Risk Stratification of Prostate Cancer: Preoperatively Assessing Aggressiveness of Favorable and Unfavorable Intermediate-Risk Groups by Advanced Diffusion-Weighted Imaging

Abstract

Background

The treatment decisions and prognostic outcomes have distinct differences between unfavorable intermediate-risk group (un-FIRG) prostate carcinoma (PCa) and favorable intermediate-risk group (FIRG) PCa. The study aimed to differentiate un-FIRG and FIRG by using advanced diffusion-weighted imaging (DWI).

Methods

From December 2018 to January 2023, 51 FIRG patients and 67 un-FIRG patients were enrolled in our study. All enrolled PCa patients underwent diffusion kurtosis imaging (DKI) and intravoxel incoherent motion (IVIM) imaging and stretched exponential model using a 3.0-T system. The statistical tests included independent samples t-test and binary logistic regression analysis.

Results

Results showed that fractional anisotropy (FA) value and mean kurtosis (MK) value of the un-FIRG (p < 0.001, p < 0.001) were significantly higher than those of the FIRG. On the contrary, mean diffusion (MD), standard apparent diffusion coefficient (ADC), diffusion coefficient (D), and distribute diffusion coefficient (DDC) of the un-FIRG (p = 0.013, p < 0.001, p < 0.001, p < 0.001) were significantly lower than those of the FIRG. Results of binary logistic regression analysis showed that the diagnostic model was statistically significant, chi-square = 90.969, p < 0.001, which could effectively distinguish the 86.40% of un-FIRG in the intermediate-risk group. The results of receiver operating characteristic analysis showed that the area under the curve was 0.9429. Sensitivity and specificity were 88.06 and 84.31%.

Conclusion

Compared with the FIRG group PCa, the un-FIRG group PCa exhibits lower standard ADC, D, DDC, and MD values, as well as higher FA and MK values. This suggests that preoperative advanced DWI may serve as an imaging biomarker for distinguishing between the FIRG group and un-FIRG group PCa.

Relevance Statement

Combining preoperative advanced DWI (DKI, IVIM imaging, and the stretched exponential model) shows potential for noninvasive subrisk stratification of intermediate-risk group PCa.

Keywords

prostate cancer - intermediate-risk group - diffusion kurtosis imaging - intravoxel incoherent motion imaging - stretched exponential model

Key Points

• Advanced diffusion-weighted imaging could subrisk stratify intermediate-risk group prostate cancer (PCa).

• Advanced diffusion-weighted imaging can effectively predict unfavorable intermediate-risk group PCa.

• Advanced diffusion-weighted imaging may be an imaging biomarker for PCa clinical decision-making.

Introduction

Prostate carcinoma (PCa) ranks among the more prevalent neoplasms in males, exhibiting a frequency that escalates with advancing age.[1] Although PCa is often slow to progress and is potentially curable, its implications from either excessive or insufficient treatment position it as a major worldwide cause of oncological impairments and a critical factor in mortality among males due to cancer.[2] [3] Selecting the most appropriate therapeutic strategy for a particular patient is complex and demands evaluation of the patient's overall health status, the tumor's properties, adverse effects from potential treatments, evidence from clinical trials with comparable patient cohorts, and forecasts of their prospective outcomes. This complexity is further intensified by the absence of readily accessible prognostic instruments that could more effectively stratify patients by risk.

One of the most common systems used to risk-stratify patients worldwide is the National Comprehensive Cancer Network (NCCN) guidelines of PCa, which uses a minimum of stage, Gleason grade, and serum prostate-specific antigen (PSA) to assign patients to risk groups.[4] This three-tier system forms the basis of treatment recommendations used for localized prostate cancer throughout the world and some studies have shown that this system provides better treatment recommendations and prognosis prediction results than guided by the clinical stage or Gleason grade system alone.[5] [6] Results of Zumsteg et al[7] showed that compared with the unfavorable intermediate-risk group (un-FIRG), the favorable intermediate-risk group (FIRG) have better prognosis, the un-FIRG have lower PSA recurrence-free survival, higher rates of distant metastasis, and PCa-specific mortality than the FIRG.

Multiparametric magnetic resonance imaging (mpMRI) has become a common method for screening, diagnosing, and monitoring PCa. In recent years, some studies have attempted to differentiate PCa with Gleason pattern ≤ 3 + 4 and ≥ 3 + 4 by diffusion-weighted imaging (DWI) and advanced DWI.[8] [9] [10] [11] Tamada et al[12] studied 457 PCa patients and results showed that the apparent diffusion coefficient (ADC) value of PCa with Gleason pattern ≥ 4 + 3 was significantly lower than those of PCa with Gleason pattern ≤ 3 + 4. Shan et al[13] used diffusion kurtosis imaging (DKI) and intravoxel incoherent motion imaging (IVIM) to study 121 PCa patients and results showed that both DKI and IVIM have the capability to differentially diagnose PCa between Gleason pattern ≤ 3 + 4 and Gleason pattern ≥ 3 + 4. To our knowledge, there is currently no research that has attempted to combine IVIM, stretched exponential model, and DKI to distinguish NCCN risk stratification of PCa.

The aim of this study was to investigate the feasibility and power of the combination of DKI, IVIM, and stretched exponential model in the differentiation of PCa between the FIRG and the un-FIRG.

Materials and Methods

Patients

This study was approved by the ethics committee of the Second Hospital of Dalian Medical University, Dalian, People Republic of China. A total of 372 PCa patients underwent prostate mpMRI from December 2018 to January 2023. In total, 131 patients were diagnosed with PCa in the intermediate-risk group according to the risk stratification criteria of the NCCN clinical practice guidelines of PCa.[4] Exclusion criteria: (1) patients underwent prostate biopsy (n = 7) within 6 weeks before prostate MRI (reducing postbiopsy changes, including hemorrhage and inflammation, may adversely affect the interpretation of prostate MRI for staging); (2) patients have comorbidities (such as tumor history), prostate treatment history (such as transurethral resection prostate), and other underlying conditions (prostatitis or abscess of the prostate), which could confound the results (n = 2); and (3) the prostate mpMRI have artifacts, which affect the assessment (n = 4). Finally, 118 PCa patients in the intermediate-risk group were enrolled in this study, including 51 FIRG patients and 67 un-FIRG patients ([Fig. 1]).

Fig. 1 Flow diagram of the study population.

MRI Techniques

Every participant was subjected to a prostate MRI examination utilizing a 3.0 T MRI system (GE Discovery MR 750W) equipped with an eight-channel phased-array coil. The scanning sequences included axial T₁-weighted imaging (T₁WI), T₂-weighted imaging (T₂WI) (axial T₂WI, coronal T₂WI, and sagittal T₂WI), DWI, dynamic contrast-enhanced imaging (DCE), IVIM, and DKI. The scanning parameter of DCE-MRI included: three-phase T₁WI plain scans with a rotation angle of 3°6°9° were performed before the DCE-MRI scanning, and the other parameters were consistent with those of the DCE-MRI scanning parameters. A high-pressure syringe was used to inject gadodiamide at 0.25 mL/kg body mass through the dorsal vein, which was rinsed with 20 mL normal saline to ensure complete injection of the drug into the bloodstream. DCE-MRI was performed in 48 continuous scans with 52 layers in each phase, and the scanning time was 7 minutes. Flip degree was 12° and the number of incentives was 0.73. IVIM scanning parameters were: repetition time/echo time (TR/TE), 3500 to 4400/87.5 ms; slice thickness, 3.6 mm; space, 0.5 mm; field of view (FOV), 28 × 28 cm; acquisition matrix, 128 × 128; 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. DKI scans utilized the following settings: TR/TE, 3800 to 4600/87.4 ms; slice thickness, 3 mm; space, 0.5 mm; FOV, 28 × 28 cm; matrix, 128 × 128; with 15 gradient directions and three b-values (0, 1000, 2000 s/mm²), and the selection of b-values was based on the study results of Chou et al.[14] Details are shown in [Table 1].

Table 1 Acquisition parameters of the multiparametic MRI protocol

Sequence	TR/TE (ms)	Slice/Gap (mm)	FOV (cm2)	Matrix	b (s/mm2)	Direction
T2WI	3200–3500/90–100	3/0.5	24 × 24	256 × 224	–	–
T1WI	750–780/9–10	3/0.5	24 × 24	256 × 224	–	–
DWI	3500/70–75	3/0.5	28 × 28	128 × 128	0,2000	–
DCE-MRI	3.8/1.4	3/0.5	32 × 28	256 × 192	–	–
DKI	3800–4600/87.4	3/0.5	28 × 28	128 × 128	0, 1000, 2000	15
IVIM	3500–4400/87.5	3.6/0.5	28 × 28	128 × 128	0, 25, 50, 75, 100, 150, 200, 400, 800, 1200, 2000

Abbreviations: DCE-MRI, dynamic contrast-enhanced MRI; DKI, diffusion kurtosis imaging; DWI, diffusion-weighted imaging; FOV, field of view; IVIM, intravoxel incoherent motion; MRI, magnetic resonance imaging; T₁WI, T₁-weighted imaging; T₂WI, T₂-weighted imaging; TE, echo time; TR, repetition time.

Imaging and Quantitative Data Analysis

The evaluation of all prostate mpMRI adhered to the Prostate Imaging–Reporting and Data System (PI-RADS) version 2 and subsequently version 2.1; PI-RADS version 2 was employed as the evaluative framework prior to the advent of version 2.1.[15] [16] Each image underwent independent appraisal by two veteran radiologists (J.B., 20 years, and D.M.L., 15 years of experience in urogenital imaging), with no prior exposure to the patients' clinical narratives, laboratory diagnostics, or ancillary imaging findings, including ultrasonography. Should discrepancies emerge in diagnostic conclusions, the radiologists would jointly reassess the lesion to establish a consensus.

Quantitative data acquisition was executed using the GE AW 4.6 workstation. Three regions of interest (ROIs) were drawn at the maximum three slices of each target lesions. Each ROI was manually outlined around the whole lesion on the maximum three slices by two radiologists (J.B., 20 years, and K.P.Z., 5 years of experience in urogenital imaging) in consensus on the related parameter maps of IVIM (standard ADC map, diffusion coefficient map, pseudo-diffusion coefficient map, and perfusion fraction map), stretched exponential model (distribute diffusion coefficient [DDC] map and heterogeneity index map), and DKI (fractional anisotropy [FA] map, mean diffusion [MD] map, and mean kurtosis [MK] map) using T₂WI or ADC map as references (if the lesion was located in the peripheral zone, ADC map was referred and if the lesion was located in the transitional zone, T₂WI was referenced). Average values were computed and employed for subsequent data analysis. Detailed visualization is provided in [Fig. 2].

Fig. 2 Quantitative data acquisition using the GE AW 4.6 workstation (GE Healthcare, United States), illustrated by the example of acquiring diffusion coefficient (D) values. Step 1: Assessment of target lesions was performed using the Prostate Imaging–Reporting and Data System (PI-RADS), version 2.1 criteria. Step 2: A region of interest (ROI) was delineated on each of the three maximum slices of the target lesion; each ROI was manually outlined around the whole lesion. Then, the diffusion (D) values for each ROI were subsequently recorded individually. Step 3: Finally, calculation of the mean D value across the target lesions was performed.

Pathology Acquisition and Analysis

All biopsies were conducted with the BK transperineal MRI/TRUS fusion biopsy system (BK Medical, Denmark). Initially, three biopsy cores were taken from each targeted lesion. Subsequently, patients received 18 to 24 systematic biopsies following the Ginsburg protocol, which involves using a spring-loaded biopsy gun equipped with an 18-gauge needle. In this phase, two cores were sampled from each of the 12 sectors, beginning with the anterior sectors. The procedures were performed by either of two experienced urologists (K.G., 11 years, and Y.J.X., 13 years of experience in urogenital pathology diagnosis).

Pathological assessment of each specimen was independently conducted by two pathologists, who were not privy to the MRI outcomes. Discrepancies between the evaluations were reconciled through consensus. The evaluative framework for the pathological analysis adhered to the International Society of Urological Pathology (ISUP) consensus guidelines established in 2014.[17] Each sample's pathological classification and Gleason grading were determined distinctly.

Statistical Analysis

The normality and homogeneity of variances were assessed using the Kolmogorov–Smirnov test and Levene's test, respectively. To test whether the related parameters between FIRG and un-FIRG were statistically significant, an independent sample t-test was used. Binary logistic regression was conducted to investigate the relationship between DKI, IVIM, and stretched exponential model parameters and the likelihood of un-FIRG in the intermediate-risk group. Following the binary logistic regression, receiver operating characteristic (ROC) curve analysis was performed on the statistically significant combination parameters to evaluate their diagnostic effectiveness in identifying un-FIRG within the intermediate-risk group. A p-value of < 0.05 was deemed to indicate statistical significance. All statistical analyses were executed using the SPSS software (IBM, Version 25).

Results

A total of 118 intermediate-risk group PCa patients were included, among which 51 were FIRG ([Fig. 3]), and 67 were un-FIRG ([Fig. 4]). The average age of patients in the FIRG and un-FIRG was 78.53 ± 7.87 (57–92) years and 78.12 ± 7.00 (59–90) years, respectively (p > 0.05). The average total-PSA (T-PSA) of the un-FIRG (16.17 ± 2.39 ng/mL) was significantly higher than that of the FIRG (13.50 ± 2.10 ng/mL) (p < 0.001). Conversely, the average free-PSA (F-PSA)/T-PSA ratio of the un-FIRG (0.14 ± 0.02) was significantly lower than that of the FIRG (0.17 ± 0.02) (p < 0.001). In our study, the average F-PSA of the FIRG (2.33 ± 0.27 ng/mL) was higher than that of the un-FIRG (2.28 ± 0.28 ng/mL), but has no statistical significance (p = 0.323). Both the FIRG group PCa lesions (47.06%) and the un-FIRG group PCa lesions (49.25%) were predominantly PI-RADS category 5, with no statistical significance among PI-RADS categories of the two groups (p = 0.092, detail showed in [Table 2]). Pathologically, in the FIRG, 19.61% was grade group 1, and 80.39% was grade group 2, whereas in the un-FIRG, 5.97% was grade group 1, 19.40% was grade group 2, and 74.63% was grade group 3. In terms of clinical T staging, the percentage of T2c stage patients in the FIRG (43.13%) was lower than that in the un-FIRG (55.22%). The detailed clinical characteristics are shown in [Table 3].

Table 2 PI-RADS classification of PCa lesions according to PI-RADS V 2 and PI-RADS V 2.1

Category	According to PI-RADS V 2 (n = 21)			According to PI-RADS V 2.1 (n = 97)			p
	FIRG (n = 8)	Un-FIRG (n = 13)	p	FIRG (n = 43)	Un-FIRG (n = 54)	p
PI-RADS 2			0.98	3 (2.5%)	2 (1.7%)	0.91	0.92
PI-RADS 3	1 (0.8%)	2 (1.7%)		8 (6.8%)	10 (8.5%)
PI-RADS 4	3 (2.5%)	5 (4.2%)		12 (10.2%)	15 (12.7%)
PI-RADS 5	4 (3.4%)	6 (5.1%)		20 (16.7%)	27 (22.9%)

Abbreviations: FIRG, favorable intermediate-risk group; PCa, prostate carcinoma; PI-RADS, Prostate Imaging–Reporting and Data System; Un-FIRG, unfavorable intermediate-risk group.

Table 3 Clinical characteristics of the patients

Variables	FIRG (n = 51)	Un-FIRG (n = 67)	p
Age (y)	78.53 ± 7.87 (57–92)	78.12 ± 7.00 (59–90)	0.766
T-PSA (ng/mL)	13.50 ± 2.10	16.17 ± 2.39	< 0.001
F-PSA (ng/mL)	2.33 ± 0.27	2.28 ± 0.28	0.323
F-PSA/T-PSA	0.17 ± 0.02	0.14 ± 0.02	< 0.001
Grade group	1 (n = 12) 2 (n = 39)	1 (n = 4) 2 (n = 19) 3 (n = 44)
Clinical T staging	cT2a (n = 5) cT2b (n = 24) cT2c (n = 22)	cT2b (n = 30) cT2c (n = 37)

Abbreviations: FIRG, favorable intermediate-risk group; F-PSA, free-PSA; PSA, prostate-specific antigen; T-PSA, total-PSA; Un-FIRG, unfavorable intermediate-risk group.

Fig. 3 A 76-year-old favorable intermediate-risk group (FIRG) prostate carcinoma (PCa) patient, prostate-specific antigen (PSA) of 10.32 ng/mL, Gleason 3 + 3, and cT2a. (A–C) The lesion with Prostate Imaging–Reporting and Data System (PI-RADS) 4 seen in the left middle peripheral zone. (D) Hematoxylin and eosin (HE) staining, ×100. (E) Fractional anisotropy (FA) of 0.325. (F) Mean diffusion (MD) of 0.897 × 10⁻³ mm²/s. (G) Mean kurtosis (MK) of 0.564. (H) Standard apparent diffusion coefficient (ADC) of 0.627 × 10⁻³ mm²/s. (I) Diffusion coefficient (D) of 0.553 × 10⁻³ mm²/s. (J) Distribute diffusion coefficient (DDC) of 0.701 × 10⁻³ mm²/s.

Fig. 4 A 79-year-old unfavorable intermediate-risk group (un-FIRG) prostate carcinoma (PCa) patient, prostate-specific antigen (PSA) of 16.32 ng/mL, Gleason 4 + 3, and cT2b. (A–C) The lesion with Prostate Imaging–Reporting and Data System (PI-RADS) 5 seen in the left middle peripheral zone. (D) Hematoxylin and eosin (HE) staining, ×100. (E) Fractional anisotropy (FA) of 0.424. (F) Mean diffusion (MD) of 0.815 × 10⁻³ mm²/s. (G) Mean kurtosis (MK) of 0.652. (H) Standard apparent diffusion coefficient (ADC) of 0.507 × 10⁻³ mm²/s. (I) Diffusion coefficient (D) of 0.451 × 10⁻³ mm²/s. (J) Distribute diffusion coefficient (DDC) of 0.539 × 10⁻³ mm²/s.

Results of Independent Samples t-Test

Results of independent samples t-test showed that the parameters FA, MK, and MD of DKI, the parameters standard ADC and diffusion coefficient (D) of IVIM, and the parameter DDC of stretched exponential model have significant difference between the FIRG and un-FIRG ([Table 4]). FA value and MK value of the un-FIRG (0.413 ± 0.066, 0.637 ± 0.066) were significantly higher than those of the FIRG (0.317 ± 0.067, 0.555 ± 0.079). On the contrary, the MD value, standard ADC value, D value, and DDC value of the un-FIRG were significantly lower than those of the FIRG. Compared with the FIRG, the MD value, standard ADC value, D value, and DDC value of the un-FIRG were decreased by 0.067 × 10⁻³ mm²/s, 0.101 × 10⁻³ mm²/s, 0.109 × 10⁻³ mm²/s, and 0.143 × 10⁻³ mm²/s, respectively.

Table 4 Results of independent samples t-test

Variables	FIRG	Un-FIRG	t	p-Value
FA	0.317 ± 0.067	0.413 ± 0.066	−7.807	< 0.001
MK	0.555 ± 0.079	0.637 ± 0.066	−6.153	< 0.001
MD (×10−3 mm2/s)	0.888 ± 0.133	0.821 ± 0.148	2.536	0.013
Standard ADC (×10−3 mm2/s)	0.621 ± 0.122	0.520 ± 0.120	4.495	< 0.001
D (×10−3 mm2/s)	0.555 ± 0.100	0.446 ± 0.957	6.016	< 0.001
DDC (×10−3 mm2/s)	0.688 ± 0.111	0.545 ± 0.084	7.944	< 0.001

Abbreviations: D, pure diffusion coefficient; DDC, distributed diffusion coefficient; FA, fractional anisotropy; FIRG, favorable intermediate-risk group; MD, mean diffusion; MK, mean kurtosis; standard ADC, standard apparent diffusion coefficient; un-FIRG, unfavorable intermediate-risk group.

Results of Binary Logistic Regression Analysis

Binary logistic regression analysis was used to analyze the relevant parameters of IVIM, DKI, and stretched exponential models and construct a diagnostic model. The results showed that the diagnostic model was statistically significant, chi-square = 90.969, p < 0.001, which could effectively distinguish the 86.40% of un-FIRG in the intermediate-risk group. Among the nine independent variables included in the model, the FA value (p < 0.001), MK value (p = 0.014), MD value (p = 0.031), D value (p = 0.041), and DDC value (p = 0.001) were statistically significant. Odds ratios (95% confidence interval [CI]) were 6.856 (2.679–17.545), 3.579 (1.292–10.013), 3.018 (1.107–8.227), 0.411 (0.175–0.962), and 0.177 (0.062–0.510), respectively ([Table 5], [Fig. 5]). The results of ROC analysis showed that the area under the curve was 0.9429, 95% CI: 0.9095–0.9805. Sensitivity, specificity, positive predictive value, and negative predictive value were 88.06, 84.31, 88.06, and 84.31%, respectively.

Fig. 5 The binary logistic regression analysis results revealed that, in the intermediate-risk group, the prevalence of unfavorable intermediate-risk group (un-FIRG) increased with rising fractional anisotropy (FA) and mean kurtosis (MK) values but decreased as mean diffusion (MD), diffusion coefficient (D), and distribute diffusion coefficient (DDC) values increased. The diagnostic efficacy of these parameters in differentiating un-FIRG from the intermediate-risk group was 0.9429.

Table 5 Prediction for un-FIRG PCa patients from intermediate-risk group

Variables	p-Value	Odds ratio (95% CI)
FA	< 0.001	6.856(2.679~17.545)
MK	0.014	3.579(1.292~10.013)
MD	0.031	3.018(1.107∼8.227)
D	0.041	0.411(0.175∼0.962)
DDC	0.001	0.177(0.062∼0.510)

Abbreviations: CI, confidence interval; D, pure diffusion coefficient; DDC, distributed diffusion coefficient; FA, fractional anisotropy; MD, mean diffusion; MK, mean kurtosis; PCa, prostate carcinoma; un-FIRG, unfavorable intermediate-risk group.

Discussion

In this study, a significant difference was found in the parameters of IVIM, stretching exponential model, and DKI between the FIRG and un-FIRG. MD value, standard ADC value, D value, and DDC value of the FIRG were significantly lower than those of the un-FIRG. Results of our study are consistent with previous studies.[13] [18] [19] Parameters MD value of DKI, standard ADC value and D value of IVIM, and DDC value of stretching exponential model reflect the diffusion restriction of water molecules in tissues. Some studies have suggested that the cause of decrease of MD value, standard ADC value, D value, and DDC value with increase of PCa Gleason pattern may be related to histological differences, including vascular (i.e., capillaries), fibromuscular stroma, epithelium, and glandular lumen.[13] [18] [19] [20] PCa with higher Gleason pattern has higher cellularity and smaller extracellular space, which leads to increased diffusion restriction of water molecules in the tissue. According to the assessment criterion of PCa pathological tissue recommended by the ISUP, the growth pattern of PCa tissue was divided into five grades by the Gleason system, with the increase of Gleason pattern, the morphology of histiocytes become more irregular, have higher cellularity, and the destruction of glandular structure become more complete.[21] In addition, our results showed that the pathology of FIRG was mainly grade group 2 (Gleason 3 + 4), accounting for approximately 76.47%, in contrast, the pathology of un-FIRG was mainly grade group 3 (Gleason 4 + 3), accounting for approximately 65.67%. Therefore, the reasonable explanation may be that compared with the FIRG, the un-FIRG have higher Gleason pattern and the tissue contains increased volumes of low-restriction diffusivity epithelial cells and decreased high-restriction diffusivity stroma and lumen space. Those factors resulted in the un-FIRG having lower D value, standard ADC value, D value, and DDC value than those of the FIRG. In fact, the diffusion motion of water molecules in tissues is non-Gaussian, which is also reflected in our research results, where the FA value and MK value of FIRG were significantly higher than that of the un-FIRG. Our results are consistent with the results of previous studies.[22] [23] [24] Results of Chatterjee et al[24] showed that the nuclear fraction, cytoplasmic fraction, and cellular fraction increase with the increase of the Gleason pattern, while the stromal fraction and luminal fraction have opposite correlation with the Gleason pattern. Those pathological differences increased diffusion restriction of water molecules in tissues, while also causing the diffusion restriction of water molecules deviating from normal distribution more significant. As a result, those factors may be responsible for the significantly higher FA value and MK value of the un-FIRG than the FIRG.

In our study, pseudo-diffusion coefficient (D*) value and perfusion fraction (f) value have no statistical significance between the FIRG and un-FIRG. This is in keeping with previous study.[20] It is known that the D* and f values mainly reflect the perfusion of the tissue.[25] Specifically, separation of perfusion from diffusion requires high signal-to-noise ratios, and there are some technical challenges to overcome, such as artifacts from other bulk flow phenomena. Vascular and tubular flow may be difficult to disentangle in some tissues. Active transport resulting from glandular secretion (such as breast ducts, salivary glands, and pancreas) may also be difficult to separate from microcapillary perfusion. What is more, IVIM imaging has a differential sensitivity to vessel sizes, according to the range of b-values that are used.²⁵ Those factors may be responsible for the failure of D* value and f value to differentiate the FIRG and un-FIRG. Some studies have shown that the parameter heterogeneity index (α) value of the stretched exponential model has the ability to diagnose PCa.[26] However, results of our study showed that the α value has no statistical significance between the FIRG and un-FIRG. The plausible explanation may be that compared with benign prostate tissue, PCa has relatively complex microscopic structure (including higher tumor density, smaller glandular lumen, and extracellular space, etc.), these factors result in α value to have the ability to differentiate PCa and benign prostatic hyperplasia. However, in PCa tissue, the Gleason pattern, as an important indicator to evaluate the aggressiveness of PCa, is based on the microstructure of PCa tissue (including the relative proportion of luminal, epithelial, and stromal components), not just tumor heterogeneity. Those microstructure differences between the FIRG and un-FIRG may not be sufficient to be reflected by α value.

Binary logistic regression analysis was used to analyze the relevant parameters of IVIM, DKI, and stretched exponential model to construct a diagnostic model, and the results showed that the diagnostic efficiency was 0.9429. This result may be related to the un-FIRG to have higher Gleason pattern than the FIRG. In our study, the pathology of the FIRG were Gleason 3 + 3 and 3 + 4, of which Gleason 3 + 4 accounted for 80.39%, while the pathology of the un-FIRG were Gleason 3 + 3, Gleason 3 + 4, and Gleason 4 + 3, and Gleason 4 + 3 accounted for 74.63%. Previous studies have observed that compared with Gleason pattern 3 + 4 PCa, Gleason pattern 4 + 3 PCa has significantly lower standard ADC value, D value, DDC value, and MD value, and have significantly higher MK value and FA value,[27] [28] and the results were confirmed in our study. Tamada et al[28] used DWI and DKI to differentiate Gleason 3 + 4 and Gleason 4 + 3 PCa, and the results showed that the diagnostic efficacy was 0.721. In contrast, the results of our study showed higher diagnostic efficacy, which may be related to the patients of Gleason 3 + 4 PCa (n = 113) being significantly more than Gleason 4 + 3 PCa (n = 48).

There are some limitations in our study. First, our study was a single-center study. Second, in our study, the ROIs were manually drawn at the slice with the maximum slice of PCa lesions, as well as the nearest levels above and below it. What is more, the ROIs were manually outlined around the whole PCa lesions. The choice of different ROIs might also lead to differences in study results due to tumor size and heterogeneity. Finally, our study only analyzed the mean values of the relevant parameters, while some research results showed that it may be a feasible method to use histogram analysis.[29]

Conclusion

In conclusion, the risk stratification modeling combining IVIM, stretched exponential model, and DKI exhibited great potential for noninvasive risk stratification of intermediate-risk group PCa, revealing significant discriminatory power.

Conflict of Interest

None declared.

Ethics

Institutional Ethics Committee of The Second Hospital of Dalian Medical University approved the study.

^* These authors contributed equally to this work.

• References

• 1 Siegel RL, Miller KD, Wagle NS, Jemal A. Cancer statistics, 2023. CA Cancer J Clin 2023; 73 (01) 17-48
• 2 Moses KA, Sprenkle PC, Bahler C. et al. NCCN Guidelines® Insights: Prostate Cancer Early Detection, Version 1.2023. J Natl Compr Canc Netw 2023; 21 (03) 236-246
• 3 Ward EM, Sherman RL, Henley SJ. 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) 1279-1297
• 4 Schaeffer EM, Srinivas S, Adra N. et al. NCCN Guidelines® Insights: Prostate Cancer, Version 1.2023. J Natl Compr Canc Netw 2022; 20 (12) 1288-1298
• 5 Reese AC, Pierorazio PM, Han M, Partin AW. Contemporary evaluation of the National Comprehensive Cancer Network prostate cancer risk classification system. Urology 2012; 80 (05) 1075-1079
• 6 Gandaglia G, Karnes RJ, Sivaraman A. et al. Are all grade group 4 prostate cancers created equal? Implications for the applicability of the novel grade grouping. Urol Oncol 2017; 35 (07) 461.e7-461.e14
• 7 Zumsteg ZS, Spratt DE, Pei I. et al. A new risk classification system for therapeutic decision making with intermediate-risk prostate cancer patients undergoing dose-escalated external-beam radiation therapy. Eur Urol 2013; 64 (06) 895-902
• 8 Grey ADR, Scott R, Shah B. et al. Multiparametric ultrasound versus multiparametric MRI to diagnose prostate cancer (CADMUS): a prospective, multicentre, paired-cohort, confirmatory study. Lancet Oncol 2022; 23 (03) 428-438
• 9 Eldred-Evans D, Neves JB, Simmons LAM. et al. Added value of diffusion-weighted images and dynamic contrast enhancement in multiparametric magnetic resonance imaging for the detection of clinically significant prostate cancer in the PICTURE trial. BJU Int 2020; 125 (03) 391-398
• 10 Glazer DI, Hassanzadeh E, Fedorov A. et al. Diffusion-weighted endorectal MR imaging at 3T for prostate cancer: correlation with tumor cell density and percentage Gleason pattern on whole mount pathology. Abdom Radiol (NY) 2017; 42 (03) 918-925
• 11 Barrett T, McLean M, Priest AN. et al. Diagnostic evaluation of magnetization transfer and diffusion kurtosis imaging for prostate cancer detection in a re-biopsy population. Eur Radiol 2018; 28 (08) 3141-3150
• 12 Tamada T, Prabhu V, Li J, Babb JS, Taneja SS, Rosenkrantz AB. Assessment of prostate cancer aggressiveness using apparent diffusion coefficient values: impact of patient race and age. Abdom Radiol (NY) 2017; 42 (06) 1744-1751
• 13 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) 3441-3452
• 14 Chou MC, Ko CW, Chiu YH, Chung HW, Lai PH. Effects of B value on quantification of rapid diffusion kurtosis imaging in normal and acute ischemic brain tissues. J Comput Assist Tomogr 2017; 41 (06) 868-876
• 15 Weinreb JC, Barentsz JO, Choyke PL. et al. PI-RADS Prostate Imaging - Reporting and Data System: 2015, Version 2. Eur Urol 2016; 69 (01) 16-40
• 16 Turkbey B, Rosenkrantz AB, Haider MA. et al. Prostate Imaging Reporting and Data System Version 2.1: 2019 update of Prostate Imaging Reporting and Data System Version 2. Eur Urol 2019; 76 (03) 340-351
• 17 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 (04) e1-e7
• 18 Chen Z, Xue Y, Zhang Z. et al. The performance of intravoxel-incoherent motion diffusion-weighted imaging derived hypoxia for the risk stratification of prostate cancer in peripheral zone. Eur J Radiol 2020; 125: 108865
• 19 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
• 20 Yang DM, Kim HC, Kim SW. et al. Prostate cancer: correlation of intravoxel incoherent motion MR parameters with Gleason score. Clin Imaging 2016; 40 (03) 445-450
• 21 van Leenders GJLH, van der Kwast TH, Grignon DJ. et al; ISUP Grading Workshop Panel Members. The 2019 International Society of Urological Pathology (ISUP) consensus conference on grading of prostatic carcinoma. Am J Surg Pathol 2020; 44 (08) e87-e99
• 22 Dominguez C, Plata M, Cataño JG. et al. Diagnostic accuracy of multiparametric magnetic resonance imaging in detecting extracapsular extension in intermediate and high - risk prostate cancer. Int Braz J Urol 2018; 44 (04) 688-696
• 23 Ding K, Yao Y, Gao Y. et al. Diagnostic evaluation of diffusion kurtosis imaging for prostate cancer: detection in a biopsy population. Eur J Radiol 2019; 118: 138-146
• 24 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 (03) 751-762
• 25 Iima M, Le Bihan D. Clinical intravoxel incoherent motion and diffusion MR imaging: past, present, and future. Radiology 2016; 278 (01) 13-32
• 26 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 (04) 1078-1085
• 27 Fukunaga T, Tamada T, Kanomata N. 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: 152-159
• 28 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 (01) 100-108
• 29 Wang Q, Li H, Yan X. et al. Histogram analysis of diffusion kurtosis magnetic resonance imaging in differentiation of pathologic Gleason grade of prostate cancer. Urol Oncol 2015; 33 (08) 337.e15-337.e24

Publication History

Article published online:
19 May 2025