High-b-value rFOV-syDWIs generated by both 5b-protocol and 13b-protocol DW acquisitions in our study showed better subjective image quality (anatomic details, lesion conspicuity, level of geometric deformation, overall image quality) and better objective assessment (CR between prostate cancer lesions and neighboring tissues) than rFOV-sDWIs when b value was set at ≥ 1000s/mm2, especially when b = 1500 s/mm2(P < 0.05). Moreover, syADC values had equivalent diagnosis performance to sADC values in distinguishing PCa from BPH in spite that some syADC values were different to sADC ones.
FOCUS DWI is viewed as an optimized reduced field-of-view DWI sequence that achieves spatial selective excitation and improves detection of lesion severity, image artifact, image blur, distortion and overall image quality compared to conventional SS-EPI DWI[11].For subjective assessment of rFOV-sDWIs and rFOV-syDWIs in our study, the scores of 5b-protocl and 13b rFOV-syDWIs were closely equivalent to each other and significantly higher than rFOV-sDWIs; rFOV-syDWIs generally had superior scores of lesion conspicuity and level of geometric deformation to rFOV-sDWIs. For objective assessment, CRs of both 5b-protocol and 13b-protocol rFOV-syDWIs were superior to rFOV-sDWIs,and rFOV-syDWIb=1500 had higher CR than rFOV-syDWIb=1000.In consistent with previous articles [16, 19, 20], elevated CR along with increasing b values indicated that synthetic DW image spossessed the advantage of displaying strong lesion-to-tissue contrasts, especially high-b-value DW images. That is, synthetic DWIs met the clinical demands for high-b-value prostatic DWI.
High-b-value synthetic DWIs can be mathematically generated from lower b-value images instead of being directly acquired [14]. An ADC map is calculated from at least two different b-value DW images using a mono-exponential modeland sufficiently distinguishes prostate cancer from normal tissues in clinical practices [21].ADC also plays an important role in cancer diagnosis and invasion status[22–24]; therefore, objective analysis of ADC values is valuable when investigating feasibility of a novel DWI in clinics. In spite that ADC values were easily influenced by selected b-value, SNR, partial volume effect, rFOV-DWI-generated ADC values in characterization of PCa with less distortion and fine lesion details are not statistically different to SS-EPI DW images-based ones [13].Therefore, to compare ADC values to SS-EPI-DWI-acquired ones is straightforward when recruited patient cohorts are highly similar. In our study, significant difference was found between mean and median of syADCb=1500 and sADCb=1500 in PCa but not in hyperplasia, consistent with the explanation of signal attenuating more in normal tissues and hyperplasia than tumor cells when high b value is used [25]. Accordingly, BPH attenuated to the noise level first on rFOV-DWIb=1000 while PCa attenuated to the noise level later on rFOV-DWIb=1500 in our study. Since synthetic ADC maps were computed with scanned b value of 0 s/mm2 and synthetic b value of 1000 or 1500 s/mm2, less noise contributed to high-b-value DW images. Significantly lower syADCmean,b=1000 than sADCmean, b=1000 and statistically higher syADCmean, b=1500 than sADCmean, b=1500 in both cancer and hyperplasia These were attributed to water in the restrictions of complex biological structures, such as organelles, cell membranes and intercellular space, diffuses in a Gaussian displacement when b value is set less than 1000 s/mm2 but in a non-Gaussian distribution when b value is set greater than 1000s/mm2 [26, 27]. In other words, less diffusion kurtosis effect can be found on DWIs when b value is less than 1000s/mm2. In our study, syADCb=1000 presented a less decreasing trend while syADCb=1500 showed significant increasing for cancer tissues, for low-b-value DW images weighed more. The mean and median of synthetic DWIs-computed ADC values might be closer to the scanned DWI-computed ADC values when syADC maps were generated using more relatively high-b-value images.
Histogram-derived ADC values—mean and median of both actually scanned and synthetic ADC values reached perfect diagnosis efficacy with AUC of almost close and equal to 1.000 in our study, implying mean and median of ADC values possess good diagnosis efficacy in detection of prostate cancer and its invasiveness [28, 29].There was no difference of 5b-protocol and 13b-protocol mean and median of syADCb=1000, 1500 in tumor tissues. In accordance to a study of synthetic high-b-value DWIs in diagnosis of cervical cancer[30], the diagnosis efficacy of both 5b-protocol and 13b-protocol syADCmean, b=1000, 1500 was rather competitive. This result might attribute to several reasons. First, there was obviously no overlap of ADC values for prostatic cancer and hyperplasia. ADC values have shown significantly lower in the cancer region than normal tissues [31, 32].Second, three ROIs were placed at the tumor core on the maximal cross-section slice of tumors, leading to lower ADC values than those obtained using the whole-tumor ROIs[33].Third, only patients who were diagnosed as Gleason score ≥ 7 were recruited in our study. Gleason score rises with decreasing ADC and reflects the significantly higher heterogeneity and complexity in PCa than BPH and the level of focal invasiveness and pathological grade[34–36]. 5b-protocol synthetic DWI is useful in clinical work with the pros of short scan time, good-image-quality, high-b-value DWIs and satisfactory diagnosis performance. The other two histogram-derived ADC values—kurtosis and skewness reflect cellularity and degree of heterogeneity; kurtosis > 0.5 and positive skewness value respectively means low differentiated cells and malignant tissues while kurtosis < 0.5 and negative skewness value respectively represents massive edema and cystic tissues [37–39]. Skewness and kurtosis of both actually scanned and synthetic ADC values in our study showed poor diagnosis performance, consistent with most research [40–41].Tissue inhomogeneity in regional tissues such as various cell density in part of prostatic lesions leads to magnificently different ADC values. ROIs were also placed in avoidance of necrotic, hemorrhagic and cystic lesions. Thus, measured skewness and kurtosis showed less sensitive to spatial distribution of proliferated cells. Therefore, it should be cautious to use histogram-derived synthetic ADC parameters — skewness and kurtosis in differentiating PCa from BPH. Interestingly, we found statistically lower 13b-prtocol kurtosis of syADCb=1000than sADCb=1000 in PCa as well as lower 13b-protocol skewness and higher 13b-protocol kurtosis of syADCb=1000 than sADCb=1000 in BPH. This could be explained that 13b-prtocol less-kurtosis syADCb=1000 was generated with six lower b values = 0, 25, 50, 75, 100, 150s/mm2 compared to 5b-prtocol kurtosis syADCb=1000 that was generated with only two lower b value = 0,100 s/mm2.Kurtosis effect was less pronounced on 13b-syADCkurtosis, b=1000but not on 5b-syADCkurtosis, b=1000 compared to 13b-sADCkurtosis, b=1000in accordance with a DWI study using b-value less than 200 s/mm2[42]. syADCs might provide more closely to the results with the dominantly diffusion-only effect, but our sample size should be expanded to validate this finding due to poor repeatability of skewness[43].
The possible reasons and limitations in our study were as follows. First, our study cohort and diversity were relatively small and this study was carried out in only one single institution. In this study, all patients with Gleason ≥ 7 prostate cancer lead to our results deviated from generalization. As in previous prostate studies, the maximum cross-sectional area was used to sketch lesion ROIs and areas such as hemorrhages, necrosis, or cysts were excluded so that target ROIs showed less heterogeneous and led to lower differentiation power. More different pathological subtypes and delineation approaches should be explored in the future to conduct a large cohort study. Secondly, synthetic high-b-value images were conducted only on 1.5 T MRI. Direct comparison of computed DWI between different magnetic fields might be impractical and should be verified especially ADC values in disease diagnosis in spite that there is no significant difference in ADC values between cervical cancer and abdominal organs between 1.5 and 3.0 T[44].Synthetic higher-b-value DWI showed less kurtosis effect and more different to scanned DWIs. We expected synthetic b-value DWI could differentiate prostate cancers with distinct Gleason levels in the future due to our present observation of decreased signal intensity with the increasing of used b value for both rFOV-sDWIs and rFOV-syDWIs. Last but not the least, synthetic formula constrained the extrapolated b-value DWI and thus no image quality and diagnosis performance of higher b-value synthetic ADC map were validated. This work will be extended in the future.
To sum up, a synthetic DWI with reliable synthetic ADC values on 1.5T MRI scanner had clinical potential in diagnosis of PCa and BPH. The synthetic DWI itself and its derived 5b-protocol scheme application shortened scan time at least half of convectional 13b-protocol scheme and provided better image quality and reliable ADC values.