Diagnostic performance of 3-tesla multiparametric MRI for assessment of the bladder cancer T stage and histologic grade

Preoperative staging of urinary bladder carcinomas using TNM system is crucial in the management of bladder cancer which is determined mainly by stage and grade of tumor at diagnosis. We aim to evaluate the diagnostic accuracy of multiparametric MRI (mp-MRI) for assessment of the bladder cancer T stage and histologic grade. The overall T2-WI diagnostic accuracy for the T stage was 72.3%, increased to be 87.1% for contrast-enhanced images, and 92.6% for DWI, reaching the maximum accuracy 94.5% using the combined multi-parametric MRI technique. Diagnostic accuracies of mp-MRI in differentiating superficial from muscle-invasive (91%) and organ-confined from non-organ confined tumors (92%) were superior to DW-MRI (89% and 87%), DCE-MRI (84% and 83%), and T2W-MRI (74% and 71%), respectively. The agreement between MRI findings and histopathological staging was greater in mp-MRI (k = 0.91; excellent agreement) than in DW-MRI (k = 0.77; moderate agreement), DCE-MRI (k = 0.76; substantial agreement), and T2W-MRI (k = 0.53; fair agreement). Mp-MRI provides useful information for evaluating the local T stages of bladder cancer and can predict the histological grades of urinary bladder cancers with high diagnostic accuracy.


Background
Clinical management of urinary bladder cancer is determined primarily on the basis of distinguishing superficial tumors (stage T1 or lower) from invasive ones (stage T2 or higher) because the treatment options differ considerably [1]. Superficial tumors are treated with transurethral resection (TUR) with or without adjuvant intravesical chemotherapy or photodynamic therapy [2], whereas invasive tumors are treated with radical cystectomy, radiation therapy, chemotherapy, or a combination [3].
The most important point for the local staging of bladder cancer is the presence or absence of muscle invasion, which has a significant impact on the management strategy as outlined in multiple guidelines including those issued by the European Association of Urology (EAU), the National Comprehensive Cancer Network (NCCN), and the American Urological Association/Society of Urologic Oncology (AUA/SUO) [4,5].
In high-resolution T2 sequence, urine has high signal intensity, tumors have intermediate signal intensity, and normal detrusor muscle appears as a hypointense line in-between, so in non-muscle-invasive disease, the low signal intensity detrusor lining adjacent to the tumor is preserved, whereas it is interrupted in case of muscleinvasive disease [4,7,8].
Diffusion-weighted imaging is a noninvasive functional imaging technique reflecting proton diffusion properties in water. It depends mainly on the Brownian motion through the inhibitory effect of cell membranes to the random motion of water molecules [9]. Changes in ADC are inversely correlated with changes in cellularity; thus, in tissues with high cellularity, decreasing water mobility will be reflected as decreased ADC values. Besides tumor localization and staging, DWI is considered a promising tool in assessment of tumor aggressiveness [10].
Dynamic contrast-enhanced MRI is useful to assess angiogenesis because tumor enhancement is attributed to pathological neovascularization and so malignant neoplasms show earlier enhancement than tissues without neovascularization and this presented as rapid contrast wash-in followed by washout pattern in case of malignant neoplasms [11].
Multiple studies have proved that 3-T MRI shows a better specificity and sensitivity than 1.5-T MRI when used to diagnose and stage bladder cancer. Therefore, the guidelines recommend the use of multi-parametric 3-T MRI to improve the diagnostic accuracy for determining bladder cancer T stage [1].
We hypothesized that mp-MRI can accurately assess the local stage and grade of bladder cancer. So, we conducted this study to prospectively evaluate the usefulness of mp-MRI for determining T stage for bladder cancers and to measure the correlation between mp-MRI and histopathological grade.

Subjects
This prospective study included 54 consecutive patients with bladder tumors diagnosed either clinically, from urine cytology or by other radiologic investigations including ultrasonography or computed tomography during the period from March 2017 to March 2019. Patients were referred to the Radiology Department (MRI Unit) from the Urology Department of the same center. Exclusion criteria included patients with general contraindications for MR examination (as with pacemaker or metallic prosthesis) or contraindications for cystoscopy (patients unfit for anesthesia or due to urethral stricture), patients with high renal function tests, not suitable for IV-gadolinium injection, and patients refusing consent. Adequate bladder distention is important and can be achieved by instructing the patient not to void for about 1-2 h before imaging or by instructing the patient to start drinking 500-1000 ml of water 30 min before the examination.

MRI technique
MRI was performed in the supine position using a high field system (3-Tesla) MRI scanner (Phillips, ingenia 3 T, the Netherlands) with 18-channel phased external array coil. Each patient was subjected to three main components of mpMRI of the pelvis including high-resolution T2W, diffusion-weighted imaging with ADC map, and DCE MRI. Additional sequences were done including T1W (with and without fat suppression) and axial T2 with fat suppression. Examination protocol was approved by the institutional board of ethics, and informed consents were obtained from all patients.

Image analysis
All scans were transferred and reviewed on a picture archiving and communication systems (Magic View, GE, Milwaukee, WI, USA). T2-WI, DWI, and DCE-MR images were interpreted referring to T1 and T2-WI image sets, and the tumors were assigned a T stage on a stageby-stage basis in each sequence. When multiple tumors were present, the highest T stage represented the T stage of the patient.
In T2-weighted, the signal intensity of bladder cancers is lower than the signal from the urine and perivesical fat and higher than the wall; tumors are staged as follows: Stage T1: Preservation of low signal intensity of the bladder wall. Stage T2: Interruption of the hypointense bladder wall in tumor site. Stage T3: Infiltration of the perivesical fat. Stage T4: Tumor extending into the adjacent organs, abdominal or pelvic wall.
In DW-MRI, signal intensity of the lesions is identified at b1000 images; in these images, the tumor is hyperintense and the bladder wall appears slightly hyperintense. While analyzing the DW MRI, the image interpretation was referred to T1 and T2 images and lesions are classified into the following: Stage T1: Hyperintense tumor is within the bladder lumen. Stage T2: Hyperintense tumor partially seen in bladder wall. Stage T3: Hyperintense tumor disrupting the bladder wall. Stage T4: Hyperintense tumor extending into the adjacent organs, abdominal or pelvic wall.
In DCE-MRI, bladder tumor, mucosa, and submucosa enhanced early, but the muscle layer maintains its hypointensity and enhanced late. Lesions are classified as follows: The time-intensity curves were constructed from signal intensity values obtained from regions of greatest enhancement selected from the lesions as follows: Grade1: Enhancement, followed by a slow increase. Grade 2: Enhancement, followed by a plateau. Grade 3: Enhancement, followed by washout.
ADC values were calculated from the most hypointense areas of the tumors, corresponding normal bladder walls and urine.
Data were regrouped to evaluate the accuracy of each separate imaging set and mp-MRI in distinguishing nonmuscle-invasive from muscle-invasive tumors and organ-confined from non-organ-confined tumors with comparison to the post-operative histopathological data for tumor type, grade, and stage.

Statistical analysis
Qualitative data were described using number and percent. Quantitative data were described using range, mean, standard deviation, and median. The used tests were chi-square test, Fisher's exact test, and Student's t test. Kappa statistic was applied to compare the extent of agreement between MRI findings and histopathological results. Receiver operating characteristic curve (ROC) analysis was used to calculate the area under the curve (AUC) for identification of ADC cutoff values differentiating normal values from bladder cancer and differentiating low-grade from high-grade cancers. Diagnostic efficacy was depicted in terms of sensitivity, specificity, negative predictive value, positive predictive value, and diagnostic accuracy. The analysis was performed by SPSS version 25 software. A p value < 0.05 was considered as significant.
So, the extent of harmony between radiologic and histopathological staging results was greatest when using mp-MRI denoting high diagnostic performance.
As regards the time-signal intensity curve of the DCE-MRI, there were three types of curves identified in our study: type I = ascending pattern, type II = plateau pattern, and type III = descending pattern. There were significant statistical correlations (P < 0.001) between the type of curve and the tumor grade by using the Monte Carlo test.
As regards the mean ± SD of the aberrant diffusion coefficient (ADC) values of the detected bladder tumors in our study, it was 0.813 ± 0.24 × 10 −3 mm 2 /s in urothelial carcinoma (no = 47), 1.02 ± 0.21 × 10 −3 mm 2 / s in SCC, and 0.92 ± 0.24 × 10 −3 mm 2 /s in adenocarcinoma. There is no statistically significant correlation between tumor ADC value and tumor subtypes. There is no statistically significant correlation between the mean ADC value of the tumor subtypes, normal urinary bladder wall ADC, and urine ADC. Also, there was no statistically significant correlation between the ADC value and tumor stage (Table 3 and Fig. 1).
The mean of ADC values of lesions (no = 54) was compared to the grade of tumors. ADC values were significantly lower in high-grade tumors and higher in lowgrade tumors (Table 4 and Fig. 2).
A receiver operating characteristic (ROC) curve based on ADC values demonstrated a perfect AUC of 0.959. A cutoff ADC value of 0.941 × 10 −3 mm 2 /s best differentiated high-grade from low-grade carcinomas with a sensitivity of 92.3%, a specificity of 90.2%, PPV of 75%, and NPV of 97.4% (Fig. 3).

Discussion
Preoperative staging of urinary bladder carcinomas using TNM system is crucial in the management of bladder cancer which is determined mainly by stage and grade of tumor at diagnosis [1,12].
TURBT is the current standard for determining the presence or absence of muscle invasion, provides an estimate of pathologic subtype and grade, and can be completely curative in some times if the entire tumor is resected [12,13]. However, TURBT has been found to underestimate T stage in up to 40% of patients, is inaccurate at determining tumor grade in up to 15% of patients, and frequently needs to be repeated [3].
Improving MRI technology had led to introduction of mp-MRI (including high-resolution T2 WI, DW MRI, and DCE MRI) providing both anatomic and functional evaluation of the local staging and grading of bladder cancer with relatively high accuracy [6].
The overall T2 staging diagnostic accuracy in the current study was 72.3%, which is higher compared to previous similar studies. In the study conducted by Takeuchi et al. [14] on 52 patients with bladder tumor using 1.5-T MRI, they found that the overall diagnostic accuracy of T2WI in local T staging of bladder tumor was 67%. A study by Afifi et al. [15] using 1.5-T magnet, conducted on 50 patients with bladder cancer, showed T2 staging diagnostic accuracy of 52%. Another interesting study by Abou-El-Ghar et al. [16] using 1.5-T MRI conducted on larger number of patients (130 patients) showed T2 staging diagnostic accuracy of 39.6%. The improved diagnostic accuracy of our study may be attributed to the use of 3-T MRI, as the most of the previous studies were done on 1.5-T magnet. The current study revealed T2 diagnostic accuracy of 71% in differentiating superficial from invasive tumors. This was better than Takeuchi et al. [14] who reported a diagnostic accuracy of 79% and Barsoum et al. [17] who reported a diagnostic accuracy of 88%.
Our study revealed T2 diagnostic accuracies of 85.2% in differentiating organ-confined from non-organconfined tumors. This was similar to Takeuchi et al. [14] who reported a diagnostic accuracy of 85% and less than Barsoum et al. [17] who reported a diagnostic accuracy of 94%.
In the current study, overall staging accuracy of DCE-MRI was 87% and overstaging was detected in 13% of patients. Our results were better than those reported by Afifi et al. [15] who reported accuracy of (74%) and overstaging of about 26%. In their study, Gupta et al. [6] detected less staging accuracy of DCE MRI (73.3%) and more overstaging (20%) compared to our results. Also, Takeuchi et al. [14] reported accuracies between 75 and 92% which is comparable to our results.
To our knowledge, the accuracy of dynamic contrastenhanced MR images for differentiating between T1 or lower tumors and T2 or higher tumors has been reported to be 75-92% [14,18,19], and it was reported to be 75-92% for differentiating between T2 or lower tumors and T3 or higher tumors. The overall accuracy for diagnosing tumor stage has been reported to be 52-93% [14,18].
Our study revealed a significant statistical correlation (P < 0.001) between grading, time-intensity curves, and corresponding histopathological grades. Grade 1 curve only identified in low-grade tumors (3.7%), grade 3 curve  mostly identified in high-grade tumors (75.9%), while grade 2 curve were more identified in low-grade (72.7%) than in high-grade carcinomas (27.3%). This was in agreement with Afifi et al. [15] and Gupta et al. [6] who showed significant correlation between time-intensity curves and tumors grading. Quantitative analysis of the ADC values potentially reflects the histological grades of urothelial tumors. To our knowledge, DWI was superior to DCE-MRI and T2W-MRI in detection and local staging of urinary bladder cancers [13,20].
The overall staging accuracy of DW-MRI in current study was 92.6% and overstaging was detected in 7.4% of patients. Our results are better than results reported by Gupta et al. [6] who detected less DWI staging accuracy (76.7%) and more overstaging (16.7%) and also Afifi et al. [15] who showed an overall staging accuracy of DW-MRI of 82% and overstaging in 18% of patients.
Our study revealed DWI diagnostic accuracies of 96.3% in differentiating both superficial from muscleinvasive and organ-confined from non-organ confined tumors. These results were higher than El-Assmy et al. [21] results who found the DWI staging accuracies of 63.6% and 69.6% in differentiating superficial from muscle-invasive and organ-confined from nonorgan confined tumors, respectively. On other hand, our results were similar to Barsoum et al. [17] results who reported the accuracies of 96% and 98%, respectively.
In the current study, the correlation between the radiologic and pathologic stages was greater with the DWI (q = 0.766) than with the T2WI (0.527) or gadolinium-enhanced (0.758) which was in agreement with results reported by Watanabe et al. [22] and Gupta et al. [6]. Similar results also were detected in Afifi et al. [15] study who also found greater correlation between the radiologic and pathologic stages with the diffusion sequence (q = 0.679) than with the T2W (0.274) or gadolinium-enhanced (0.566).
By comparing multiparametric acquisition protocols, we found that the combination of T2-weighted, DWI, and DCE-MRI was most accurate for staging. Overall staging accuracy of mp-MRI in current study was 94.5% and overstaging was detected only in 5.9% of patients. These results were better than study of Afifi et al. [15] who reported mp-MRI overall staging accuracy of 88% and overstaging was detected only in 12% of their patients. Takeuchi et al. [14] also stated that theT2-WI diagnostic accuracy for the T stage was 67%, increased to be 88% for T2WI plus DWI, 79% for T2 WI plus contrast-enhanced images, and 92% for all three image types reaching the maximum accuracy using the combined multi-parametric assessment.
Our study revealed mp-MRI diagnostic accuracy of 94.4% in differentiating superficial from invasive tumors. This was better than van der Pol et al. [1] who reported diagnostic accuracy of 84 to 86% and Afifi et al. [15] who reported diagnostic accuracy of 88%. Our study revealed mp-MRI diagnostic accuracy of 98.1% in differentiating organ-confined from non-organconfined tumors. This was better than van der Pol et al. [1] who reported diagnostic accuracy of 81 to 83% and Afifi et al. [15] who reported diagnostic accuracy of 88%.
In this study, we found that the mean ADC value of the detected malignant bladder lesions was 0.813 ± 0.24 × 10 −3 mm 2 /s, which is significantly lower than that of urine (mean 2.71 ± 0.35) and that of normal bladder wall (mean 1.56 ± 0.41). This was similar to Afifi et al. [15] who reported mean ADC value in bladder cancers of 0.87 ± 0.26 × 10 −3 mm 2 /s, significantly lower than that of urine (mean 2.90 ± 0.25) and that of normal bladder wall (mean 1.57 ± 0.17). Our results also agreed with Kobayashi et al. [23] study.
In current study, we detected a cutoff ADC value of < 1.36 × 10 −3 mm 2 /s differentiating bladder carcinomas from normal bladder wall and a cutoff ADC value of 0.941 × 10 −3 mm 2 /s differentiating high-grade from low-grade carcinomas. Similar results were detected by Afifi et al. [15] who found that a cutoff ADC value of < 1.36 × 10 −3 mm 2 /s differentiating bladder carcinomas from normal bladder wall and a cutoff ADC value of < 1.012 × 10 −3 mm 2 /s differentiating high-grade from low-grade carcinomas.
On the other hand, Kobayashi et al. [23] stated a cutoff ADC value of 0.86 × 10 −3 × 10 −3 mm 2 /s differentiating aggressive from low-grade disease which is lower than cutoff value detected in our study. Avcu et al. [20] stated a cutoff ADC value 1.545 × 10 −3 mm 2 /s differentiating malignant and benign bladder wall pathologies and a cutoff ADC value 1.135× × 10 −3 mm 2 /s differentiating high and lowgrade carcinomas which is higher than our results.
We detected inverse correlation between bladder carcinomas ADC values and T stages, as ADC values were significantly lower in high stage tumors P T2 (mean 0.79 ± 0.28 × 10 −3 mm 2 /s) than the superficial tumorsT1 (mean 1.04 ± 0.26 × 10 −3 mm 2 /s). This was similar to Afifi et al. [15], who reported that the ADC values were significantly lower in high stage tumors P T2 (mean 0.69 ± 0.13 × 10 −3 mm 2 /s) than the superficial tumors T1 (mean 0.97 ± 0.26 × 10 −3 mm 2 /s). Similar to our results, Kobayashi et al. found that patients with higher T stages exhibited significantly lower ADC values (P < 0.001).
Our study had a number of limitations; the distribution of T stages was uneven, with a large number of T2 and T3 and a small number of T1 and T4 tumors; therefore, larger studies are warranted to more fully define the role of mpMRI in T staging of bladder tumors. Limitations of the current study included increased cost, time consumed by the technique especially including the cost of the contrast, and some patients were less cooperative with motion artifacts as the study mandates full UB. Finally, the radiologic-pathologic correlation was not perfect because the cut surfaces on surgical specimens were not always completely identical to those seen on MR images.
It is expected that utilization of mp-MRI for bladder cancer local T staging will increase as awareness of its accuracy in the urology community improves. Several recent meta-analyses assessing the topic of local staging of bladder cancer with mp-MRI confirm an increase in the potential benefits of performing mp-MRI [1].

Conclusions
Mp-MRI is an excellent tool for determining the different local T stages and the histological grades of urinary bladder cancers with high diagnostic accuracy. So, before cystoscopy and histopathological sampling, more accurate predictions can be made using Mp-MRI.