New RESOLVE-Based Diffusional Kurtosis Imaging in MRI-Visible Prostate Cancer: Effect of Reduced b Value on Image Quality and Diagnostic Effectiveness

Transcription

1 Genitourinary Imaging Original Research Zhang et al. Diffusional Kurtosis Imaging of Prostate Cancer Genitourinary Imaging Original Research Yu-Dong Zhang 1 Chen-Jiang Wu 1 Mei-Ling Bao 2 Hai Li 2 Xu Yan 3 Xi-Sheng Liu 1 Hai-Bin Shi 1 Zhang YD, Wu CJ, Bao ML, et al. Keywords: b value, diffusional kurtosis imaging, MRI, prostate cancer, readout segmentation of long variable echo-trains, RESOLVE DOI: /AJR Received December 4, 2015; accepted after revision February 26, Department of Radiology, the First Affiliated Hospital with Nanjing Medical University, 300 Guangzhou Rd, Nanjing, China, Address correspondence to H. B. Shi (njmu_shb@163.com). 2 Department of Pathology, the First Affiliated Hospital with Nanjing Medical University, Nanjing, China. 3 MR Collaboration NE Asia, Siemens Healthcare, Shanghai, China. Supplemental Data Available online at AJR 2016; 207: X/16/ American Roentgen Ray Society New RESOLVE-Based Diffusional Kurtosis Imaging in MRI-Visible Prostate Cancer: Effect of Reduced b Value on Image Quality and Diagnostic Effectiveness OBJECTIVE. The purpose of this article was to investigate whether a new readout segmentation of long variable echo-trains (RESOLVE) based diffusional kurtosis imaging (DKI) with reduced b value technique can affect image quality and diagnostic effectiveness in MRI-visible prostate cancer (PCA). SUBJECTS AND METHODS. Prostatic RESOLVE DKI ( s/mm2) was prospectively performed for 12 volunteers. The optimal protocol was then performed in 108 MRI-visible PCAs to determine whether it can compete against a preferred b-value set ( s/mm 2 ) regarding image quality and diagnostic effectiveness. Images were interpreted by two independent radiologists using the prostate imaging reporting and data system (PI-RADS). Readers concordance and diagnostic effectiveness were tested with the Fleiss kappa and area under the ROC curve (A z ) analyses. RESULTS. A b value of 1400 s/mm 2 generated a larger apparent diffusion coefficient of gaussian distribution (Dapp) (1.35 ± 0.31 vs 1.30 ± 0.30 mm 2 /s; p < 0.001) and apparent kurtosis coefficient (Kapp) (1.11 ± 0.26 vs 1.00 ± 0.21; p < 0.001) in PCA than did a b value of 2000 s/mm 2. Interreader agreement using PI-RADS was relatively low when Dapp and Kapp maps were excluded from image interpretations (κ = vs κ = with Dapp and Kapp maps). Interreader agreement in staging PCA was relatively high (κ > 0.80) and was not influenced by reducing the b value. The power of Dapp and Kapp to differentiate PCA from normal tissue (A z = ), tissue with a Gleason score less than or equal to from tissue with a Gleason score greater than (A z = ), and PCA stage lower than pt3 from stage pt3 and higher PCA (A z = ) was not significantly degraded by reducing the b value. CONCLUSION. We found that b values significantly influenced image quality, PI-RADS score, and DKI outputs but did not degrade the diagnostic effectiveness of DKI parameters to detect and classify PCA. P rostate cancer (PCA) is the most common malignancy in men and the second leading cause of cancer deaths in developed countries [1, 2]. To noninvasively determine the clinically significant cancer and preoperatively assess the pretherapeutic risk is crucial for patient management decisions and is helpful for improving active surveillance in patients with PCA [3 5]. DWI is an important MRI technique that reveals special biologic information of intravital tissues [6 8]. Diffusional kurtosis imaging (DKI) is a more recently described metric for optimization of DWI analysis [9]. The DKI model, which works on the assumption that water diffusion is nongaussian in behavior, is more likely to be useful for tissue with microstructural complexity [10, 11]. The application of DKI in prostate imaging has been gaining increasing interest in recent years [12 16]. It has resulted in several studies showing that DKI has better performance in lesion detection and classification of PCA compared with conventional DWI [12, 15, 16]. However, with different imaging parameters, discrepant postprocessed values were reported in those studies. An important reason for the difference is the choice of b values. The optimal b values for prostatic DKI in the literature were s/mm 2, with four to seven b values [12 16]. Theoretically, increasing the b value may be more beneficial for robust estimation of DKI parameters [17, 18], which will, unfortunately, require more acquisition time. Readout segmentation of long variable echo-trains (RESOLVE) is a newly devel- AJR:207, August

2 Zhang et al. oped sequence for DWI acquisition that has been associated with increased image quality and higher lesion conspicuity than conventional echo-planar imaging (EPI) sequences [19, 20]. However, the performance of prostatic DKI using RESOLVE may face up to the challenges regarding long acquisition time and increasing noise scales when multiple and high b values are used [17, 21, 22]. To optimize a DKI protocol that can make a compromise between image quality and time consumption is, thus, necessary. However, there is not yet a standard consensus regarding the optimization of the number, range, and position of b values for current prostatic DKI using RESOLVE. The aim of this study was thus to introduce an optimized DKI protocol using RESOLVE that can be readily applied to clinically indicated pelvic MRI examinations for the diagnosis of PCA. We first optimized the new approach in a prospectively performed volunteer study and then validated whether the optimal approach can make a compromise on accuracy regarding DKI output and diagnostic ability compared with a preferred b value set in a cohort of patients with PCA. Subjects and Methods Patients This single-institution study was approved by the institutional review board of Nanjing Medical University. The study included two groups of subjects: group 1 included 12 volunteers (age range, years; mean age, 63.5 years) with clinically suspected prostate disease, such as benign prostatic hyperplasia, prostatitis, or PCA. The volunteers were informed that they would undergo a specialized DKI protocol for research purpose. Written informed consent was obtained from them. Image data in the volunteer group were used to determine an optimal b value set for prostatic RESOLVE DKI. On the basis of this work, a RESOLVE DKI protocol optimized with b values of 0, 700, 1400, and 2000 s/mm 2 was applied to our routine pelvic MRI examination protocol (for more details, see the supplemental data, which can be viewed in the AJR electronic supplement to this article, available at Group 2 included 108 patients with biopsy-confirmed PCA diagnosed between January 2013 and October All patients included in this group underwent a preoperative prostate MRI examination (including the optimized RESOLVE protocol) and underwent a single session of radical prostatectomy (RP) therapy within 2 weeks after the MRI. Because this imaging procedure did not add additional economic burden and was noninvasive, the institutional review board waived the requirement for informed consent from these patients. Prostatic MRI One hundred eight patients with biopsy-confirmed PCA underwent preoperative MRI examination using the following imaging protocols: axial T1-weighted turbo spin-echo (TR/TE, 700/14; flip angle, 180 ; section thickness, 3.5 mm; intersection gap, 0.5 mm; FOV, 25 cm 2 ; and matrix, ) and axial, coronal, and sagittal T2-weighted turbo spin-echo (TR/TE, 6000/124; flip angle, 180 ; section thickness, 3.5 mm; intersection gap, 0.3 mm; FOV, 25 cm 2 ; and matrix, ). For axial DKI sequences using RESOLVE (TR/TE, 4000/76; flip angle, 90 ; section thickness, 3.5 mm; intersection gap, 0.3 mm; FOV, 25 cm 2 ; and matrix, ), diffusion was measured using b values of 0, 700, 1400, and 2000 s/mm 2, and the number of readout segments was three with two repeated segments, producing a total scanning time of 7 minutes 52 seconds. Finally, dynamic contrast-enhanced (DCE) imaging (3D- FLASH; TR/TE, 3.96/1.46; flip angle, 12 ; section thickness, 3.5 mm; FOV, 25 cm; and matrix, ) was performed with the IV injection of gadopentetate dimeglumine (0.1 mmol/kg; Magnevist, Schering); the scanning period was 5 minutes. Diffusion images were postprocessed using FireVoxel software (version 201, Center for Advanced Imaging Innovation and Research [CAI 2 R]). For the DKI model, the relationship between signal intensity and b values can be expressed as follows: 1 s (b) = s 0 e ( bdapp + b 2 Dapp 2 Kapp) 6 (1), where s is the signal intensity (arbitrary units), b is the b value in seconds per square meter, Dapp is the apparent diffusion coefficient (ADC) of gaussian distribution ( 10 3 mm 2 /s), and Kapp, a dimensionless parameter, is the apparent kurtosis coefficient. Two combinations of b values were selected for DKI calculation: 0, 700, and 1400 s/mm 2 were the minimized b values for DKI, and 0, 700, 1400, and 2000 s/mm 2 were the preferred b values for DKI. Data Assessment Two readers experienced in interpreting crosssectional images were recruited to independently assess the four datasets according to different imaging combinations: dataset A included T2-weighted, DCE, and DW images with a b value of 1400 s/mm 2 ; dataset B included T2-weighted, DCE, and DW images with a b value of 2000 s/mm 2 ; dataset C included T2-weighted, DCE, and DW images with a b value of 1400 s/mm 2 and corresponding Dapp and Kapp maps; and dataset D included T2-weighted, DCE, and DW images with A B C Fig. 1 MR images of two representative cases to indicate imaging registration, ROI drawing, and identification of leading lesion. A and B, 63-year-old man who presented with solid tumor (pathologic Gleason score 4 + 4, pt2c) in left peripheral zone (PZ; green outline). DW image (A) was obtained with b value of 2000 s/mm 2, and registration and fusion of images between T2-weighted image and diffusional kurtosis imaging was performed using DICOM-tag metrics (B). C, 71-year-old man who presented with two solid tumor foci in right PZ (pathologic Gleason score 4 + 3; green outline) and right transition zone (TZ; pathologic Gleason score 4 + 4; red outline). Because tumor in left TZ had higher pathologic Gleason score and larger size than right PZ lesion, it was selected as leading lesion for quantitative measure. 2 AJR:207, August 2016

3 Diffusional Kurtosis Imaging of Prostate Cancer Fig year-old man with multiple tumor foci within prostate (pathologic Gleason score 4 + 5, pt3b). Cancer showed diffusional hypointense signal intensity on axial T2-weighted image and ill-defined boundary; seminal vesicle invasion was seen on sagittal T2-weighted image (not shown). Images show comparison of evaluation with diffusional kurtosis imaging (DKI) using different b values. A and B, T2-weighted DKI with b values of 1400 s/mm 2 (A) and 2000 s/mm 2 (B) indicated multitumor foci (white outlines) within prostate (red outline). Small lesion (arrowhead) located at left-anterior peripheral zone was invisible on image obtained with b value of 2000 s/mm 2 (B) but was clearly seen on image obtained with b value of 1400 s/mm 2 (A). Numbers on scale are arbitrary units. C and D, Apparent diffusion coefficient of gaussian distribution (Dapp) maps corresponding to DKI performed with b values of 1400 s/mm 2 (C) and 2000 s/mm 2 (D). By using Dapp maps, lesion (arrowhead and white outline) was scored as Prostate Imaging Reporting and Data System category 3 on both DKI images. Difference in Dapp maps under two b value sets was not statistically significant. Numbers on scale indicate Dapp values ( 10 3 mm 2 /s). E and F, Apparent kurtosis coefficient (Kapp) maps corresponding to DKI performed with b values of 1400 s/mm 2 (E) and 2000 s/mm 2 (F). Difference in Kapp values was remarkably manifested because low b value set produced higher Kapp values. Numbers on scale are arbitrary units. Arrowheads denote lesion. a b value of 2000 s/mm 2 and corresponding Dapp and Kapp maps. Image data were presented to the radiologists in alphabetic order. To avoid reading bias, the reading interval between each dataset was 2 weeks, and the reading order for each patient was different for each dataset. The four representative datasets were displayed simultaneously without annotations (e.g., patient names or b values) and were explored slice by slice using FireVoxel. The software allows spatial registration and fusion of images between T2-weighted images and DKI using DICOM-tag metrics (Fig. 1). The readers independently reviewed the sole or fused images using Prostate Imaging and Reporting and Data System (PI-RADS) version 2 [23]. In this procedure, the prostate was divided into 12 regions for purposes of review, with eight in the peripheral zone (PZ; right and left regions at the level of mid gland, each in turn divided into anterior and posterior portions) and four in the transition zone (TZ; right and left regions at the level of mid gland). The following imaging features were recorded: the presence or absence of extracapsular extension (ECE) and seminal vesicle invasion (SVI) and a separate PI-RADS score from 1 to 5 for each region. To facilitate the image screening, the 5-point PI-RADS score was simplified to three scales: score 1 2 was coded as benign, score 3 was coded as uncertain, and score 4 5 was coded as a malignant lesion. Quantitative assessment of all image data was performed by one radiologist and a dedicated urologic pathologist. In this procedure, the radiologist was aware that all patients had been diagnosed with PCA and treated with RP but was unaware of any other clinical or outcome data. Tumor boundaries were determined by manually outlining the ROIs on axial T2-weighted images and DW images slice by slice (Figs. 1A and 1B). For patients with multitumor foci, only the lesion with the highest pathologic Gleason score that was also histologically larger than 0.5 cm 3 was considered to be the leading lesion for calculation (Fig. 1C). Each lesion identified by the radiologist was correlated to the histologic findings from post-rp specimens. Finally, an ROI within normal tissue (mean volume, 0.61 cm 3 ; range, cm 3 ) was outlined as a contrast to the tumor in the PZ or center gland in locations that the radiologist-pathologist indicated as normal. Normalized Dapp (D N ) and Kapp (K N ) of the cancer were calculated as follows: D N = K N = Dapp Ca Dapp normal Kapp Ca Kapp normal (2), (3). Histopathologic Examination After RP, the prostatic specimens were uniformly processed and submitted for histologic investigation as described elsewhere [12]. The prostatectomy specimens were fixed in 10% neutral buffered formalin and stored overnight after surgical resection. Prostatectomy specimens were fixed in 5% buffered formalin, processed, and cut serially into 3.5-mm-thick blocks from apex AJR:207, August

4 Zhang et al. to base in transverse planes. Each block was then halved or quartered (depending on its size), and 7- to 8-μm microtome slices were stained with H and E. Two dedicated urologic pathologists reviewed all the sections and outlined the location of the tumor on the photographs for each slice. Histopathologic variables included the tumor number, location (PZ or TZ), tumor size, primary and secondary pathologic Gleason score, and the presence or absence of ECE, SVI, perineural invasion, and lymph node metastasis. Statistical Analysis All statistical analyses were performed by using SPSS, (version 16.0). Continuous variables were summarized using means and ranges. Categoric variables were summarized with counts and proportions. Interreader agreement was evaluated with the Fleiss kappa coefficient. The kappa values were interpreted as poor agreement ( ), fair agreement ( ), moderate agreement ( ), good agreement ( ), or excellent agreement ( ). Statistical comparison of ranked variables (PI-RADS scores, ECE, and SVI) was made by a Mann-Whitney U test. Comparison of continuous variables (Dapp, Kapp, normalized Dapp, and normalized Kapp) was made by paired-samples t test. A Bland-Altman plot was created using the average of Dapp and Kapp measured from both b value sets as the x-axis and the difference between them as the y-axis. The mean percentage difference and 95% limits of agreement between the two measurements were reported. Spearman rank correlation coefficients were computed between DKI outputs (Dapp and Kapp) and pathologic findings (Gleason score and T category). Patients were classified into two groups according to Gleason score ( vs > 3 + 4) and T category (< pt3a vs pt3a) of the PCA. Statistical comparisons of the imaging ability to differentiate normal tissue from PCA, low-gleason-score from high-gleason-score PCA, and low- from high-stage PCA were made by area under the ROC curve (A z ) analysis. The diagnostic sensitivity, specificity, accuracy, positive predictive value, and negative predictive value were calculated at a cutoff point that maximized the value of the Youden index, and p < 0.05 determined statistical significance. Results Baseline Characteristics Of 108 patients, 227 solid foci were identified as positive for tumor at histopathologic analysis. Of these, 162 of 227 (71.4%) were in the PZ and 65 of 227 (28.6%) were in the TZ. One hundred four tumor foci (45.4%) were histologically smaller than 0.5 cm 3 or were MRI TABLE 1: Summary of Clinical and Pathologic Characteristics of Patients With Prostate Cancer (PCA) Variable invisible, and 17 lesions (7.4%) were larger than 0.5 cm 3 but were not identified as the leading lesions and thus were excluded from further statistical analysis. Baseline characteristics of the patients are summarized in Table 1. Value Clinical information (n = 108 patients) Age (y), median (range) 69.5 (48 86) Prostate-specific antigen level (ng/ml), median (range) 12.1 ( ) MRI-invisible PCA (n = 104 lesions) Tumor size (cm 3 ) < 0.5 or unable to be evaluated Pathologic Gleason score Pathologic T category pt2a MRI-visible PCA (n = 108 lesions) Tumor size (cm 3 ), median (range) 2.6 ( ) Located in peripheral zone 78 (72.2) Located in transition zone 30 (27.8) Pathologic Gleason score < (59.3) Pathologic Gleason score (40.7) < pt3 79 (73.1) pt3a 29 (26.9) pt3b 13 (12.0) Note Except where noted otherwise, data are number (%) of lesions. The histopathologic reports of tumor size, location, and pathologic Gleason score are only for the leading lesions of the included patients. TABLE 2: Summaries of Ranking Results Using Prostate Imaging Reporting and Data System per Reader for Scoring 1296 Regions in 108 Patients Imaging Combination Benign Uncertain Ranking Results Prostate Cancer Extracapsular Extension Seminal Vesicle Invasion Reader 1 Dataset A Dataset B Dataset C Dataset D Reader 2 Dataset A Dataset B Dataset C Dataset D Histopathologic indication Note Data are number of findings. Dataset A included T2-weighted MRI, dynamic contrast-enhanced (DCE) MRI, and diffusional kurtosis imaging (DKI) at a b value of 1400 s/mm 2 ; dataset B included T2-weighted MRI, DCE-MRI, and DKI at a b value of 2000 s/mm 2 ; dataset C included T2-weighted MRI, DCE-MRI, DKI, and apparent diffusion coefficient of gaussian distribution (Dapp) and apparent kurtosis coefficient (Kapp) maps at a b value of 1400 s/mm 2 ; and dataset D included T2-weighted MRI, DCE-MRI, DKI, and Dapp and Kapp maps at a b value of 2000 s/mm 2. Ranking Scores of Prostate The distribution of PI-RADS scores per reader for stratifying the 1296 prostatic regions in 108 patients are summarized in Table 2. Observations in dataset A were scored 4 AJR:207, August 2016

5 Diffusional Kurtosis Imaging of Prostate Cancer TABLE 3: Comparisons of Diffusional Kurtosis Imaging Coefficients Between Low and High b Values Variable higher, and those in dataset B were scored lower by the two readers using the PI-RADS scale scheme, compared with those in datasets C and D (chi-square test, p < 0.01). Ranking results for the observations did not reflect statistically significant differences between datasets C and D (p = 0.097). The presence of ECE was relatively overestimated by the two readers using any of the four datasets Normal Tissue ROIs Prostate Cancer Apparent diffusion coefficient of gaussian distribution (mm 2 /s) b = 1400 s/mm ± ± 0.31 < b = 2000 s/mm ± ± 0.30 < p < Apparent kurtosis coefficient b = 1400 s/mm ± ± 0.26 < b = 2000 s/mm ± ± 0.21 < p < Normalized apparent diffusion coefficient of gaussian distribution b = 1400 s/mm ± 0.15 b = 2000 s/mm ± 0.15 p < Normalized apparent kurtosis coefficient b = 1400 s/mm ± 0.61 b = 2000 s/mm ± 0.45 p < Note Except for p values, data are mean ± SD. (Dapp at b value of 1400 Dapp at b value of 2000) / Average Dapp (%) Average Dapp at Both b Values ( 10 3 mm 2 /s) (positive predictive value, %). The presence of SVI using any of the four datasets was in good accordance with the histopathologic indication (positive predictive value, %). The overall agreement between the two readers in determination of PI-RADS scores, ECE, and SVI is summarized in Table S1, which can be viewed in the AJR electronic 1.96 SD 3.4 Mean SD 11.7 (Kapp at b value of 1400 Kapp at b value of 2000) / Average Kapp (%) p supplement to this article (available at www. ajronline.org). Interreader agreement for the PI-RADS score was relatively low in dataset A (κ = 0.39) and dataset B (κ = 0.41). Interreader agreement for the PI-RADS score was somewhat improved by adding Dapp and Kapp maps (dataset C, κ = 0.66; dataset D, κ = 0.68). Interreader agreement for the determination of ECE and SVI was relatively high and was consistent among the four datasets (ECE, κ = ; SVI, κ = ). The Mann-Whitney U test showed that the ranking results for PI-RADS score, ECE, and SVI interpreted by the two radiologists were not significantly changed when the b value was reduced to 1400 s/mm 2 (Fig. 2 and Table S2, which can be seen in the AJR electronic supplement to this article, available at Quantitative Diffusional Kurtosis Imaging The quantitative DKI results of 108 leading lesions are summarized in Table 3. Reducing the b value highly influenced both the absolute and normalized Dapp and Kapp values for PCA (all p < 0.001) but did not significantly influence the absolute Dapp and Kapp values in normal tissue. Bland- Altman plots (Fig. 3) show that mean differences in percentage and limits of agreement in Dapp between a b value of 1400 s/mm 2 and a b value of 2000 s/mm 2 are 4.2% (95% CI, 11.7% to 3.4%; Pearson correlation coefficient, ρ = 0.987; p < 0.001). The mean differences in percentage and limits of agreement in Kapp between minimized and preferred b values are 9.7% (95% CI, 2.5% to 22.0%; ρ = 0.963; p < 0.001) SD 22.0 Mean SD Average Dapp at Both b Values (Arbitrary Units) A Fig. 3 Bland-Altman plots of average apparent diffusion coefficient of gaussian distribution (Dapp) and apparent kurtosis coefficient (Kapp) values. A and B, Graphs show average Dapp (A) and Kapp (B) values calculated using b values of 1400 and 2000 s/mm 2. Crosses denote data points. B AJR:207, August

6 Zhang et al. TABLE 4. Performance Ability of Diffusional Kurtosis Imaging Variables of Apparent Diffusion Coefficient of Gaussian Distribution (Dapp) and Apparent Kurtosis Coefficient (Kapp) Calculated With Two b Value Sets for Prostate Cancer (PCA) Diagnosis and Pathologic Grade Prediction Variable Lesion diagnosis (normal vs PCA) Dapp There was a statistically significant negative correlation between tumor Dapp and pathologic Gleason score, with Spearman ρ ranging between for a b value of 1400 s/mm 2 and for a b value of 2000 s/mm 2 (all p < 0.001). There was a statistically significant positive correlation between tumor Kapp and pathologic Gleason score, with Spearman ρ ranging between for a b value of 1400 s/mm 2 and for a b value of 2000 s/mm 2 (p < 0.001). The diagnostic ability of DKI parameters under two b value sets in the detection and classification of PCA are summarized in Table 4. The power of DKI parameters (Dapp and Kapp) to discriminate PCA from normal tissue, low-gleason-score from high-gleason-score PCA, and low-stage from highstage PCA was good to excellent for both b value sets, with individual A z ranging from 0.70 to 0.98 for the respective classifications. No statistically significant difference in ROC A z was determined between b values of 1400 Area Under ROC Curve, Mean (95% CI) Sensitivity Specificity Accuracy and 2000 s/mm 2 for the diagnosis of PCA, grading of Gleason score, and T category (all p > 0.05) (Fig. 4). Discussion Our volunteer experiment showed that the choice of b value significantly influenced the image quality and quantitation of DKI. Clinical data from 108 tumors showed that there was substantial interreader discordance between b values of 1400 and 2000 s/mm 2 in the classification of PI-RADS score when Dapp and Kapp maps were excluded from the image interpretation. Reducing the b value did not significantly influence the image interpretation in the determination of ECE and SVI. Reducing the b value generated larger DKI-associated Dapp, Kapp, normalized Dapp, and normalized Kapp values than those measured at a b value of 2000 s/mm 2. However, reducing the b value did not degrade the diagnostic effectiveness of DKI parameters to detect and classify PCA. Positive Predictive Value Negative Predictive Value Cutoff Value ( 10 3 mm 2 /s) b = 1400 s/mm ( ) 88.0 (95/108) 94.4 (102/108) 91.2 (197/216) 94.1 (95/101) 88.7 (102/115) 1.73 b = 2000 s/mm ( ) 89.8 (97/108) 94.4 (102/108) 92.1 (199/216) 94.2 (97/103) 90.3 (102/113) 1.68 Kapp b = 1400 s/mm ( ) 94.4 (102/108) 89.8 (97/108) 92.1 (199/216) 90.3 (102/113) 94.2 (97/103) > 0.74 b = 2000 s/mm ( ) 93.5 (101/108) 89.8 (97/108) 91.7 (198/216) 90.2 (101/112) 93.3 (97/104) > 0.7 Gleason Score ( vs > 3 + 4) Dapp b = 1400 s/mm ( ) 88.1 (37/42) 68.2 (45/66) 75.9 (82/108) 63.8 (37/58) 90.0 (45/50) 1.40 b = 2000 s/mm ( ) 85.7 (36/42) 68.2 (45/66) 75.0 (81/108) 63.2 (36/57) 88.2 (45/51) 1.33 Kapp b = 1400 s/mm ( ) 78.6 (33/42) 72.7 (48/66) 75.0 (81/108) 64.7 (33/51) 84.2 (48/57) > 1.08 b = 2000 s/mm ( ) 78.6 (33/42) 63.6 (42/66) 69.4 (75/108) 57.9 (33/57) 82.4 (42/51) > 0.95 T category (< T3a vs T3a) Dapp b = 1400 s/mm ( ) 93.1 (27/29) 53.2 (42/79) 63.9 (69/108) 42.2 (27/64) 95.4 (42/44) 1.42 b = 2000 s/mm ( ) 89.7 (26/29) 58.2 (46/79) 66.7 (72/108) 44.1 (26/59) 93.9 (46/49) 1.32 Kapp b = 1400 s/mm ( ) 86.2 (25/29) 49.4 (39/79) 59.3 (64/108) 38.5 (25/65) 90.7 (39/43) > 1.02 b = 2000 s/mm ( ) 89.7 (26/29) 43.0 (34/79) 55.6 (60/108) 36.6 (26/71) 91.9 (34/37) > 0.91 Note Except where noted otherwise data are percentage (number/total). Therefore, we concluded that the optimal DKI protocol using RESOLVE (b values of 0, 700, and 1400 s/mm 2 ) could be readily applied to a clinically indicated pelvic MRI examination and that it would not degrade the diagnostic effectiveness but could reduce the examination time (4 minutes 31 seconds vs 7 minutes 52 seconds) compared with the preferred b value set. The application of DKI in prostate imaging has been gaining great interest in recent years. It has been prospectively used in several studies by Rosenkrantz et al. [15, 16], Suo et al. [24], Quentin et al. [25], Tamura et al. [14], and Wang et al. [12] and has shown better individual features in distinguishing PCA from normal tissue and discriminating low-grade from high-grade PCA than does conventional DWI. In this study, we used DKI using a RESOLVE protocol in clinically indicated pelvic MRI. RESOLVE has been shown to decrease image distortion and susceptibility artifacts and improve spa- 6 AJR:207, August 2016

7 Diffusional Kurtosis Imaging of Prostate Cancer tial resolution in the adult brain and pediatric brain and in spine imaging [21, 26 28]. More recently, it has been used in breast [29 31], prostate [20], and abdominal imaging protocols [32] and showed better image quality and higher diagnostic accuracy than did conventional EPI sequences. However, long acquisition time and increased noise scales are two important disadvantages when RESOLVE is used for high-b-value DKI. The ratio of scanning time for RESOLVE to that of single-shot EPI is approximately equal to the number of readout segments in RESOLVE, and the low signal-to-noise ratio efficiency from RESOLVE is due to reduced readout time [20]. Higher b values can yield the best signal suppression for benign prostate tissue Sensitivity (%) Sensitivity (%) Specificity (%) Specificity (%) and can minimize the effects of T2 weighting, which is beneficial for robust diagnostic specificity [33]. Conversely, a low b value (e.g., < 1000 s/mm 2 ) can yield a better signalto-noise ratio and reduce time consumption but cannot suppress the signal intensity of benign tissue in the PZ and sometimes can obscure tumor lesions because of persistent T2 shine-through effects [34]. Our data showed substantial discordance in image interpretation when PI-RADS reporting schema were used. Image interpretation using datasets A, B, C, and D (with various b value and image combinations) may have different implications: for example, a lesion may present with low signal intensity on a T2-weighted MR image and Dapp b = 1400 s/mm 2 (A z, 0.969) Kapp b = 1400 s/mm 2 (A z, 0.972) Dapp b = 2000 s/mm 2 (A z, 0.974) Kapp b = 2000 s/mm 2 (A z, 0.971) Dapp b = 1400 s/mm 2 (A z, 0.751) Kapp b = 1400 s/mm 2 (A z, 0.711) Dapp b = 2000 s/mm 2 (A z, 0.751) Kapp b = 2000 s/mm 2 (A z, 0.704) A C Sensitivity (%) isointense signal intensity on a low-b-value DW image and may be assigned a PI-RADS score of 3, whereas the diagnostic specificity may be improved by increasing the b value or using Dapp or Kapp maps. From the results in Table 2, we can see that great differences in assigned PI-RADS scores are seen when different combinations of images are used. Unfortunately, in this study, radiologic findings of 1296 prostatic observations were not correlated to the histopathologic results region by region, which could be quite challenging to perform reproducibly. From these results we can conclude that the choice of b value and the use or disuse of Dapp and Kapp maps may have a great influence on the individual interpretation when a scaled Dapp b = 1400 s/mm 2 (A z, 0.817) Kapp b = 1400 s/mm 2 (A z, 0.783) Dapp b = 2000 s/mm 2 (A z, 0.820) Kapp b = 2000 s/mm 2 (A z, 0.768) Specificity (%) Fig. 4 ROC analysis of diffusional kurtosis imaging average apparent diffusion coefficient of gaussian distribution (Dapp) and apparent kurtosis coefficient (Kapp) values calculated with different b values. A C, Graphs show ROC analysis for differentiating prostate cancer (PCA) from normal tissue (A), discriminating low-gleason-score ( 3 + 4) from high-gleasonscore (> 3 + 4) PCA (B), and discriminating low-category (< T3a) from highcategory ( T3a) PCA (C). No statistically significant differences in area under ROC curve (A z ) were determined between b values of 1400 and 2000 s/mm 2 for diagnosis of PCA (p > 0.05), grading of Gleason score (p > 0.05), and grading of T stage (p > 0.05). Diagonal lines denote reference values. B AJR:207, August

8 Zhang et al. PI-RADS scheme is used. The use of a high b value and corresponding Dapp and Kapp maps may help to reduce the interreader discordance and improve diagnostic specificity [20, 33, 35, 36]. We compared the quantitative DKI outputs obtained from b values of 1400 and 2000 s/mm 2 (Table S3, which can be viewed in the AJR electronic supplement to this article, available at We found that a reduced b value produced larger absolute and normalized Dapp and Kapp values in malignant tissues, but did not significantly influence the measures in normal tissue. Our finding was partly consistent with the study of Fukunaga et al. [37]. Similar results were reported by the studies of Peng et al. [36] and Manenti et al. [33], in which a higher b value produced lower ADCs in PCA and normal tissue. However, the Bland-Altman plot and Pearson test showed that Dapp and Kapp values obtained at a b value of 1400 s/mm 2 were linearly correlated with those obtained at a b value of 2000 s/mm 2. The coefficients of variation were below 10%, on average. In addition, we compared the diagnostic ability of DKI parameters under two b value sets for the diagnosis and grading of PCA. All the parameters using the optimal cutoff values performed well in cancer detection and prediction of Gleason score and T category, and the diagnostic power (A z ) was not significantly affected by the change of b values. This was partly consistent with the study of Peng et al. [38], who found that, although the use of different b values can cause systematic changes in ADCs, the diagnostic effectiveness of the ADCs is not necessarily degraded. Our findings may be confined by the following limitations. First, we used an abbreviated PI-RADS scoring scheme to interpret the imaging findings of prostatic observations, which might reduce the true interreader discordance. Second, imaging findings of 1296 prostatic observations were not correlated to corresponding histopathologic results region by region, which could be quite challenging to perform reproducibly, and deviation was, to some extent, unavoidable. Third, only the MRI-visible PCAs were included in the investigation; subadvanced lesions were excluded and not analyzed. Because of this selection bias, the sample may not fully represent the features of the population of patients with PCA, and the results are not applicable to other treatment options. Conclusion With this study, we optimized a DKI protocol using RESOLVE to analyze clinically indicated pelvic MRI examinations. Our volunteer experience and clinical study showed that the choice of b value significantly influenced the image quality, image quantitation of DKI, and image interpretation using PI-RADS. Although the optimal b value set (0, 700, and 1400 s/mm 2 ) can produce systematic changes in Dapp and Kapp values compared with the standard reference, their diagnostic effectiveness in discriminating PCA and predicting the pathologic grade of PCA was not necessarily degraded. Therefore, we conclude that optimal RESOLVE DKI is readily applied to a clinically indicated pelvic MRI examination for the diagnosis of PCA, which saves examination time. Reference 1. Siegel R, Desantis C, Jemal A. Colorectal cancer statistics, CA Cancer J Clin 2014; 64: Carlsson S, Vickers AJ, Roobol M, et al. Prostate cancer screening: facts, statistics, and interpretation in response to the US Preventive Services Task Force Review. J Clin Oncol 2012; 30: Barentsz JO, Richenberg J, Clements R, et al. ESUR prostate MR guidelines Eur Radiol 2012; 22: Mottet N, Bellmunt J, Bolla M, et al. EAU guidelines on prostate cancer. Part II. Treatment of advanced, relapsing, and castration-resistant prostate cancer. Eur Urol 2011; 59: Heidenreich A, Bellmunt J, Bolla M, et al. EAU guidelines on prostate cancer. Part I. Screening, diagnosis, and treatment of clinically localised disease. Eur Urol 2011; 59: Yakar D, Debats OA, Bomers JG, et al. Predictive value of MRI in the localization, staging, volume estimation, assessment of aggressiveness, and guidance of radiotherapy and biopsies in prostate cancer. J Magn Reson Imaging 2012; 35: Jung SI, Donati OF, Vargas HA, Goldman D, Hricak H, Akin O. Transition zone prostate cancer: incremental value of diffusion-weighted endorectal MR imaging in tumor detection and assessment of aggressiveness. Radiology 2013; 269: Chong Y, Kim CK, Park SY, Park BK, Kwon GY, Park JJ. Value of diffusion-weighted imaging at 3 T for prediction of extracapsular extension in patients with prostate cancer: a preliminary study. AJR 2014; 202: Jensen JH, Helpern JA, Ramani A, Lu H, Kaczynski K. Diffusional kurtosis imaging: the quantification of non-gaussian water diffusion by means of magnetic resonance imaging. Magn Reson Med 2005; 53: Nakanishi A, Fukunaga I, Hori M, et al. Microstructural changes of the corticospinal tract in idiopathic normal pressure hydrocephalus: a comparison of diffusion tensor and diffusional kurtosis imaging. Neuroradiology 2013; 55: Paydar A. Diffusional kurtosis imaging: a promising technique for detecting microstructural changes in neural development and regeneration. Neural Regen Res 2014; 9: 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; 33:337. e e Roethke MC, Kuder TA, Kuru TH, et al. Evaluation of diffusion kurtosis imaging versus standard diffusion imaging for detection and grading of peripheral zone prostate cancer. Invest Radiol 2015; 50: Tamura C, Shinmoto H, Soga S, et al. Diffusion kurtosis imaging study of prostate cancer: preliminary findings. J Magn Reson Imaging 2014; 40: Rosenkrantz AB, Prabhu V, Sigmund EE, Babb JS, Deng FM, Taneja SS. Utility of diffusional kurtosis imaging as a marker of adverse pathologic outcomes among prostate cancer active surveillance candidates undergoing radical prostatectomy. AJR 2013; 201: Rosenkrantz AB, Sigmund EE, Johnson G, et al. Prostate cancer: feasibility and preliminary experience of a diffusional kurtosis model for detection and assessment of aggressiveness of peripheral zone cancer. Radiology 2012; 264: Tachibana Y, Obata T, Tsuchiya H, et al. Diffusion-tensor-based method for robust and practical estimation of axial and radial diffusional kurtosis. Eur Radiol 2015 Oct 7 [Epub ahead of print] 18. Yokosawa S, Sasaki M, Bito Y, et al. Optimization of scan parameters to reduce acquisition time for diffusion kurtosis imaging at 1.5T. Magn Reson Med Sci 2016; 15: Thian YL, Xie W, Porter DA, Weileng Ang B. Readout-segmented echo-planar imaging for diffusion-weighted imaging in the pelvis at 3T: a feasibility study. Acad Radiol 2014; 21: Medved M, Soylu-Boy FN, Karademir I, et al. High-resolution diffusion-weighted imaging of the prostate. AJR 2014; 203: Yeom KW, Holdsworth SJ, Van AT, et al. Comparison of readout-segmented echo-planar imaging (EPI) and single-shot EPI in clinical application of diffusion-weighted imaging of the pediatric brain. AJR 2013; 200:[web]W437 W Bogner W, Pinker-Domenig K, Bickel H, et al. Readout-segmented echo-planar imaging improves the diagnostic performance of diffusion- 8 AJR:207, August 2016

9 Diffusional Kurtosis Imaging of Prostate Cancer weighted MR breast examinations at 3.0 T. utility of reduced-distortion readout-segmented 33. Manenti G, Nezzo M, Chegai F, Vasili E, Radiology 2012; 263:64 76 echo-planar imaging in the head and neck region: Bonanno E, Simonetti G. DWI of prostate cancer: 23. Weinreb JC, Barentsz JO, Choyke PL, et al. PI-RADS prostate imaging: reporting and data system 2015, version 2. Eur Urol 2016; 69: Suo S, Chen X, Wu L, et al. Non-Gaussian water diffusion kurtosis imaging of prostate cancer. Magn Reson Imaging 2014; 32: Quentin M, Pentang G, Schimmoller L, et al. Feasibility of diffusional kurtosis tensor imaging in prostate MRI for the assessment of prostate cancer: preliminary results. Magn Reson Imaging 2014; 32: Robson MD, Anderson AW, Gore JC. Diffusionweighted multiple shot echo planar imaging of humans without navigation. Magn Reson Med 1997; 38: Porter DA, Heidemann RM. High resolution diffusion-weighted imaging using readout-segmented echo-planar imaging, parallel imaging and a two-dimensional navigator-based reacquisition. Magn Reson Med 2009; 62: Koyasu S, Iima M, Umeoka S, et al. The clinical initial experience. Eur Radiol 2014; 24: Park JY, Shin HJ, Shin KC, et al. Comparison of readout segmented echo planar imaging (EPI) and EPI with reduced field-of-view diffusion-weighted imaging at 3T in patients with breast cancer. J Magn Reson Imaging 2015; 42: Sun K, Chai W, Fu C, et al. Diffusion-weighted imaging-guided MR spectroscopy in breast lesions using readout-segmented echo-planar imaging. Eur Radiol 2015 Sep 19 [Epub ahead of print] 31. Bogner W, Pinker K, Zaric O, et al. Bilateral diffusion-weighted MR imaging of breast tumors with submillimeter resolution using readout-segmented echo-planar imaging at 7 T. Radiology 2015; 274: Tokoro H, Fujinaga Y, Ohya A, et al. Usefulness of free-breathing readout-segmented echo-planar imaging (RESOLVE) for detection of malignant liver tumors: comparison with single-shot echoplanar imaging (SS-EPI). Eur J Radiol 2014; 83: optimal b-value in clinical practice. Prostate Cancer 2014; 2014: Rosenkrantz AB, Kong X, Niver BE, et al. Prostate cancer: comparison of tumor visibility on trace diffusion-weighted images and the apparent diffusion coefficient map. AJR 2011; 196: Thörmer G, Otto J, Reiss-Zimmermann M, et al. Diagnostic value of ADC in patients with prostate cancer: influence of the choice of b values. Eur Radiol 2012; 22: Peng Y, Jiang Y, Antic T, et al. Apparent diffusion coefficient for prostate cancer imaging: impact of b values. AJR 2014; 202:[web]W247 W Fukunaga I, Hori M, Masutani Y, et al. Effects of diffusional kurtosis imaging parameters on diffusion quantification. Radiol Phys Technol 2013; 6: Peng Y, Jiang Y, Antic T, Giger ML, Eggener SE, Oto A. Validation of quantitative analysis of multiparametric prostate MR images for prostate cancer detection and aggressiveness assessment: a cross-imager study. Radiology 2014; 271: AJR:207, August

10 本文献由 学霸图书馆 - 文献云下载 收集自网络, 仅供学习交流使用 学霸图书馆 ( 是一个 整合众多图书馆数据库资源, 提供一站式文献检索和下载服务 的 24 小时在线不限 IP 图书馆 图书馆致力于便利 促进学习与科研, 提供最强文献下载服务 图书馆导航 : 图书馆首页文献云下载图书馆入口外文数据库大全疑难文献辅助工具