 4) patients compared to PGG 3 (Gleason 4 + 3) patients, 88% (95% CI, 85–89) and 63% (95% CI, 61–65), respectively [4, 5].
Reactive stroma, a microenvironment adjacent to epithelial cells, orchestrates a multitude of processes including homeostatic alterations, wound healing, and interactions with neoplastic complexes, playing a pivotal role in the genesis, progression, and metastasis of PCa. The evolving characteristics of reactive stroma during tumorigenesis encompass an augmentation of stromal cells such as fibroblasts and myofibroblasts, a remodeling of the extracellular matrix including collagen fibers and fibronectin, as well as angiogenesis and immune cell infiltration [6]. Reactive stroma grading (RSG) is employed to quantify and evaluate the distribution and abundance of these stromal components, thereby assessing the alterations in the tumor microenvironment and their potential impact on tumor behavior [7, 8]. A higher RSG generally portends a poorer prognosis. Patients with PCa exhibiting high RSG (> 50% reactive stroma) tend to have shorter biochemical recurrence-free survival compared to those with low RSG (≤ 50% reactive stroma) [9]. Currently, there is a lack of research on the non-invasive preoperative assessment of RSG [9].
Diffusion-weighted imaging (DWI), by measuring the diffusion motion of water molecules within biological tissues, indirectly reflects microscopic structural alterations, thereby providing a non-invasive approach for assessing tissue structural characteristics. In the detection and evaluation of PCa, DWI has demonstrated high sensitivity and specificity, particularly showing marked advantages for identifying lesions located in the peripheral zone [10, 11]. However, tumor tissues frequently exhibit diverse cellular arrangements, complex stromal compositions, and structural heterogeneity, resulting in non-Gaussian diffusion behavior of water molecules. Consequently, conventional DWI often fail to adequately reflect this complex microscopic environment. In contrast, advanced diffusion models such as the stretched-exponential model (SEM) and diffusion kurtosis imaging (DKI) can effectively capture non-Gaussian diffusion phenomena, thus providing a more comprehensive characterization of tumor microstructural properties. Several studies have indicated that both SEM and DKI exhibit substantial potential for evaluating PCa aggressiveness, demonstrating promising clinical applicability [12–14]. Parameter distributed diffusion coefficient (DDC) of the SEM can effectively differentiate between benign prostatic hyperplasia and PCa, and DDC value exhibit a significant negative correlation with Gleason pattern in assessing PCa aggressiveness [15]. What’s more, parameters mean diffusivity (MD) and mean kurtosis (MK) of DKI could effectively differentiate PGG 2 and PGG 3 [16].
The aim of the study was to clarify whether RSG differ between PGG 2 and PGG 3. Furthermore, the study attempted to use SEM and DKI to noninvasively evaluate RSG preoperatively, and to compare the diagnostic efficacy of SEM and DKI for high RSG.
Materials and methods
Patients
The study was approved by the ethics committee of the Second Affiliated Hospital of Dalian Medical University. From December 2019 to June 2023, we prospectively enrolled 97 PCa patients, including 65 low RSG patients and 32 high RSG patients. Exclusion criteria: (1) patients underwent prostate biopsy within 6 weeks before prostate MRI (Reducing post biopsy changes, including hemorrhage and inflammation, may adversely affect the interpretation of prostate MRI) (n = 7); (2) radiotherapy, chemotherapy, and endocrine therapy prior to prostate mpMRI (n = 6); (3) the prostate mpMRI have artifacts, which affect the assessment(n = 4). Finally, 80 PCa patients were enrolled in this study, including 54 low RSG patients and 26 high RSG patients.
MRI techniques
Each participant was subjected to a prostate MRI examination utilizing a 3.0 T MRI system (GE Discovery MR 750 W) equipped with an eight-channel phased-array coil. According to Prostate Imaging–Reporting and Data System (PI-RADS) Version 2.1, the prostate MRI scanning sequence included T2-weighted imaging (T2WI), T1-weighted imaging (T1WI), DWI and dynamic contrast-enhanced imaging MRI (DCE-MRI) [11]. SEM imaging parameters were [17]: repetition time (TR) / echo time (TE), 3500/87.5 milliseconds; slice thickness, 3.0 mm; space, 0.5 mm; field of view (FOV), 24 × 24 cm; acquisition matrix, 256 × 224; 11 b values (b values [number of excitations], 0 [1], 25 [1], 50 [1], 75 [1], 100 [4], 150 [4], 200 [6], 400 [8], 800 [10], 1200 [12] and 2000 [14] s/mm2) were employed. Parameters of DKI [16]: TR/TE, 3800/87.4 milliseconds; slice thickness, 3 mm; space, 0.5 mm; FOV, 24 × 24 cm; matrix, 256 × 224; with 15 gradient directions and five b values (b values [number of excitations], 0 [1], 500 [2], 1000 [4], 1500 [6], 2000 [8] s/mm²). Details showed in Table 1.
Table 1.
Acquisition parameters of the multiparametic MRI protocol
| Sequence | Imaging plane | TR/TE (ms) | Slice/Gap (mm) | FOV (cm2) | Matrix | b (s/mm2) |
|---|---|---|---|---|---|---|
| T2WI | Oblique axial, Oblique coronal, sagittal | 3200–3500/95.0 | 3.0/0 | 20 × 20 | 256 × 256 | - |
| T1WI | Oblique axial | 767/9.6 | 3.0/0 | 24 × 24 | 256 × 224 | - |
| DWI | Oblique axial | 3500/73.0 | 3.0/0 | 24 × 24 | 256 × 224 | 0,2000 |
| DCE-MRI | Oblique axial | 3.8/1.4 | 3.0/0 | 24 × 24 | 256 × 224 | |
| SEM | Oblique axial | 3500/87.5 | 3.0/0.5 | 24 × 24 | 256 × 224 | 0,25,50,75,100,150,200, 400,800,1200,2000 |
| DKI | Oblique axial | 3800/87.4 | 3.0/0.5 | 24 × 24 | 256 × 224 | 0, 500, 1000, 1500, 2000 |
DWI, diffusion weighted imaging; DCE-MRI, dynamic contrast-enhanced imaging MRI; SEM, stretched-exponential model; DKI, diffusion kurtosis imaging; TR, repetition time; TE, echo time; FOV, field of view
Imaging and quantitative data analysis
All prostate MRI assessments adhered to the recommendations of PI-RADS version 2.1 [11]. Two radiologists (DM. L., with 10 years of experience in abdominal imaging diagnosis; KP. Z., with 7 years of experience in abdominal imaging diagnosis) jointly evaluated the prostate MRI while blinded to the patients’ laboratory data and results from ancillary imaging examination results (such as ultrasound). In cases where there was a discrepancy in the diagnostic conclusions, a third senior radiologist (QY. L., with more than 15 years of experience in abdominal imaging) was consulted to reach a final consensus.
The parameter maps of SEM and DKI were reconstructed on a GE AW 4.6 workstation. Firstly, region of interest (ROI) was drawn along the lesion outline at the largest slice on T2WI or apparent diffusion coefficient (ADC) map (If the lesion was located in peripheral zone, ADC map was referred. If the lesion was located in transitional zone, T2WI was referenced). Subsequently, the ROI was copied onto the related parameter maps of SEM and DKI for measurement.
Pathology acquisition and analysis
All patients underwent radical prostatectomy performed by experienced urologic surgeons. Following surgery, the excised prostate specimens containing prostate cancer were immediately fixed in formalin. After paraffin embedding, serial sections were obtained at 3 mm intervals from the apex to the base of the prostate. All sections were subsequently stained with hematoxylin and eosin (H&E) for histopathological evaluation. To ensure optimal correspondence between the imaging-based ROI measurement slice and the histopathological evaluation slice, the following procedure was used for RSG assessment. First, the largest cross-sectional area of the lesion was identified on axial T2WI or ADC map, and a ROI was delineated on that slice. Next, the vertical distance from this slice to the apex of the prostate was measured on sagittal T2WI and recorded as D (in millimeters). Finally, based on the measured distance D and the pathological sectioning interval of 3 mm, the corresponding histological slice number was estimated to identify the appropriate section for RSG evaluation.
Pathological assessment of each specimen was independently conducted by two pathologists (both K. G. and J. X., with more than 10 years of experience in pathological diagnosis), who blind to the magnetic resonance imaging outcomes. Discrepancies between the evaluations were reconciled through consensus. The reactive stroma appears pale pink, exhibits fibrosis, and has a disorganized arrangement, with stromal cells characterized by elongated nuclei and scant cytoplasm. RSG was classified into four grades: Grade 0 (0% ~ 5% reactive stroma), Grade 1 (6% ~ 15% reactive stroma), Grade 2 (16% ~ 50% reactive stroma), and Grade 3 (51% ~ 100% reactive stroma or a gland-to-stroma ratio of at least 1:1) [9, 18]. Patients were divided into high RSG group (> 50% reactive stroma) and low RSG group (≤ 50% reactive stroma).
Statistical analysis
In the study, the normality and homogeneity of continuous variances were assessed using the Shapiro-Wilk test and Levene’s test, respectively. Clinical characteristics of patients were compared between low RSG and high RSG, continuous variances were compared by independent sample t-test, and categorical variances were compared by Fisher’s exact test or χ2 test. Independent sample t-test was used to analyze the differences of parameters of mono-exponential model, SEM and DKI between between low RSG and high RSG. To analysis the correlation between RSG and parameters of mono-exponential model, SEM and DKI, Spearman’s rank correlation analysis was used. To evaluate diagnostic performance and identify optimal cut-off values for the prediction of high RSG, receiver operating characteristics (ROC) curve analysis was used and the area under the curve (AUC), sensitivity, and specificity were also calculated. Subsequently, the DeLong’s test was used to assess whether the differences of AUC for each parameter were statistically significant. Binary logistic regression analysis was performed to identify independent predictors of high RSG from parameters derived from the mono-exponential model, SEM, and DKI. All statistical analysis was performed using SPSS software (v. 27.0, SPSS, Chicago, IL) and Medcalc (V. 22.0; MedCalc Software, Mariakerke, Belgium). p value < 0.05 was considered statistically significant.
Results
Patient characteristics
A total of 80 PCa patients were included in the study, among which 54 were low RSG patients (Fig. 1), and 26 were high RSG patients (Fig. 2). In PGG 2, 29 cases were classified as low RSG, while 7 cases were high RSG. In contrast, PGG 3 included 25 cases with low RSG and 19 cases with high RSG. Chi-square analysis revealed a statistically significant difference in the distribution of RSG between PGG 2 and PGG 3 (p = 0.024), indicating a potential association between higher RSG and more advanced PGG. The detailed clinical characteristics are shown in Table 2.
Fig. 1.
A 76-year-old prognostic grade group (PGG) 2 patient, PSA of 10.32 ng/ml, Gleason 3 + 4. (a-c) The lesion with PI-RADS 4 seen in the left middle peripheral zone. (d) Mean diffusion (MD) of 0.887 × 10− 3 mm2/s. (e) Mean kurtosis (MK) of 0.554. (f) Distribute diffusion coefficient (DDC) of 0.701 × 10− 3 mm2/s
Fig. 2.
A 79-year-old prognostic grade group (PGG) 3 patient, PSA of 16.32 ng/ml, Gleason 4 + 3. (a-c) The lesion with PI-RADS 5 seen in the left middle peripheral zone. (d) Mean diffusion (MD) of 0.758 × 10− 3 mm2/s. (e) Mean kurtosis (MK) of 0.652. (f) Distribute diffusion coefficient (DDC) of 0.579 × 10− 3 mm2/s
Table 2.
Clinical characteristics of the patients
| Variables | RSG | p | |
|---|---|---|---|
| Low RSG (n = 54) | High RSG (n = 26) | ||
| Age (year) | 77.56 ± 8.41 | 77.42 ± 7.83 | 0.946 |
| T-PSA (ng/ml) | 11.08 ± 2.73 | 12.33 ± 2.51 | 0.054 |
| F-PSA (ng/ml) | 1.71 ± 0.29 | 1.74 ± 0.25 | 0.602 |
| F/T-PSA (ng/ml) | 0.16 ± 0.03 | 0.15 ± 0.03 | 0.067 |
| PI-RADS category | 0.096 | ||
| 2 | 1 (1.25) | 2 (2.50) | |
| 3 | 7 (8.75) | 1 (1.25) | |
| 4 | 20 (25.00) | 5 (6.25) | |
| 5 | 26 (32.50) | 18 (22.50) | |
| lesion location | 0.270 | ||
| TZ | 20 (25.00) | 13 (16.25) | |
| PZ | 34 (42.50) | 13 (16.25) | |
| PGG | 0.024 | ||
| 2 | 29 (36.25) | 7 (8.75) | |
| 3 | 25 (31.25) | 19 (23.75) | |
PGG, prognostic grade group; T-PSA, total prostate specific antigen; F-PSA, free PSA; PI-RADS, Prostate Imaging–Reporting and Data System; RSG, reactive stromal grade; TZ, transition zone; PZ, peripheral zone
Correlation analysis of mono‑exponential model, SEM and DKI with RSG
Spearman’s rank correlation analysis showed that MK value (r = 0.658, p < 0.001) had a moderate positive correlation with RSG, whereas MD value (r = − 0.597, p < 0.001) and DDC value (r = − 0.579, p < 0.001) had moderate negative correlations with RSG. ADC value (r = − 0.352, p = 0.001) had low negative correlations with RSG. However, 𝛼 value (r = 0.037, p = 0.743) only exhibited a low positive correlation with RSG, which was not statistically significant.
Comparison of mono-exponential model, SEM, and DKI parameters between low and high RSG groups
Compared with low RSG PCa, high RSG PCa had a higher MK value (0.567 ± 0.079 vs. 0.656 ± 0.061), had a lower MD value (0.873 ± 0.128 × 10-³ mm²/s vs. 0.752 ± 0.077 × 10-³ mm²/s), DDC value (0.693 ± 0.095 × 10-³ mm²/s vs. 0.603 ± 0.071 × 10-³ mm²/s) and ADC value (1.064 ± 0.176 × 10-³ mm²/s vs. 0.926 ± 0.159 × 10-³ mm²/s). Results of independent samples t-test showed that the differences in MK (t = − 5.021, p < 0.001), MD (t = 4.442, p < 0.001), DDC (t = 4.257, p < 0.001) and ADC (t = 3.395, p = 0.001) values between low RSG and high RSG were statistically significant. However, the difference in 𝛼 values between low RSG (0.565 ± 0.199) and high RSG (0.595 ± 0.201) have no statistically significance (p = 0.541). The details are shown in Table 3.
Table 3.
Independent samples t-Test results for SEM and DKI parameters between low RSG and high RSG PCa
| Parameters | Low RSG | High RSG | F | t | p |
|---|---|---|---|---|---|
| MK | 0.567 ± 0.079 | 0.656 ± 0.061 | 2.340 | -5.021 | < 0.001 |
| MD (×10− 3mm2/s) | 0.873 ± 0.128 | 0.752 ± 0.077 | 9.328 | 4.442 | < 0.001 |
| DDC (×10− 3mm2/s) | 0.693 ± 0.095 | 0.603 ± 0.071 | 2.243 | 4.257 | < 0.001 |
| ADC (×10− 3mm2/s) | 1.064 ± 0.176 | 0.926 ± 0.159 | 0.032 | 3.395 | 0.001 |
| 𝛼 | 0.565 ± 0.199 | 0.595 ± 0.201 | 0.425 | -0.614 | 0.541 |
SEM, stretched-exponential model; DKI, diffusion kurtosis imaging; RSG, reactive stromal grade; PCa, prostate cancer; MK, mean kurtosis; MD, mean diffusivity; DDC, distributed diffusion coefficient; ADC, apparent diffusion coefficient; 𝛼, heterogeneity index
Diagnostic performance of Mono-exponential model, SEM, and DKI in differentiating low and high RSG PCa
Results of ROC curve analysis showed that MK, MD, DDC and ADC values could effectively differentiate high RSG and low RSG, AUCMK=0.816 (p < 0.001;), AUCMD=0.774 (p < 0.001), AUCDDC=0.781 (p < 0.001) and AUCADC=0.717 (p < 0.001), respectively. When MK value ≥ 0.638, the sensitivity and specificity were 69.23% and 81.48%, respectively. For MD value ≤ 0.852 × 10-³ mm²/s, the sensitivity and specificity were 91.15% and 55.56%, respectively. When the cut-off for the DDC value was set at 0.657 × 10-³ mm²/s, the sensitivity and specificity were 84.62% and 66.67%, respectively. When the cut-off for the ADC value was set at 0.855 × 10-³ mm²/s, the sensitivity and specificity were 46.15% and 88.89%, respectively. The details are shown in Fig. 3 and Table 4. Although MK value had a higher AUC compared to MD value, DDC value and ADC value, but the differences have no statistically significance (all p > 0.05). The details are shown in Table 5.
Fig. 3.
Receiver operating characteristic curves of mean kurtosis (MK) (a), mean diffusion (MD) (b), distribute diffusion coefficient (DDC) (c) and combined parameters (d) for the diagnosis ability of low reactive stromal grade and high reactive stromal grade
Table 4.
Diagnostic performance of Mono-exponential model, SEM, and DKI in differentiating low and high RSG PCa
| Parameters | AUC | Youden Index | Cut-off Value | Sensitivity (%) | Specificity (%) | 95% CI | p |
|---|---|---|---|---|---|---|---|
| MK | 0.816 | 0.507 | 0.638 | 69.23 | 81.48 | 0.713 ~ 0.893 | < 0.001 |
| MD | 0.774 | 0.517 | 0.852 (10-³ mm²/s) | 91.15 | 55.56 | 0.666 ~ 0.860 | < 0.001 |
| DDC | 0.781 | 0.513 | 0.657 (10-³ mm²/s) | 84.62 | 66.67 | 0.674 ~ 0.865 | < 0.001 |
| ADC | 0.717 | 0.350 | 0.855 (10-³ mm²/s) | 46.15 | 88.89 | 0.605 ~ 0.812 | < 0.001 |
SEM, stretched-exponential model; DKI, diffusion kurtosis imaging; RSG, reactive stromal grade; PCa, prostate cancer; MK, mean kurtosis; MD, mean diffusivity; DDC, distributed diffusion coefficient; ADC, apparent diffusion coefficient; AUC, area under the ROC curve
Table 5.
Results of Delong test
| Parameters | Difference between AUC | Z statistic | 95% CI | p |
|---|---|---|---|---|
| AUCMK vs. AUCMD | 0.042 ± 0.058 | 0.731 | -0.071 ~ 0.155 | 0.465 |
| AUCMK vs. AUCDDC | 0.035 ± 0.061 | 0.577 | -0.084 ~ 0.153 | 0.564 |
| AUCMK vs. AUCADC | 0.099 ± 0.060 | 1.640 | -0.019 ~ 0.217 | 0.101 |
| AUCMD vs. AUCDDC | 0.007 ± 0.060 | 0.119 | -0.111 ~ 0.125 | 0.906 |
| AUCMD vs. AUCADC | 0.057 ± 0.065 | 0.871 | -0.071 ~ 0.184 | 0.384 |
| AUCDDC vs. AUCADC | 0.064 ± 0.067 | 0.930 | -0.071 ~ 0.198 | 0.352 |
MK, mean kurtosis; MD, mean diffusivity; DDC, distributed diffusion coefficient; ADC, apparent diffusion coefficient; AUC, area under the ROC curve
Binary logistic regression analysis of DWI-Based parameters for predicting high RSG
Binary logistic regression analysis demonstrated a statistically significant model (χ² =43.222, p < 0.001) with an overall accuracy of 80.00% in predicting high RSG. The model yielded a sensitivity of 70.83%, specificity of 83.93%, positive predictive value of 65.39%, and negative predictive value of 87.04%. Among the evaluated parameters, MK (odds ratio = 10.185; 95% CI: 2.467 ~ 21.694; p < 0.001) and MD (odds ratio = 0.014; 95% CI: 0.003 ~ 0.367; p < 0.001) were identified as independent predictors of high RSG. The details are shown in Table 6.
Table 6.
Results of binary logistic regression analysis
| Parameters | β (SE) | Wald χ² | df | p | OR | 95%CI |
|---|---|---|---|---|---|---|
| MK | 2.321(0.555) | 17.52 | 1 | < 0.001 | 10.185 | 2.467 ~ 21.694 |
| MD | -4.269(1.226) | 12.13 | 1 | < 0.001 | 0.014 | 0.003 ~ 0.367 |
| DDC | -1.398(0.717) | 3.799 | 1 | 0.051 | 0.247 | 0.039 ~ 0.971 |
| 𝛼 | 1.468(1.055) | 1.938 | 1 | 0.164 | 4.341 | 0.337 ~ 19.684 |
| ADC | -1.127(0.665) | 2.873 | 1 | 0.090 | 0.324 | 0.093 ~ 0.967 |
| T-PSA | -0.128(0.670) | 0.036 | 1 | 0.849 | 0.880 | 0.237 ~ 3.273 |
| F-PSA | -3.817(4.743) | 0.648 | 1 | 0.421 | 0.022 | 0.007 ~ 0.193 |
| F/T-PSA | 0.913(1.582) | 0.333 | 1 | 0.563 | 2.492 | 0.467 ~ 7.372 |
| PGG | 0.922(1.043) | 0.782 | 1 | 0.377 | 2.515 | 0.326 ~ 19.433 |
| Lesion Location | -0.518(0.761) | 0.463 | 1 | 0.496 | 0.596 | 0.134 ~ 2.650 |
MK, mean kurtosis; MD, mean diffusivity; DDC, distributed diffusion coefficient; ADC, apparent diffusion coefficient; 𝛼, heterogeneity index; T-PSA, total prostate specific antigen; F-PSA, free PSA; PGG, prognostic grade group; OR, odds ratio
Discussion
The study demonstrated the role of quantitative mono-exponential model, SEM and DKI parameters for intratumor RSG in PGG 2 and PGG 3 prostate cancer patients. In particular, ADC, DDC, MK, and MD values and their combinations were able to distinguish high and low RSG groups. However, only MK and MD values were independent risk factors for high RSG.
Previous studies have demonstrated that both the parameter DDC value of SEM and the parameter MD value of DKI exhibit similarities to the ADC value of the mono-exponential model, and they are negatively correlated with the Gleason pattern of PCa. In contrast, MK value, which reflects the non-Gaussian behavior of water diffusion within tissues, exhibits a positive correlation with tumor aggressiveness [10, 15, 19]. Our findings are consistent with these studies: compared to low RSG group, high RSG group exhibited a significantly lower ADC, DDC, and MD values, and a higher MK value.
As RSG increases, the stromal reactivity becomes more pronounced [7, 8], leading to increased diffusion restriction and larger diffusion peak of water molecule within the tumor tissue. Furthermore, results of Hectors et al. [12] showed that, key contributors to restricted water diffusion in tissues include nuclear-to-cytoplasmic ratio, cell density, and stromal complexity. This may partly explain why high RSG is associated with lower DDC and MD values and higher MK value compared to low RSG. Previous studies have shown that with an increase of Gleason pattern, the volume of low-diffusion epithelial cells increases, while the volume of high-diffusion stroma and lumens decreases [20, 21]. This results in lower DDC and MD values and higher MK value in PGG 3 tumor compared to PGG 2 tumor. Notably, in our study, the proportion of PGG 3 in high RSG PCa was significantly higher than that in low RSG PCa. This may provide another explanation for the observed lower DDC and MD values and higher MK value in the high RSG compared to the low RSG.
RSG is a key factor in PCa progression and resistance, and higher RSG is associated with increased biochemical recurrence rate [7, 9, 22]. Our findings showed that the feasibility of using mono-exponential model, SEM and DKI for non-invasive evaluation of RSG. Although MK demonstrated the highest AUC (0.816) compared to ADC, DDC, and MD, the differences in diagnostic performance among these parameters for identifying high RSG were not statistically significant. Some studies have reported that, compared to the mono-exponential model, neither SEM nor DKI provides additional diagnostic value in assessing the Gleason score of PCa [10, 13–16]. However, conflicting evidence exists; such as, studies by Fukunaga and Ding et al. [17, 23]. have suggested that SEM and DKI offer superior diagnostic performance in this context.
The reactive stroma in PCa primarily consists of tumor-associated fibroblasts, immune cells, extracellular matrix, and blood vessels within the tumor microenvironment. Cancer-associated fibroblasts play a critical role in tumor stromal remodeling by promoting fibrosis and enhancing the accumulation of collagen and glycosaminoglycans, which increases the complexity of the tumor stroma [24–26]. These histological changes lead to a deviation from Gaussian diffusion behavior within prostate cancer tissues. Binary logistic regression analysis revealed that parameters MK and MD values of DKI were independent predictors of high RSG. In contrast, parameters of SEM and mono-exponential model were not statistically significant.
Notably, our study observed no significant correlation between the parameter α value of SEM and RSG, with no statistically significant differences detected between low RSG and high RSG. Although previous studies have demonstrated the diagnostic value of α value in distinguishing benign prostatic hyperplasia (BPH) from PCa [27], this may be attributed to the more complex microstructural characteristics of PCa tissue, such as higher tumor cellularity, reduced glandular lumen structure, and compressed extracellular spaces. However, several studies have reported that a lack of significant association between the α value and Gleason pattern [15], and our results keep line with the previous studies. This suggests that the microstructural differences reflected by both the Gleason system and the RSG may not have reached the sensitivity threshold required for the α value detection.
There are some limitations in our study. Firstly, this was a single-center retrospective study with a relatively small sample size, which may limit the generalizability of the findings. Further validation using multicenter datasets with larger cohorts is warranted. Secondly, an imbalance in sample size between the low and high RSG groups may have reduced the statistical power of the analysis and increased the risk of model overfitting. Thirdly, given the heterogeneous nature of RSG, histogram- or texture-based imaging feature analysis may offer greater diagnostic value than mean value–based measurements. However, due to limitations in research conditions, we were unable to perform such advanced analyses in this study and instead relied on mean parameter extraction. This methodological constraint may have limited the comprehensiveness of our imaging-based evaluation of RSG. Fourthly, Due to the small volume of some lesions, which were visible on only a single axial slice of the T2WI or ADC map, it was not feasible to accurately delineate their full 3D volume. To ensure methodological consistency across the entire cohort, ROI analysis was performed on the axial slice showing the largest cross-sectional area of the lesion. However, this single-slice approach may have limited our ability to fully capture the spatial heterogeneity of the tumor and its associated stroma.
Conclusions
The study demonstrated that ADC, DDC, and MD values were significantly negatively correlated with RSG, while MK was significantly positively correlated. Although mono-exponential model, SEM, and DKI all showed good performance in distinguishing between low and high RSG, only parameters MD and MK values of DKI were identified as independent predictors of high RSG. These findings suggest that DKI provides superior diagnostic performance compared to the mono-exponential model and SEM, and may serve as a promising noninvasive approach for assessing RSG in PCa.
Acknowledgements
Thanks are due to Jia-Wen Luo for assistance with the experiments.
Abbreviations
- PCa
Prostate cancer
- PGG
Prognostic grade group
- PSA
Prostate-specific antigen
- RSG
Reactive stromal grade
- DWI
Diffusion-weighted imaging
- SEM
Stretched-exponential model
- DKI
Diffusion kurtosis imaging
- DDC
Distributed diffusion coefficient
- α
Heterogeneity index
- MD
Mean diffusivity
- MK
Value and mean kurtosis
- TR
Repetition time
- TE
Echo time
- FOV
Field of view
- ADC
Apparent diffusion coefficient
- ROI
Region of interest
- ROC
Receiver operating characteristics
- AUC
Area under the curve
Author contributions
Conceptualization, QY L and KP Z; Data curation, HB H, SY L and DM L; Formal analysis, DM L; Investigation, HB H, SY L and KP Z; Methodology, QY L, KP Z and DM L; Project administration, QY L and KP Z; Software, HB H and SY L; Supervision, QY L and KP Z; Validation, ZX L and XW C; Visualization, SY L; Writing – original draft, KP Z, HB H and SY L; Writing – review & editing, QY L, JB, DM L. All authors reviewed the manuscript.
Funding
The authors did not receive support from any organization for the submitted work.
Data availability
The data that support the findings of this study are available from the authors but restrictions apply to the availability of these data, which were used under license from the Second Affiliated Hospital of Dalian Medical University for the current study, and so are not publicly available. Data are, however, available from the authors upon reasonable request and with permission from the Second Affiliated Hospital of Dalian Medical University.
Declarations
Ethics approval and consent to participate
The study was approved by the Institutional Ethics Committee of the Second Affiliated Hospital of Dalian Medical University (PR/AG- 166/2021, approved on 4 November 2021), and in accordance with the Declaration of Helsinki. Informed consent was obtained from all subjects involved in the study.
Consent for publication
Written informed consent for publication was obtained from all participants.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Jie Bian and Di-Min Liu co-corresponding authors and contributed equally to this work.
Kun-Peng Zhou, Hua-Bin Huang and Shu-Yi Li contributed equally to this work and share first authorship.
Contributor Information
Di-Min Liu, Email: liudimin@sysush.com.
Jie Bian, Email: bianjie@163.com.
Qing-Yu Liu, Email: liuqingyu@sysush.com.
References
- 1.Siegel RL, Giaquinto AN, Jemal A. Cancer statistics, 2024. CA Cancer J Clin. 2024;74(1):12–49. [DOI] [PubMed] [Google Scholar]
- 2.Schaeffer EM, Srinivas S, Adra N, An Y, Barocas D, Bitting R, et al. Prostate cancer, version 4.2023, NCCN clinical practice guidelines in oncology. J Natl Compr Canc Netw. 2023;21(10):1067–96. [DOI] [PubMed] [Google Scholar]
- 3.Ward EM, Sherman RL, Henley SJ, Jemal A, Siegel DA, Feuer EJ et al. Annual Report to the Nation on the Status of Cancer, Featuring Cancer in Men and Women Age 20–49 Years. J Natl Cancer Inst. 2019;111(12):1279-97. [DOI] [PMC free article] [PubMed]
- 4.Epstein JI, Zelefsky MJ, Sjoberg DD, Nelson JB, Egevad L, Magi-Galluzzi C, et al. A contemporary prostate cancer grading system: A validated alternative to the Gleason score. Eur Urol. 2016;69(3):428–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Kamel MH, Khalil MI, Alobuia WM, Su J, Davis R. Incidence of metastasis and prostate-specific antigen levels at diagnosis in Gleason 3 + 4 versus 4 + 3 prostate cancer. Urol Ann. 2018;10(2):203–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Sandberg TP, Stuart M, Oosting J, Tollenaar R, Sier CFM, Mesker WE. Increased expression of cancer-associated fibroblast markers at the invasive front and its association with tumor-stroma ratio in colorectal cancer. BMC Cancer. 2019;19(1):284. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Ding Y, Bu P, Assylbekova B, Ruder S, Miles B, Sayeeduddin M, et al. Quantification of collagen content and stromal cellularity within reactive stroma is predictive of prostate cancer biochemical recurrence and specific death. Hum Pathol. 2024;144:1–7. [DOI] [PubMed] [Google Scholar]
- 8.Doldi V, Lecchi M, Ljevar S, Colecchia M, Campi E, Centonze G et al. Potential of the stromal matricellular protein Periostin as a biomarker to improve risk assessment in prostate cancer. Int J Mol Sci. 2022;23(14). [DOI] [PMC free article] [PubMed]
- 9.Nastaly P, Smentoch J, Popeda M, Martini E, Maiuri P, Zaczek AJ et al. Low Tumor-to-Stroma ratio reflects protective role of stroma against prostate cancer progression. J Pers Med. 2021;11(11). [DOI] [PMC free article] [PubMed]
- 10.Shan Y, Chen X, Liu K, Zeng M, Zhou J. Prostate cancer aggressive prediction: preponderant diagnostic performances of intravoxel incoherent motion (IVIM) imaging and diffusion kurtosis imaging (DKI) beyond ADC at 3.0 T scanner with Gleason score at final pathology. Abdom Radiol (NY). 2019;44(10):3441–52. [DOI] [PubMed] [Google Scholar]
- 11.Turkbey B, Rosenkrantz AB, Haider MA, Padhani AR, Villeirs G, Macura KJ, et al. Eur Urol. 2019;76(3):340–51. Prostate Imaging Reporting and Data System Version 2.1: 2019 Update of Prostate Imaging Reporting and Data System Version 2. [DOI] [PubMed]
- 12.Hectors SJ, Semaan S, Song C, Lewis S, Haines GK, Tewari A, et al. Advanced Diffusion-weighted imaging modeling for prostate cancer characterization: correlation with quantitative histopathologic tumor tissue Composition-A Hypothesis-generating study. Radiology. 2018;286(3):918–28. [DOI] [PubMed] [Google Scholar]
- 13.Liu W, Liu XH, Tang W, Gao HB, Zhou BN, Zhou LP. Histogram analysis of stretched-exponential and monoexponential diffusion-weighted imaging models for distinguishing low and intermediate/high Gleason scores in prostate carcinoma. J Magn Reson Imaging. 2018;48(2):491–8. [DOI] [PubMed] [Google Scholar]
- 14.Tamada T, Prabhu V, Li J, Babb JS, Taneja SS, Rosenkrantz AB. Prostate cancer: Diffusion-weighted MR imaging for detection and assessment of Aggressiveness-Comparison between conventional and kurtosis models. Radiology. 2017;284(1):100–8. [DOI] [PubMed] [Google Scholar]
- 15.Kim E, Kim CK, Kim HS, Jang DP, Kim IY, Hwang J. Histogram analysis from stretched exponential model on diffusion-weighted imaging: evaluation of clinically significant prostate cancer. Br J Radiol. 2020;93(1106):20190757. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Park H, Kim SH, Lee Y, Son JH. Comparison of diagnostic performance between diffusion kurtosis imaging parameters and mono-exponential ADC for determination of clinically significant cancer in patients with prostate cancer. Abdom Radiol (NY). 2020;45(12):4235–43. [DOI] [PubMed] [Google Scholar]
- 17.Fukunaga T, Tamada T, Kanomata N, Takeuchi M, Ueda Y, Miyaji Y, et al. Quantitative diffusion-weighted imaging and dynamic contrast-enhanced MR imaging for assessment of tumor aggressiveness in prostate cancer at 3T. Magn Reson Imaging. 2021;83:152–9. [DOI] [PubMed] [Google Scholar]
- 18.Epstein JI, Amin MB, Reuter VE, Humphrey PA. Contemporary Gleason grading of prostatic carcinoma: an update with discussion on practical issues to implement the 2014 international society of urological pathology (ISUP) consensus conference on Gleason grading of prostatic carcinoma. Am J Surg Pathol. 2017;41(4):e1–7. [DOI] [PubMed]
- 19.Rajabi P, Rezakhaniha B, Galougahi MHK, Mohammadimehr M, Sharifnia H, Pakzad R, et al. Unveiling the diagnostic potential of diffusion kurtosis imaging and intravoxel incoherent motion for detecting and characterizing prostate cancer: a meta-analysis. Abdom Radiol (NY). 2025;50(1):319–35. [DOI] [PubMed]
- 20.Kobus T, van der Laak JA, Maas MC, Hambrock T, Bruggink CC, Hulsbergen-van de Kaa CA, et al. Contribution of histopathologic tissue composition to quantitative MR spectroscopy and Diffusion-weighted imaging of the prostate. Radiology. 2016;278(3):801–11. [DOI] [PubMed] [Google Scholar]
- 21.Chatterjee A, Watson G, Myint E, Sved P, McEntee M, Bourne R. Changes in epithelium, stroma, and lumen space correlate more strongly with Gleason pattern and are stronger predictors of prostate ADC changes than cellularity metrics. Radiology. 2015;277(3):751–62. [DOI] [PubMed] [Google Scholar]
- 22.De Vivar AD, Sayeeduddin M, Rowley D, Cubilla A, Miles B, Kadmon D, et al. Histologic features of stromogenic carcinoma of the prostate (carcinomas with reactive stroma grade 3). Hum Pathol. 2017;63:202–11. [DOI] [PubMed] [Google Scholar]
- 23.Ding K, Yao Y, Gao Y, Lu X, Chen H, Tang Q, et al. Diagnostic evaluation of diffusion kurtosis imaging for prostate cancer: detection in a biopsy population. Eur J Radiol. 2019;118:138–46. [DOI] [PubMed] [Google Scholar]
- 24.Bedeschi M, Marino N, Cavassi E, Piccinini F, Tesei A. Cancer-associated fibroblast: role in prostate cancer progression to metastatic disease and therapeutic resistance. Cells. 2023;12(5). [DOI] [PMC free article] [PubMed]
- 25.Yu S, Jiang Y, Wan F, Wu J, Gao Z, Liu D. Immortalized cancer-associated fibroblasts promote prostate cancer carcinogenesis, proliferation and invasion. Anticancer Res. 2017;37(8):4311–8. [DOI] [PubMed] [Google Scholar]
- 26.Bordanaba-Florit G, Royo F, Alboniga OE, Clayton A, Falcon-Perez JM, Webber J. Integration of proteomic and metabolomic analysis reveal distinct metabolic alterations of prostate cancer-associated fibroblasts compared to normal fibroblasts from patient’s stroma samples. Biochim Biophys Acta Mol Basis Dis. 2024;1870(6):167229. [DOI] [PubMed] [Google Scholar]
- 27.Liu X, Zhou L, Peng W, Wang H, Zhang Y. Comparison of stretched-Exponential and monoexponential model diffusion-Weighted imaging in prostate cancer and normal tissues. J Magn Reson Imaging. 2015;42(4):1078–85. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The data that support the findings of this study are available from the authors but restrictions apply to the availability of these data, which were used under license from the Second Affiliated Hospital of Dalian Medical University for the current study, and so are not publicly available. Data are, however, available from the authors upon reasonable request and with permission from the Second Affiliated Hospital of Dalian Medical University.