Introduction

It has been estimated that about 248,530 new cases of prostate cancer (PCa) will be diagnosed in the United States during 2021, and this disease is the second-leading cause of cancer death for men in most Western countries [1, 2]. Accordingly, early detection of PCa with accurate assessment of tumor aggressiveness and local staging is essential to improve mortality rates as well as patient prognosis. Multiparametric MRI (mpMRI) comprising T2-weighted imaging (T2WI), diffusion-weighted imaging (DWI), and dynamic contrast-enhanced MRI (DCE-MRI) prior to prostate biopsy, followed by MR-guided prostate biopsy (such as MRI–ultrasound fusion-guided prostate biopsy), is the recommended protocol to replace standard systematic prostate biopsy for detection of PCa in patients with elevated PSA levels [3,4,5,6]. mpMRI, especially DWI, has already made a strong contribution to the accumulation of research results with regard to the detection and localization of primary clinically significant PCa (csPCa) and local recurrence, assessment of tumor aggressiveness by such as Gleason score (GS) and Gleason grade (GG), local staging, active surveillance (AS), and standardization of prostate MRI diagnosis [the Prostate Imaging Reporting and Data System version (PI-RADS)] [7,8,9,10,11,12,13,14,15,16,17,18,19,20,21]. Furthermore, in recent years, numerous studies regarding biparametric MRI (bpMRI), which does not include DCE-MRI, have reported comparable diagnostic accuracy between bpMRI and mpMRI for detecting csPCa [22,23,24,25,26,27,28]. Therefore, the role of DWI in csPCa detection is becoming increasingly important, and its optimization is an urgent issue. In this review article, we discuss the problems of standard DWI with single-shot echo-planar imaging (ssEPI) and their countermeasures, as well as the possibility of clinical application of quantitative DWI, and the latest DWI technology.

New technology to overcome problems in standard DWI

Susceptibility artifact and geometric anatomic distortion

Currently, clinical DWI acquisition is routinely performed using two-dimensional (2D) ssEPI technique. DWI with 2D ssEPI (ssEPI DWI) has high signal-to-noise ratio (SNR) and is minimally affected by motion artifact due to its rapid acquisition. However, the image quality of ssEPI DWI suffers from susceptibility artifact caused by gas within the adjacent rectum and by hip implants, and from marked geometric anatomical distortion resulting from factors such as the very rapid acquisition, B0- and B1-field inhomogeneities, and eddy currents [29, 30]. These effects become more pronounced at higher field strength of 3T, which is widely used in prostate mpMRI. Therefore, we need to continue to verify the usefulness of bowel preparation techniques and develop MRI techniques that can improve image quality.

The basic countermeasures against susceptibility artifact and geometric anatomical distortion in ssEPI DWI include dietary restrictions and administration of hyoscine N-butylbromide to reduce intestinal peristalsis before the MRI examination. Schmidt et al. have reported that among hyoscine N-butylbromide, microenema, and dietary restrictions for artifact reduction and image quality in prostate mpMRI, only microenema appeared to significantly improve the image quality of DWI and the whole mpMRI image set of the prostate [31]. A review article that compared the effectiveness of antispasmodics and rectal enemas concluded that intravenous hyoscine butylbromide was the optimum patient preparation method for improving T2W and DWI image quality in prostate mpMRI, and did not recommend the use of a preparatory rectal enema [32]. Therefore, it is necessary to take active measures to improve image quality of ssEPI DWI using MRI techniques while continuing to verify the clinical usefulness of bowel preparation techniques.

Turbo spin-echo DWI (TSE DWI)

SsTSE DWI combines TSE readout with a single-shot acquisition. TSE readout is not sensitive to susceptibility artifact, whereas single-shot acquisition can shorten the scan time and is not sensitive to motion effect. Initially, signal loss and artifact were drawbacks of ssTSE DWI, because the refocusing pulses after DW preparation violate the Carr–Purcell–Meiboom–Gill (CPMG) condition. Imperfect 180° refocusing of RF pulses will generate both spin echoes (SEs) and stimulated echoes (STEs). If the signal phase after the excitation RF pulse aligns with the phase of subsequent refocusing pulses, STE does not occur. However, as spins are left with unpredictable magnetization phases after DW preparation and this condition is significantly worsened by the presence of tissue motion during DW preparation, STEs are generated. Therefore, the combination of imperfect refocusing RF pulses and random signal phase after DW preparation leads to destructive interference between SEs and STEs, inconsistent signal losses, and consequent image artifacts. Alsop proposed a method that applies a 90° pulse, dephasing gradient, and rephrasing gradient to eliminate non-CPMG (i.e., STE) components [33]. The CPMG condition, which requires alignment of the initial transverse magnetization with the axis of the refocusing pulses, is violated due to random phase error accumulated during the diffusion preparation, which is termed non-CPMG condition. The non-CPMG component of the signal is effectively “hidden” along the longitudinal axis by a 90° pulse prior to the echo train. Because the non-CPMG component that causes artifacts is not acquired, SNR is reduced.

Split acquisition of fast spin-echo signals for diffusion imaging (SPLICE) [34] has been developed to overcome low SNR in ssTSE DWI. This sequence acquires the SE and STE components separately using unbalanced readout gradient and then reconstructs each k-space separately to avoid destructive phase interference. SPLICE with high SNR could be a new option against severe distortion. Figure 1 shows a clinical case in which prostate imaging was obtained using ssEPI DWI, ssTSE DWI, and SPLICE.

Fig. 1
figure 1

A 40-year-old male without prostate cancer. ssEPI DWI image (a) shows distortion of the prostate due to air in the rectum. In comparison, ssTSE DWI (b) and SPLICE (c) are less affected by such distortion. Note the higher signal-to-noise ratio of the prostate with SPLICE (c) than with ssTSE DWI (b). ssEPI single-shot echo-planar imaging, DWI diffusion-weighted imaging, ssTSE single-shot turbo spin-echo; SPLICE, split acquisition of fast spin-echo signals for diffusion imaging

Full size image

Reduced field-of-view DWI

A recently introduced technology for EPI acquisition termed “reduced field-of-view” (rFOV)-DWI has potential for improving image quality, including issues associated with standard ssEPI DWI such as anatomic distortion and susceptibility artifact. rFOV-DWI can be acquired using commercially available sequences such as ZOOM DWI (Philips), ZOOMit (Siemens), and FOCUS (GE), and employs 2D spatially selective excitation pulses instead of the 1D excitation pulse of standard EPI DWI; therefore, it allows excitation of only a small inner volume along the phase-encoding direction and reduces the number of phase-encoding steps [35,36,37]. This 2D excitation prolongs the time required for the initial radiofrequency pulse. However, the more recent use of parallel transmission (pTx) with independent radiofrequency transmitter channels enables acceleration of the 2D spatially selective excitation pulse, thus improving the practicality of rFOV-DWI within clinical protocols [38, 39]. Several clinical studies have already investigated the impact of rFOV-DWI for optimizing prostate imaging [36, 37, 40,41,42,43,44]. These studies have reported substantial improvement in image quality with rFOV-DWI, including reduced anatomical distortion and artifacts. However, two previous studies (using b values of 50 and 800–1000 s/mm2 and FOV of 104 × 64 mm2 and 88 × 148 mm2) that assessed tumor detection in PCa reported no improvement in diagnostic performance for rFOV-DWI compared with standard DWI [36, 44]. Tamada et al. considered that the lack of improvement in tumor detection ability using rFOV-DWI may be related to the lower SNR of rFOV-DWI, and the impact of this lower SNR on image contrast such as contrast-to-noise ratio (CNR), if the acquisition time of rFOV-DWI is the almost the same as that of standard DWI [36, 37]. Therefore, further technical optimization of the acquisition method is required to improve SNR and image contrast and thus stabilize image quality in rFOV-DWI. Hausmann et al. has reported that reduced FOV in high b value DWI (b value of 2000 s/mm2) in combination with T2WI could be useful for detecting csPCa [45].

Blurring

Another drawback in ssEPI DWI is blurring due to T2* attenuation, which shows an increasing tendency at higher field strengths [46,47,48]. Although parallel imaging enables a dramatic reduction in blurring [49, 50], it remains a problem in ssEPI DWI, especially in high-resolution images.

DWI with multishot EPI

Multi-shot EPI (msEPI) DWI, in which k-space data are acquired in multiple excitations, can reduce blurring due to shorter shot length, but is sensitive to motion caused by phase differences between shots [51, 52]. msEPI DWI can be acquired using commercially available sequences such as image reconstruction using the image-space sampling function (IRIS) in Philips, read-out-segmented EPI multi-shot (RESOLVE) in Siemens, and multiplexed sensitivity encoding (MUSE) in GE [52,53,54]. In msEPI DWI, multiple excitation is performed in the phase direction for IRIS and MUSE and in the frequency direction for RESOLVE. Regarding the application of msEPI DWI to prostate imaging, a recent study at 3T has reported that subjective image quality taking into account artifacts, delineation of anatomic structures and borders, overall sharpness, contrast, and overall subjective impression; and CNR of PCa and benign tissue, were all higher for msEPI DWI than for ssEPI DWI, but that SNR was lower for msEPI DWI than for ssEPI DWI [55] (Fig. 2). Because distortion and blurring are reduced in msEPI DWI compared with ssEPI DWI, we would expect msEPI DWI to contribute to improved diagnostic performance in local staging for such as extracapsular extension (Fig. 3). However, acquisition time is much longer in msEPI DWI than in ssEPI DWI, which must be improved before msEPI DWI can be applied for clinical prostate MRI.

Fig. 2
figure 2

An 80-year-old male with prostate cancer (PSA level, 13.69 ng/mL; Gleason score, 3 + 4) in the transitional zone. A homogeneous hypointense lesion is seen on T2-weighted imaging (a) (arrow). A focal hyperintensity is depicted clearly on ssEPI DWI (b) and msEPI DWI (c) (arrow) DWI. SNR is higher in ssEPI DWI (b) than msEPI DWI (c), whereas sharpness is better in msEPI DWI (c) than ssEPI DWI (b). CNR between prostate cancer and benign prostate is comparable between the two DWI sequences (b and c). PSA prostate-specific antigen, DWI diffusion-weighted imaging, ssEPI single-shot echo-planar imaging, msEPI multi-shot echo-planar imaging, SNR signal-to-noise ratio, CNR contrast-to-noise ratio

Full size image
Fig. 3
figure 3

An 83-year-old male with prostate cancer (PSA level, 6.6 ng/mL; Gleason score, 3 + 4) in the peripheral zone. No lesion is apparent on ssEPI DWI (a) or msEPI DWI (b), both of which show distortion of the prostate due to rectal air. msEPI DWI (b) is less affected by distortion induced by rectal gas compared with ssEPI DWI (a). PSA prostate-specific antigen, DWI diffusion-weighted imaging, msEPI multi-shot echo-planar imaging, ssEPI single-shot echo-planar imaging

Full size image

Image contrast between benign and malignant tissues

Ultra-high b value DWI (uhDWI)

In prostate mpMRI, high b value DWI (hDWI) with b values of 1500–2000 s/mm2 is commonly used in daily clinical practice [20, 21]. However, even hDWI shows insufficient image contrast between benign and malignant tissues. Therefore, it would be useful to determine whether higher b value DWI could improve the clinical performance of PCa detection. As one possible solution, computed DWI obtained as a calculation image using standard b values such as 800–1000 s/mm2 may be useful for improving the image contrast [20, 21, 56]. A study that used a wide range of computed b values (1000–5000 s/mm2) has demonstrated that those in the range of 1500–2500 s/mm2 are optimal for PCa detection; and that computed b values of 1000 and 3000–5000 s/mm2 exhibit lower performance related to insufficient signal suppression at the low b value and excessive signal suppression leading to diminished anatomic clarity at the higher b values, respectively [57]. A recent study showed that PCa detection rates in computed DWIs with b values of 2000 and 2500 s/mm2 were similar to that of native-acquired DWI with b value of 2000 s/mm2 [58]. Vural et al. showed that lesion detection rates were the same for computed b values of 2000 and 3000 s/mm2, both of which were better than for a computed b value of 1500 s/mm2 [59]. Therefore, the clinical application of computed uhDWI with high b value such as 3000 s/mm2 may not be realistic for improving PCa detection ability. In addition, computed DWI requires dedicated post-processing software as well as extra post-processing time. As a second solution in the acquisition of higher b value DWI that may have potential for the clinical application of uhDWI, improvement in SNR may be gained by state-of-the-art 3T MRI with extremely high gradient waveform fidelity (≈100%), achieved by improved eddy current calibration with maximum gradient strength and maximum slew rate due to precisely calculated coil design and high cooling efficiency. In the study of Zang et al. that evaluated the efficacy of native-acquired uhDWI for PCa detection, native-acquired uhDWI with a b value of 3000 s/mm2 had higher area under the curve (AUC) for PCa tumor detection compared with native-acquired DWIs with b values of 1000 and 2000 s/mm2 [60] (Fig. 4). Thus, such diagnostic performance of uhDWI for PCa detection should be further confirmed in prospective multi-institutional clinical trials with a larger number of patients.

Fig. 4
figure 4

A 79-year-old male with prostate cancer (PSA level, 8.01 ng/mL; Gleason score, 4 + 3) in the posterior left region of the peripheral zone. A homogeneous hypointense lesion with mass effect is seen on T2-weighted imaging (a) (arrow). Signal intensity of the benign whole prostate is lower in DWI with b values of 0 and 3000 s/mm2 (c) than in DWI with b values of 0 and 2000 s/mm2 (b). The lesion is depicted clearly as a focal hyperintensity (arrows) in DWI with b values of 0 and 3000 s/mm2 (c), compared with DWI with b values of 0 and 2000 s/mm2 (b). PSA prostate-specific antigen, DWI diffusion-weighted imaging

Full size image

Quantitative DWI for assessment of tumor aggressiveness in PCa

Clinical impact of accurate assessment of tumor aggressiveness in PCa

PCa can be classified as csPCa, for which curative therapies such as radical prostatectomy are indicated, or as clinically insignificant PCa (cisPCa), for which AS using serum prostate-specific antigen (PSA) is indicated. Therefore, accurate pre-treatment risk stratification of PCa is essential in determining the treatment strategy. In general, csPCa is defined based on histopathology as Gleason score (GS) ≥ 3 + 4, and/or tumor volume ≥ 0.5 cc, and/or extracapsular extension [20]. Among these determining elements of pre-treatment risk stratification in PCa, a tumor GS that reflects tumor aggressiveness would be strongly associated with signal intensity change on DWI and ADC map, as in the DWI scoring system in PI-RADS v2.1 [21] (Table 1). Systematic ultrasonography-guided prostate biopsy, which is a standard method for histopathological evaluation of PCa before treatment in patients with elevated PSA levels has several limitations, including underestimation of indicators of tumor aggressiveness such as the GS of PCa [61, 62]. Therefore, clinical study using quantitative parameters of DWI such as ADC is being actively performed to discriminate PCa aggressiveness.

Table 1 Scoring system of diffusion-weighted imaging for assessment of transition zone and peripheral zone in PI-RADS v2.1
Full size table

DWI with standard mono-exponential model

DWI reflects the Brownian movement of water molecules mainly within extracellular space and is influenced by changes in the relative volumes of glandular, epithelial, and stromal components, as occurs in malignancy [63, 64]. DWI is a key method for tumor detection in prostate mpMRI and also for assessment of tumor aggressiveness in patients with PCa [9, 65]. Previous studies that have employed ADC calculated using a mono-exponential model from DWI acquisitions could discriminate between low-risk and moderate-to-high-risk PCa [9]. However, because there is much overlap between the ADC values of low- and high-grade tumors [66,67,68,69,70], further optimization of DWI methodology is needed to improve the diagnostic performance of diffusion measurements.

The ADC computation assumes Gaussian behavior of water diffusion, whereby water molecules are treated as moving freely and a simple mono-exponential fitting model is applied to characterize the signal decay (Fig. 5). ADC is calculated for a pair of b values (e.g., 0 and 1000 s/mm2) using the following equation:

Fig. 5
figure 5

Mathematical models employed in diffusion-weighted imaging. ADC, apparent diffusion coefficient; K, diffusional kurtosis; IVIM, intravoxel incoherent motion; f, perfusion fraction; D*, perfusion-related diffusion coefficient, D, molecular diffusion coefficient, ɑ, stretching parameter; DDC, distributed diffusion coefficient

Full size image

Sb = S0 ·exp{-b·ADC}.

ADC histogram analysis

ADC histogram analysis is noteworthy as a measure for improving the diagnostic performance of standard ADC, and includes metrics such as the mean, median, various percentile values, skewness as a measure of the asymmetry of the distribution, entropy as a measure of the randomness of the distribution, and kurtosis as a measure of the sharpness of the peak of the distribution [16]. ADC parameters such as the 10th percentile derived from ADC histogram analysis have been found to have higher discrimination ability for tumor aggressiveness compared with the mean ADC value [71, 72]. Lower tumor ADC regions such as the 10th percentile of ADC derived from the ADC histogram analysis may be related to its sensitivity to more aggressive sub-regions within a heterogeneous tumor in PCa [71, 73] that might not be appreciated using conventional metrics such as mean value by volume averaging. A study by Tamada et al. using PCa patients under AS have reported that the mean 0–10th percentile value of 3D whole-lesion ADC histogram analysis in the baseline MRI examination had the best performance for predicting lesion growth on serial MRI examinations, and the change in lesion volume on serial examinations was associated with tumor aggressiveness on follow-up targeted biopsy [16]. Furthermore, histogram analysis can be applied to various other Gaussian and non-Gaussian fitting models as well as to standard ADC. ADC histogram analysis has recently been performed using a simple region of interest (ROI) placement technique on standard Picture Archiving and Communication System (PACS), without using dedicated software such as non-Gaussian fitting models.

Bi-exponential model and non-Gaussian fitting model

It is considered that because water molecule diffusion is obstructed by microstructural complexity (including cellular membranes) in PCa tissues, more complex Gaussian and non-Gaussian models may better reflect the diffusion behavior in PCa, which is characterized by tissue heterogeneity within the tumor [74]. The bi-exponential and non-Gaussian fitting models for which there are abundant research results mainly include intravoxel incoherent motion (IVIM), stretched exponential model, and diffusion kurtosis imaging (DKI).

The bi-exponential fitting and non-Gaussian behavior of diffusion can be investigated using DWI with high b values and with a relatively large number of b values, following recent advances in hardware and software that have enabled the use of higher b values such as 2000 s/mm2 and advanced DWI acquisition and modeling methods [75]. In IVIM and the stretched exponential model, it is desirable to use a large number of b values ranging from very low to high, which might be a more reliable and reproducible method for assessing tumor aggressiveness compared with the mono-exponential model [76,77,78].

IVIM

IVIM is a bi-exponential fitting model that separately evaluates extravascular molecular diffusion and microcirculation of blood within the capillaries [76, 77] (Fig. 5). It is calculated using bi-exponential fitting with multiple b-values, with the following equation:

$${S}_{b}={S}_{0}\left\{F\bullet \mathrm{exp}\left(-b{D}^{*}\right)+\left(1-F\right)\bullet \mathrm{exp}\left(-bD\right)\right\}$$

where D* and F are the perfusion-related diffusion coefficient and microvascular volume fraction, respectively; and D is the molecular diffusion coefficient.

Numerous previous investigations of IVIM DWI have reported comparable diagnostic performance of PCa risk stratification between IVIM (mainly D) and mono-exponential model ADC [70, 79,80,81,82,83,84], whereas only one study has observed better performance of D in IVIM compared with mono-exponential model ADC [85]. Among these studies, two assessed IVIM DWI using b values of 2000 s/mm2 or higher [80, 84]. In contrast, many previous studies have shown insufficient diagnostic performance of f in IVIM [70, 81, 83, 85].

Stretched exponential model

The stretched exponential model reflects the deviation of the curve from mono-exponential behavior (Fig. 5). It is performed with multiple b values, using the following equation:

$${S}_{b}={S}_{0}\bullet \mathrm{exp}\left\{-{\left(bDDC\right)}^{\alpha }\right\}$$

where DDC is the distributed diffusion coefficient, which shows the rate of signal decay with increasing b values; and α is the stretching parameter, which characterizes the deviation of the signal attenuation from the mono-exponential model. A value close to one indicates high homogeneity in apparent diffusion, whereas a low-value result from the non-exponential model is caused by the addition of multiple components.

The diagnostic performance of the stretched exponential model (mainly DDC) for risk stratification of PCa tends to be similar to that of mono-exponential ADC [78, 86, 87]. These studies regarding the stretched exponential model were performed using high b value DWI with multiple b values such as 1500 s/mm2 or higher [78, 86, 87]. Several investigators have reported that the similar diagnostic capabilities for assessment of tumor aggressiveness among mono-exponential ADC, IVIM, and the stretched exponential model may be due to strong correlations between these models, or between GS and these models, which may suggest that these models individually provide similar information in PCa [78, 80,81,82, 86].

DKI

DKI is a non-Gaussian DWI fitting model that is believed to better reflect the microstructural complexity of biologic tissue compared with mono-exponential ADC [88]. Like IVIM and the stretched exponential model, DKI also requires higher b values during acquisition, such as 2,000 s/mm2, and quantifies the deviation of tissue diffusion from a Gaussian pattern as diffusional kurtosis (K) [89] (Fig. 5). An elevation in K indicates greater tissue complexity and deviation from Gaussian behavior, and the K value has been reported to be higher in various malignant tumors than in normal tissues. DKI is calculated with multiple b values, using the following equation:

$$S = S0 \, \cdot {\text{ exp}}\left( { - {\text{b }}\cdot D + {\text{ b}}^{{2}} \cdot D^{{2}} \cdot K/{6}} \right)$$
(1)

where K has no units and represents excess kurtosis relative to a mono-exponential fitting, being 0 for perfectly Gaussian diffusion, and increasing for greater deviation from the Gaussian pattern; and D is an analog of ADC that is adjusted for non-Gaussian diffusion behavior.

Among numerous studies with relatively small sample size that have compared mono-exponential ADC and DKI for assessing PCa aggressiveness [78, 88,89,90,91,92,93], some have observed better performance of either DKI or mono-exponential ADC [88,89,90, 93], whereas others have reported similar diagnostic performance [78, 91, 92]. A previous large study at 3 T that included 285 PCa patients compared the discrimination ability of PCa tumor aggressiveness using radical prostatectomy as reference standard between mono-exponential ADC (with three b values of 0, 500, and 1000 s/mm2) and DKI (with five b values of 0, 500, 1000, 1500, and 2000 s/mm2) for characterization of PCa [74]. ADC and K were highly correlated (r = –0.82), with similar diagnostic performance for GS ≤ 3 + 3 tumors vs. GS ≥ 3 + 4 tumors (AUC 0.744 for ADC and 0.715 for K) and for GS ≤ 3 + 4 tumors vs. GS ≥ 4 + 3 tumors (AUC 0.720 for ADC and 0.694 for K) [75]. In addition, a recent DWI study that used 11 b values (0, 50, 100, 200, 900, 1100, 1400, 1800, 2200, 2500, and 3000 s/mm2) at 3T has reported diagnostic performance of the IVIM and DKI models comparable to that of mono-exponential ADC for prediction of PCa tumor aggressiveness (GS ≤ 3 + 4 tumors vs. GS ≥ 4 + 3 tumors; AUC 0.744 for ADC, 0.732 for D, and 0.766 for K) [80]. Therefore, these metrics appear to be providing, to some extent, overlapping information for the measures of diffusivity and kurtosis [74, 88, 90]. Furthermore, it should be noted that analysis with the non-Gaussian fitting models requires dedicated post-processing software and longer post-processing time than that required for mono-exponential ADC.

Thus, at the present time, it can be considered that no non-Gaussian and bi-exponential fitting models have diagnostic capabilities that clearly outperform mono-exponential ADC. In the clinical setting, the use of simple mono-exponential ADC with histogram analysis may be appropriate for assessing PCa tumor aggressiveness. In addition, recent investigations have reported the potential of mpMRI including mono-exponential ADC-based radiomics models, as a machine-aided approach, for predicting PCa aggressiveness [94,95,96].

Microstructural MRI

An advanced imaging technique based on DWI for mpMRI-invisible csPCa

Although mpMRI is a well-established imaging method that is widely used in conjunction with a standardized reading method (PI-RADS) in prostate imaging, a certain percentage of csPCas are missed as mpMRI-invisible tumors [97,98,99,100]. To address this unmet clinical need, microstructural MRI has been newly developed as a quantitative technique that focuses on structural change in three typical microstructures of prostatic tissue: epithelial cells, stroma, and glandular lumen [101]. Diffusion-based microstructural MRI includes restriction spectrum imaging (RSI), Vascular, Extracellular and Restricted Diffusion for Cytometry in Tumors (VERDICT), hybrid multidimensional MRI (HM MRI), and diffusion-time dependent diffusion MRI (dMRI) using oscillating and pulsed gradient spin-echo sequences (OGSE) [102, 103]. A summary of these diffusion-based microstructural MRI methods is provided in Table 2 (Fig. 6).

Table 2 Diffusion-based microstructural imaging
Full size table
Fig. 6
figure 6

Representation of prostate histopathology in benign tissue (a) and prostate cancer (b). Microstructural MRI focuses on change in four types of prostatic microstructure, represented here as follows: epithelial cells (dark gray areas), stroma (gray area), glandular lumen (white areas), and microvasculature (small black circles)

Full size image

RSI

RSI is a novel diffusion-based technique initially developed for neuroimaging. It uses the data obtained from a broad range of multiple b values obtained in multi-directional diffusion images to model a distribution or a spectrum of isotropic and anisotropic water compartments in tissue [104, 105]. This method might possibly enable isolation of signal from intracellular restricted water, and simultaneously minimize signals from extracellular hindered and free water, which currently hinders conventional DWI [106]. Therefore, as an indicator of signal arising primarily from intracellular water (in other words, cellular components) RSI highlights highly cellular tumors [102]. In a recent study that included a large number of patients with suspected PCa, comparable PCa detection ability was found for RSI and conventional DWI; however, RSI had superior specificity for transition zone (TZ) lesions [107]. Such improvement in diagnostic specificity for TZ lesions by RSI could contribute to prediction of csPCa in PI-RADS category 3 equivocal lesions, which are associated with a low detection rate of csPCa in the TZ [28, 108].

VERDICT

VERDICT is a diffusion-based imaging technique that combines a DWI acquisition with a mathematical model and assigns the diffusion signal to three separate water compartments: microvascular, extracellular–extravascular, and intracellular space [101, 109]. PCa is typically characterized by an increase in the volume of intracellular and microvascular space and a decrease in the volume of extracellular–extravascular space [102]. In a study of 70 patients with suspected PCa or undergoing AC, intracellular volume fraction obtained from VERDICT MRI using b values of 90, 500, 1500, 2000, and 3000 s/mm2 (total imaging time, 12 min 25 s) had a higher AUC for discriminating PCa with a Gleason 4 component from benign tissue and/or PCa with GS = 3 + 3 compared with ADC obtained with standard ADC using b values of 90, 500, 1500, 2000, and 3000 s/mm2 (total imaging time, 5 min 16 s) (AUC 0.93 vs. 0.85, respectively) [110].

HM MRI

Previous studies have shown that T2 and ADC are strongly interdependent, and that distinct populations of water molecules in each voxel with specific paired T2 and ADC values can be identified [104, 105]. HM MRI measures change in ADC and T2 as a function of echo time (TE) and b value, respectively, and uses these changes as a source of information about underlying tissue microstructure components such as stroma, epithelium, and lumen change [102, 111]. HM MRI is acquired using a spin-echo module with diffusion-sensitizing gradients placed symmetrically about the 180° pulse, followed by ssEPI readout and different combinations of TE and b values (TE of 47, 75, and 100 ms; and b values of 0, 750, and 1500 s/mm2) (total imaging time, 8–15 min) [102, 111]. Investigation into the use of HM MRI for the prostate is being actively conducted at the University of Chicago [111,112,113]. In this method, microstructure tissue component volumes are calculated by fitting the HM MRI data to the three-compartment signal model. Distinct paired ADC and T2 values are associated with each compartment, using the following equation:

$$\frac{S}{{S}_{0}}={\sum }_{n=1}^{n=3}{V}_{n}\times \mathrm{exp}\left({-ADC}_{n}\times b- \frac{TE}{{T2}_{n}}\right)$$
(2)

PCas are characterized by increased epithelium volume and reduced lumen and stroma fractional volumes by HM MRI [111]. In addition, it is noteworthy that correlation coefficients were higher between the fraction volumes of tissue components and tumor GS than between T2 values and tumor GS and between ADCs and tumor GS [113].

Thus, the findings of previous studies indicate that VERDICT MRI and HM MRI are expected to improve characterization in PCa, but that continued technical optimization of these advanced MRI sequences is required to shorten the acquisition time before they can be considered for broader clinical application.

dMRI using OGSE

dMRI using OGSE is a novel DWI technique with extremely shortened diffusion time that enables the calculation of microstructural components such as cellularity in prostate tissue, using a mathematical model. The total acquisition time is around 5 min [103]. A recent preliminary study has demonstrated that intracellular fraction and cellularity obtained from dMRI using OGSE had a positive correlation with GS, and AUC of the cellularity for discriminating between low-grade PCa and intermediate-to-high-grade PCa was 0.964 [103]. Although dMRI using OGSE acquired with a clinically acceptable imaging time is expected to have clinical application for PCa characterization, this method suffers from low spatial resolution, which requires improvement; in addition, it is necessary to verify the association between the microstructural MR findings and the pathological findings.

Conclusion

The role of prostate DWI is continuing to increase in the clinical management of PCa in patients with elevated PSA levels, in such as tumor detection, localization, and characterization. Technological innovations in MRI have led to the proposal of various DWI sequences and post-processing technologies as alternatives to standard ssEPI for optimizing qualitative visual assessment in DWI. Regarding the clinical application of quantitative DWI in the risk stratification of PCa, there appears to be no DWI method with bi-exponential fitting model and non-Gaussian fitting model that outperforms mono-exponential ADC with histogram analysis at the present time. In the future, we can expect state-of-the-art technologies and sequences in DWI, including microstructural MRI, to play a more important role in evaluating csPCa.