Sub-differentiation of PI-RADS 3 lesions in TZ by advanced diffusion-weighted imaging to aid the biopsy decision process

Background Performing biopsy for intermediate lesions with PI-RADS 3 has always been controversial. Moreover, it is difficult to differentiate prostate cancer (PCa) and benign prostatic hyperplasia (BPH) nodules in PI-RADS 3 lesions by conventional scans, especially for transition zone (TZ) lesions. The purpose of this study is sub-differentiation of transition zone (TZ) PI-RADS 3 lesions using intravoxel incoherent motion (IVIM), stretched exponential model, and diffusion kurtosis imaging (DKI) to aid the biopsy decision process. Methods A total of 198 TZ PI-RADS 3 lesions were included. 149 lesions were BPH, while 49 lesions were PCa, including 37 non-clinical significant PCa (non-csPCa) lesions and 12 clinical significant PCa (csPCa) lesions. Binary logistic regression analysis was used to examine which parameters could predict PCa in TZ PI-RADS 3 lesions. The ROC curve was used to test diagnostic efficiency in distinguishing PCa from TZ PI-RADS 3 lesions, while one-way ANOVA analysis was used to examine which parameters were statistically significant among BPH, non-csPCa and csPCa. Results The logistic model was statistically significant (χ2 = 181.410, p<0.001) and could correctly classify 89.39% of the subjects. Parameters of fractional anisotropy (FA) (p=0.004), mean diffusion (MD) (p=0.005), mean kurtosis (MK) (p=0.015), diffusion coefficient (D) (p=0.001), and distribute diffusion coefficient (DDC) (p=0.038) were statistically significant in the model. ROC analysis showed that AUC was 0.9197 (CI 95%: 0.8736-0.9659). Sensitivity, specificity, positive predictive value and negative predictive value were 92.1%, 80.4%, 93.9% and 75.5%, respectively. FA and MK of csPCa were higher than those of non-csPCa (all p<0.05), while MD, ADC, D, and DDC of csPCa were lower than those of non-csPCa (all p<0.05). Conclusion FA, MD, MK, D, and DDC can predict PCa in TZ PI-RADS 3 lesions and inform the decision-making process of whether or not to perform a biopsy. Moreover, FA, MD, MK, D, DDC, and ADC may have ability to identify csPCa and non-csPCa in TZ PI-RADS 3 lesions.


Introduction
The pathological and clinical features of prostate cancer (PCa) vary in different anatomical regions of the prostate. Compared to peripheral zone (PZ) cancers, transition zone (TZ) cancers have lower Gleason scores, higher tumor volumes, higher prostate-specific antigen (PSA) levels, and are often confined to the prostate (1). Although patients with TZ cancers have a more favorable prognosis than patients with PZ cancers, TZ cancers are more difficult to detect, especially those with small tumor volume (1,2).
Multiparametric MRI has been increasingly used to detect and locate tumors in patients highly suspected of PCa based on clinical examination results, such as PSA level and digital rectal exam. The recently updated version 2.1 of Prostate Imaging-Reporting and Data System (PI-RADS) guidelines have standardized the imaging acquisition and interpretation, providing more detailed principles for evaluating prostate lesions to reduce interreader variability (3).
Several studies have shown that csPCa in PI-RADS 3 have higher PSA density, lower prostate volume, and apparent diffusion coefficient (ADC) value compared with benign lesions in PI-RADS 3 (11)(12)(13)(14). Using conventional scanning sequences makes it difficult to distinguish TZ cancers from fibromuscular (stromal) benign prostatic hyperplasia (BPH) as both of these conditions can manifest low T2 signal intensity on MR images. Hansen et al. (15) have found that PI-RADS 3 lesions with low ADC value located in anterior of TZ, as well as the irregular shape, ill-defined border, and homogenous T2 signal intensity can predict the PCa in TZ; however, only ill-defined border and low ADC value help predict scPCa. We found that the evaluation criteria for PI-RADS 3 lesions in Hansen's study were PI-RADS Version 1 and Version 2. Tamada et al. (16) found that compared with PI-RADS Version 2.1, PI-RADS Version 2 could classify some PI-RADS 4 or 5 lesions as PI-RADS 3 lesions. Advanced diffusion-weighted imaging (DWI) such as intravoxel incoherent motion (IVIM), stretched exponential model, and diffusion kurtosis imaging (DKI) can reflect microstructural complexity of tumor tissues, what's more, DKI can provide information about non-Gaussian diffusion. some studies have demonstrated that parameters derived from IVIM, stretched exponential model, and DKI was useful in the detection and assessment of PCa aggressiveness and may even be superior to DWI (17)(18)(19).
In strict compliance with the evaluation criteria of PI-RADS V2.1, the present study attempted to differentiate PCa and BPH in PI-RADS 3 lesions in TZ with advanced diffusion-weighted imaging (IVIM, DKI, stretched exponential model) and to analyze which parameters have potential value in predicting csPCa.

Patients
This study was approved by the ethics committee (subheading on ethics committee incompleted). A total of 637 patients with an abnormal increase in serum PSA (T-PSA>10 ng/mL; T-PSA, 4-10 ng/mL and F-PSA/T-PSA<0.25) underwent prostate MRI examination between December 2018 and January 2021. Inclusion criteria: patients with the highest PI-RADS category 3 lesions in TZ with a diameter larger than 10 mm. Exclusion criteria (1): patients underwent prostate biopsy (n=11) within 2 weeks before prostate MRI (2); the interval between prostate MRI examination and magnetic resonance imaging-Transrectal ultrasound (MRI-TRUS) biopsy >1 month (n=2) (3); pathology was not obtained by MRI-TRUS fusion biopsy (n=14) (4); the lesions were located at the base of the prostate so that biopsy specimens could not be obtained (n=2). Finally, 145 patients with 198 PI-RADS category 3 lesions located in TZ were included in the present study. Among 198 PI-RADS 3 lesions in TZ, 149 lesions were pathologically confirmed as BPH, while 49 lesions as PCa, among which 37 were non-clinically significant prostate cancer (non-cs PCa, Gleason 3 + 3) and 12 were csPCa ( Figure 1).
IVIM, stretched exponential model and DKI models IVIM model and its parameters of diffusion coefficient (D), pseudo-diffusion coefficient (D*), and perfusion fraction (f) are fit for a biexponential equation: Where D characterizes extravascular diffusion of water, D* represents signal changes attributing to the intravascular movement of water, f is the perfusion fraction. S b is the DWI signal intensity at a specified b value, and S 0 is the baseline signal at b = 0 s/mm 2 (20).
Stretched exponential model and its parameters of distribution diffusion coefficient (DDC) and heterogeneity index (a) are fit by the following equation: Where S b represents the signal intensity at a specified b value, S 0 is the signal intensity based on b = 0 s/mm 2 . DDC is a measure of the rate of signal decay with various b values, representing mean intravoxel diffusion rates. The heterogeneity index (a) is the water molecular diffusion heterogeneity and related to intravoxel water diffusion heterogeneity (range, 0-1). A higher a value indicates low intravoxel diffusion heterogeneity, which approaches pure monoexponential decay. Conversely, an alpha = 0 indicates a higher degree of multiexponential signal decay (21).
The DKI model is based on the following equation: Where S b represents the signal intensity at a specified b value, S 0 is the signal intensity based on b = 0 s/mm 2 . When S 0 is known, mean diffusion (MD) and mean kurtosis (MK) are obtained. The parameter MK represents the apparent diffusional kurtosis, and MD is the diffusion coefficient that is corrected to account for the observed non-Gaussian behavior.

Imaging and quantitative data analysis
According to PI-RADS version 2.1 (3), we screened the PI-RADS 3 lesions strictly located in TZ ( Table 2). All the images were independently evaluated by two radiologists (JW. L., J.B.) with more than 10 years of experience in abdominal imaging diagnosis without knowing the clinical history, laboratory examination results,

Biopsy
The BK transperineal MRI/TRUS fusion biopsy system (BK Medical, Denmark) was used for all biopsies. Three target biopsy cores were taken from each target lesion before the systematic biopsies ( Figure 3). According to the Ginsburg protocol, all patients had 18-24 systematic biopsies taken using a spring-loaded biopsy gun with an 18 gauge needle. During the systematic biopsy, 2 biopsy cores were sampled from each of 12 sectors, starting with the anterior sectors. All procedures were undertaken by 1 of 2 urologists with more than 10 years of experience in transperineal biopsy using the BK MRI/TRUS fusion biopsy system.

Pathological analysis
All the pathological samples were separately evaluated by two pathologists (Y. L., K. G.) who were blinded to MR imaging results. In case of deviation in the results, an agreement was reached. The criteria for pathological evaluation were based on ISUP recommendations in 2014 (22). The pathological type and Gleason score of each specimen were given separately.

Statistical analysis
One-way ANOVA was used to test whether the clinical characteristics of patients were statistically significant. In order  to examine the association between parameters of DKI, IVIM, and stretched exponential model with the probability of PCa in PI-RADS 3 lesions, binary logistic regression analysis was applied. After binary logistic regression analysis, the receiver operating characteristic (ROC) curve of combination parameters, which were statistically significant in binary logistic regression analysis, were analyzed to test the diagnostic efficiency of detecting PCa in PI-RADS 3 lesions. One-way ANOVA was used to analyze whether parameters of DKI, IVIM, and stretched exponential model was helpful in differentiating BPH, PCa, and csPCa in PI-RADS 3 lesions. A p-value< 0.05 was considered statistically significant. All analyses were performed with SPSS (IBM, SPSS version 25).

Results of one-way ANOVA analysis
Results of one-way ANOVA analysis showed that parameters FA, MD, MK, ADC, D, DDC were statistically significant ( Table 5). The

Discussion
Our study included 198 PI-RADS category 3 lesions located in TZ, where the detection rate of non-csPCa was 18.69%, and the detection rate of csPCa was 6.06%. In our study population, 23.45% of patients were non-csPCa, and thus required active surveillance at the time of initial biopsy according to NCCN guidelines. In addition, we detected 8.28% of csPCa requiring active treatment. Hansen et al. (15) found that 45.88% were PCa in PI-RADS 3 lesions located in TZ, and 21.18% were csPCa, resulting in a higher detection rate compared to the one detected in the present study. This may be because their study population was a mixed population, including the initial biopsy and the second-biopsy populations.
Typically, PCa in TZ has homogeneous low T2 signal intensity, ill-defined margins, lack of capsule, lenticular shape, and is more prone to invasion in the anterior fibromuscular stroma (1). Usually, lesions with these manifestations are assessed as PI-RADS category 4 or 5. However, in PI-RADS 3 lesions, detection of PCa using conventional multiparameter prostate MRI can be challenging. Our results showed that combining quantitative parameters FA, MD, and MK of DKI, parameter D of IVIM, and parameter DDC of the stretched exponential model could effectively identify PCa from PI-RADS 3 lesions. D value, DDC value and MD value of PCa (0.583×10 -3 mm/s, 0.662×10 -3 mm/s, 0.989×10 -3 mm/s) were significantly lower than that of BPH (0.778×10 -3 mm/s, 0.965×10 -3 mm/s, 1.219×10 -3 mm/ s), which is consistent with the results of previous studies (23)(24)(25). D, DDC, and MD all reflect the diffusion of water molecules in tissues. Results of our study may be attributed to the difference in histological changes of PCa and BPH, including vascular (i.e., capillaries), fibromuscular stroma, epithelium, and glandular lumen. Compared with BPH, PCa contains increased volumes of low-restriction diffusivity epithelial cells and decreased high-restriction diffusivity stroma and lumen space. In fact, the diffusion motion of water molecules in tissues is non-Gaussian (26 [26][27][28]. Although it is difficult to distinguish PCa from BPH nodules by morphological features, PCa has a more complex microstructure than BPH, such as more obvious cell atypia, higher cell density, and neovascularization. These changes make the FA and MK values of PCa higher than those of PBH. Several studies have also found that a low ADC value can predict PCa in PI-RADS 3 lesions (13, 15). However, our results showed that the ADC value was not significant in our model. A plausible explanation for this may be that ADC value does not reflect the true diffusion of tissue but a combination of restricted diffusion and perfusion of tissue. Active treatment is needed for csPCa (31). Screening csPCa from PI-RADS 3 lesions may be the most important task. Our research showed that parameters FA and MK of csPCa were significantly higher than those of non-csPCa and BPH. On the contrary, compared with non-csPCa and BPH, csPCa had lower MD, ADC, D, and DDC values. Park et al. (32) also found that csPCa had a higher MK value than non-csPCa and BPH, as well as a lower MD value. This undoubtedly indicated that csPCa had a more complex microstructure compared to non-csPCa and BPH. Chatterjee et al. (33) showed that the microstructure of PCA, such as gland component volumes of epithelium, stroma, and lumen space, were significantly correlated with Gleason. Also, these factors increased the degree of restriction diffusion of water molecules in tissues. Moreover, the diffusion restriction of water molecules deviating from normal distribution was more significant. As a result, the FA and MK values of csPCa increased. Several studies have found that csPCa has a lower D value and DDC value than non-csPCA and BPH. Moreover, D and DDC showed a negative relationship with the Gleason pattern (19, 25,34), which is in line with our results. These findings strongly suggest that parameters FA, MD, MK of DKI, parameter ADC, D of IVIM, and parameter D of the stretched exponential model could be potentially used to detect csPCa.
Our results revealed that D* and f had no statistical significance among BPH, non-csPCa, and csPCa. On the contrary, Beyhan et al. (35) studied 31 patients, revealing that f of PCa in TZ was significantly lower than that of BPH. In addition, D* of PCa in TZ was also significantly higher than that of BPH, which might be because their study population was smaller than ours. Moreover, Oto et al. (36) have  found that BPH nodules, dominated by stromal hyperplasia, can mimic PCA on T2WI and demonstrate early enhancement on DCE images. These findings suggest that D* and f cannot distinguish BPH, PCa, and csPCa. Our results showed that a wast no statistical significance, which is consistent with the previous studies (34,37). On the contrary, Liu et al. (38) have found that a could differentiate PCa and BPH. In our clinical practice, we found that the signal of PCa was relatively homogeneous and rarely necrotic. In PCa tissue, the Gleason pattern is based on tissue architecture, including luminal, epithelial, and stromal components, and its change is based on their relative sizes rather than only tumor heterogeneity (39). These changes may support that a may not have the ability to distinguish BHP, PCa and csPCa.

Limitations
The present research has some limitations. First, in our study, the diameter of the lesion was larger than 10 mm. Therefore, we do not know whether our results are applicable to lesions with a diameter that is less than 10 mm. Second, our sample size was small, especially csPCa, which only included 12 patients. Third, this was a single-center study, and the study results may be biased. In order to make the study results more robust and generalizable, multi-center and large sample studies may be needed to confirm our current results.

Conclusion
In the summary, the relevant parameters of IVIM, stretched exponential model and DKI were used to establish the model, in which FA, MD, MK, D, and DDC can predict PCa in TZ PI-RADS 3 lesions and inform the decision-making process of whether or not to perform a biopsy. Moreover, FA, MD, MK, D, DDC, and ADC could differentiate csPCa and non-csPCa in TZ PI-RADS 3 lesions. The results of binary logistic regression analysis showed that in pi-rads 3 lesions, the prevalence of PCa increased with the increase of FA value and MK value but decreased with the increase of MD value, D value, and DDC value. The diagnostic efficiency in distinguishing PCa from PI-RADS 3 lesions of these parameters was 0.9197.

Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethics statement
The studies involving human participants were reviewed and approved by the ethics committee of the second hospital of Dalian Medical university. The patients/participants provided their written informed consent to participate in this study.
Author contributions K-PZ, H-BH, CB: these authors contributed equally to this work and share first authorship. JB: this author is the corresponding author. Q-YL: this author is the co-corresponding author and contributed equally to this work and share corresponding authorship. All authors contributed to the article and approved the submitted version.

Conflict of interest
Author LZ-X was employed by GE Healthcare. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.