Original Investigations
Histogram Analysis of Apparent Diffusion Coefficient at 3.0 T in Urinary Bladder Lesions: Correlation with Pathologic Findings Shi-Teng Suo, ME, Xiao-Xi Chen, MD, Yu Fan, BS, Lian-Ming Wu, MD, Qiu-Ying Yao, BS, Meng-Qiu Cao, MD, Qiang Liu, MD, Jian-Rong Xu, MD, PhD Rationale and Objectives: To investigate the potential value of histogram analysis of apparent diffusion coefficient (ADC) obtained at standard (700 s/mm2) and high (1500 s/mm2) b values on a 3.0-T scanner in the differentiation of bladder cancer from benign lesions and in assessing bladder tumors of different pathologic T stages and to evaluate the diagnostic performance of ADC-based histogram parameters. Materials and Methods: In all, 52 patients with bladder lesions, including benign lesions (n = 7) and malignant tumors (n = 45; T1 stage or less, 23; T2 stage, 7; T3 stage, 8; and T4 stage, 7), were retrospectively evaluated. Magnetic resonance examination at 3.0 T and diffusionweighted imaging were performed. ADC maps were obtained at two b values (b = 700 and 1500 s/mm2; ie, ADC-700 and ADC-1500). Parameters of histogram analysis included mean, kurtosis, skewness, and entropy. The correlations between these parameters and pathologic results were revealed. Receiver operating characteristic (ROC) curves were generated to determine the diagnostic value of histogram parameters. Results: Significant differences were found in mean ADC-700, mean ADC-1500, skewness ADC-1500, and kurtosis ADC-1500 between bladder cancer and benign lesions (P = .002–.032). There were also significant differences in mean ADC-700, mean ADC-1500, and kurtosis ADC-1500 among bladder tumors of different pathologic T stages (P = .000–.046). No significant differences were observed in other parameters. Mean ADC-1500 and kurtosis ADC-1500 were significantly correlated with T stage, respectively (r = 0.614, P < .001; r = 0.374, P = .011). ROC analysis showed that the combination of mean ADC-1500 and kurtosis ADC-1500 has the maximal area under the ROC curve (AUC, 0.894; P < .001) in the differentiation of benign lesions and malignant tumors, with a sensitivity of 77.78% and specificity of 100%. AUCs for differentiating low- and high-stage tumors were 0.840 for mean ADC-1500 (P < .001) and 0.696 for kurtosis ADC1500 (P = .015). Conclusions: Histogram analysis of ADC-1500 at 3.0 T can be useful in evaluation of bladder lesions. A combination of mean ADC-1500 and kurtosis ADC-1500 may be more beneficial in the differentiation of benign and malignant lesions. Mean ADC-1500 was the most promising parameter for differentiating low- from high-stage bladder cancer. Key Words: Histogram analysis; apparent diffusion coefficient; magnetic resonance imaging; bladder; pathology. ªAUR, 2014
B
ladder cancer is the most common type of malignant tumor in urinary tract, which is hazardous heavily to human health among both men and women (1). Preoperative assessment of the bladder cancer pathologic T stage,
Acad Radiol 2014; -:1–8 From the Department of Radiology, Ren Ji Hospital, School of Medicine, Shanghai Jiao Tong University, No. 160, Pujian Rd, Shanghai 200127, China (S.-T.S., X.-X.C., Y.F., L.-M.W., Q.-Y.Y., M.-Q.C., J.-R.X.) and Department of Pathology, Ren Ji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai, China (Q.L.). Received January 2, 2014; accepted March 4, 2014. S.-T.S. and X.-X.C. contributed equally to this work. The present research was supported by National Natural Science Foundation of China (No. 81371622) and by Medical Engineering Cross Research Foundation of Shanghai Jiao Tong University (No. YG2013MS37). Address correspondence to: J.-R.X. e-mail: [email protected] ªAUR, 2014 http://dx.doi.org/10.1016/j.acra.2014.03.004
which is a measure of clinical aggressiveness, is the most primary factor in choosing the most appropriate treatment method. As a result, disease prognosis can be different among patients depending on different tumor stages. Low-stage superficial tumors (T1 stage or lower) are associated with low risk of progression and can be effectively treated by local endoscopic resection with a favorable survival rate. On the other hand, high-stage invasive tumors (T2 stage or higher) often develop metastatic disease and are treated either by curative cystectomy or by radiation therapy or chemotherapy (2). Diffusion-weighted imaging (DWI) as a functional magnetic resonance imaging (MRI) technique has shown its ability to diagnose bladder cancer and distinguish tumors of different stages (3–6). DWI is based on the microscopic diffusion movements in the protons of the tissues’ water molecules and can reveal information about microstructure complexity, such 1
SUO ET AL
as, cellularity, integrity of the cellular membranes, and aggregation of macromolecules (7). Apparent diffusion coefficient (ADC), which is a quantitative parameter derived from DWI, can reflect water mobility within various tissues according to the pathophysiologic state. In recent years, DWI has been used as part of a routine MRI protocol for bladder examination. Although DWI has proven to be a useful imaging tool to distinguish among benign and malignant lesions and evaluate tumor stage (8), especially to reduce overstaging rate compared to gadolinium-enhanced imaging (5), the accuracy in differentiating superficial from invasive tumors is relatively low (4,9). In addition, ADC values showed a substantial overlap in lowand high-stage tumors, thus limiting the usefulness of DWI for the individual patient (10). As an additional technique to analyze ADC maps, histogram analysis based on pixel distribution yields more diffusion metrics and provides more statistical information over the mean ADC values, such as the heterogeneity of water diffusivity within tissue. Histogram analysis has shown its potential to contribute to clinical diagnosis in a complementary or more effective way (11–13). To the best of our knowledge, quantitative analysis of bladder tumors has mainly be assessed by determined the ADC values within one region of interest (ROI) at 1.5 T, but studies have rarely reported the usefulness of histogram analysis of ADC in assessment of bladder benign lesions and malignant tumors. In addition, few quantitative studies on bladder DWI have been conducted using high–field strength (3.0-T) MRI that enables increased signal-to-noise ratio and increased spatial resolution. Moreover, various studies show that different b values result in different ADC values (8,14–16). No evidence has indicated the ideal b value for ADC histogram analysis. Therefore, the purpose of the study was to retrospectively investigate the potential value of histogram analysis of ADC obtained at standard (700 s/mm2) and high (1500 s/mm2) b values on a 3.0-T scanner in the differentiation of bladder cancer from benign lesions and in distinguishing bladder tumors of different pathologic T stages and to evaluate the diagnostic performance of ADC-based histogram parameters. MATERIALS AND METHODS Study Population
This retrospective study was approved by our institutional review board, and written informed consent was obtained from all patients. Between October 2010 and December 2011, 97 patients suspected of bladder cancer underwent MRI examination including DWI. From these patients, 45 were excluded for the following reasons: 1) no pathologic records of diagnosis were established within 30 days after MRI examination (n = 22), 2) DWI was not obtained at b values of 700 and 1500 s/mm2 (n = 18), 3) no lesion was detected on magnetic resonance (MR) images (n = 3); and 4) low image quality caused by artifacts was difficult for observers to interpret (n = 2). 2
Academic Radiology, Vol -, No -, - 2014
Consequently, a final cohort of 52 patients (age range, 44–89 years; mean, 66 years) was included in this study. There were 43 men (82.7%) and 9 women (17.3%). Pathologic diagnosis for bladder benign lesions was obtained via cystoscopy biopsy and for bladder cancer via transurethral resection (n = 31) or via radical cystectomy (n = 14). Pathologic T stage was classified according to the 2002 tumor-node-metastasis (TNM) system (17). Pathologic outcomes were obtained within 22 days (mean, 8 days) after MRI. MRI Examination
Before MRI examination, patients were instructed not to urinate for at least 1 hour to achieve adequate distension of the urinary bladder. All patients underwent routine pelvic MRI and DWI on a 3.0-T MR scanner (HDxt; GE Medical Systems, Waukesha, WI) using a phased-array eight-channel cardiac coil. Routine pelvic MR images were acquired as follows: axial T1weighted spin-echo images (repetition time [TR]/echo time [TE], 420 ms/7 ms; 352 192 matrix; 28-cm field of view; 5-mm slice thickness; 1-mm intersection gap; number of excitation, 2), axial and sagittal turbo spin-echo T2-weighted images (TR/TE, 3920 ms/131 ms; 320 386 matrix; 28-cm field of view; 5-mm slice thickness; 1-mm intersection gap; number of excitation, 2), axial and sagittal fat-suppressed T2-weighted images (TR/TE, 3920 ms/131 ms; 320 386 matrix; 28-cm field of view; 5-mm slice thickness; 1-mm intersection gap; number of excitation, 2). Parameters for T1-weighted imaging and T2-weighted imaging were slightly floating because of individual difference. Subsequently, axial DWI images were obtained using a single-shot spin-echo echo-planar sequence (TR/TE, 4000 ms/66 ms for b value = 700 s/mm2; TR/TE, 5200 ms/75 ms for b value = 1500 s/mm2; 96 130 matrix; 28-cm field of view; 3-mm slice thickness; 1-mm intersection gap; number of excitation, 6). Image Analysis
All Digital Imaging and Communications in Medicine (DICOM) data were transferred from the picture archiving and communication system workstation to a personal computer for further analysis. Postprocessing codes were written in Matlab version R2011b (MathWorks, Natick, MA). ADC was calculated on a pixel-by-pixel basis using the following equation: ADC = ln (Sb/S0)/b, where b is the diffusion-sensitizing factor (b value) and Sb and S0 the signal intensity at a nonzero b value and zero b value, respectively. All ROIs were determined in consensus by two radiologists (X.-X.C. and L.-M.W., with 2 and 5 years of experience in pelvic MRI, respectively) who were blind to the intraoperative clinical findings and the postoperative pathologic outcomes. For each case, the observers were presented with T2weighted images, DWI images, and ADC maps (ADC-700 and ADC-1500). All measurements were performed on ADC maps and the other images were used to help locate the lesion. A freehand ROI was placed over the entire lesion
Academic Radiology, Vol -, No -, - 2014
HISTOGRAM ANALYSIS OF ADC IN BLADDER LESIONS
Figure 1. A representative case of histogram analysis from a 71-year-old male patient with urinary bladder carcinoma pathologically diagnosed as stage T3. (a) Axial T2-weighted image shows a soft tissue mass invading the perivesical fat on the right lateral bladder wall (arrow). (b,c) Apparent diffusion coefficient (ADC) maps obtained at standard (700 s/mm2) and high (1500 s/mm2) b values, respectively, show the mass with a hypointense signal corresponding to a restriction in diffusion. (d,e) Histograms of the mass on ADC maps of (b) and (c), respectively.
on the ADC map slice with the maximum diameter of lesion. In patients with multiple lesions, only the largest lesion was included in the analysis. ROIs were placed carefully so as to avoid normal bladder wall, urine, or the areas of tumor necrosis. Therefore, a total number of 52 ROIs were obtained with areas in the range of 16–220 mm2. Once the ROI was determined, histogram analysis was done. Parameters derived from the histogram were listed as follows: mean ADC, kurtosis (a measure of peakedness of the distribution, equals three if the histogram is Gaussian, larger than three if the histogram has a sharper peak, and smaller than three if the histogram is less peaked), skewness (a measure of asymmetry of the distribution, is positive if more points lie to the left of the mean and negative if the opposite), and entropy (a measure of variation in a histogram, equals zero if all data are the same and increases as the data distribution becomes more irregular). Statistical Analysis
Statistical analyses were performed with MedCalc version 12.7.0.0 for Windows (MedCalc Software, Mariakerke, Belgium). Data are presented as medians with interquartile ranges. P < .05 was considered indicative of a statistically significant difference. The nonparametric Mann–Whitney U test was used to compare histogram parameters between bladder benign and
malignant lesions. To assess the differences in histogram parameters among bladder cancer of different T stages, the Kruskal– Wallis test was used, followed by the Mann–Whitney U test with Bonferroni correction for pairwise comparisons. Spearman correlation coefficients (r) were calculated to examine the correlations between histogram parameters and T stage. Receiver operating characteristic (ROC) curves for histogram parameters that were significantly different and the combination of significant measures were generated to help differentiate benign and malignant lesions and differentiate low- and high-stage tumors. For ROC analyses, the combination of significant measures were first calculated with logistic regression models and then used for the construction of the ROC curves. Diagnostic performance was determined by calculating the area under the ROC curve (AUC). Sensitivity and specificity for the classification were calculated. Cutoff values were established by calculating the maximal Youden index (Youden index = sensitivity + specificity 1). RESULTS Histopathologic Findings
Of the 52 patients, 7 were diagnosed with benign lesions (glandular cystitis, 2; eosinophilic cystitis, 3; and granulomatous inflammation, 2). In the 45 patients with bladder cancer, the 3
SUO ET AL
4 TABLE 1. Histogram Parameters of ADC-700 and ADC-1500 for Bladder Benign and Malignant Lesions b = 700 s/mm2
Benign Malignant P value
b = 1500 s/mm2
Mean
Skewness
Kurtosis
Entropy
Mean
Skewness
Kurtosis
Entropy
1.74 (1.50–2.02) 1.30 (1.03–1.51) .002
0.38 (0.11–0.61) 0.62 (0.21–1.06) .198
3.14 (2.17–4.11) 3.02 (2.71–4.50) .520
4.08 (3.73–4.52) 4.06 (3.62–4.35) .592
1.38 (1.30–1.55) 1.07 (0.88–1.25) .002
0.30 ( 0.17–0.65) 0.76 (0.36–1.28) .032
2.65 (2.20–3.64) 4.23 (3.13–5.63) .004
3.97 (3.77–4.32) 3.78 (3.39–4.04) .217
ADC, apparent diffusion coefficient. Data are expressed as medians with interquartile ranges. ADC mean values are expressed as 10 ADC skewness, kurtosis, and entropy are dimensionless.
3
mm2/s.
TABLE 2. Histogram Parameters of ADC-700 and ADC-1500 for Bladder Cancer of Different T Stages
#T1 T2 T3 T4 P value
b = 1500 s/mm2
Mean
Skewness
Kurtosis
Entropy
Mean
Skewness
Kurtosis
Entropy
1.38 (1.16–1.62) 1.39 (1.07–1.49) 0.97 (0.85–1.11) 0.95 (0.88–1.30) .001
0.64 ( 0.09–1.21) 0.59 (0.52–0.63) 0.66 (0.31–1.82) 0.69 (0.16–1.05) .892
3.06 (2.64–4.51) 2.82 (2.65–4.13) 3.44 (2.88–7.74) 2.89 (2.72–4.07) .548
4.06 (3.62–4.22) 4.33 (3.88–4.42) 3.74 (3.42–4.27) 4.13 (3.36–4.50) .277
1.23 (0.99–1.32) 1.15 (0.87–1.21) 0.81 (0.73–0.93) 0.78 (0.76–1.13)
Histogram Analysis of Apparent Diffusion Coefficient at 3.0 T in Urinary Bladder Lesions: Correlation with Pathologic Findings Shi-Teng Suo, ME, Xiao-Xi Chen, MD, Yu Fan, BS, Lian-Ming Wu, MD, Qiu-Ying Yao, BS, Meng-Qiu Cao, MD, Qiang Liu, MD, Jian-Rong Xu, MD, PhD Rationale and Objectives: To investigate the potential value of histogram analysis of apparent diffusion coefficient (ADC) obtained at standard (700 s/mm2) and high (1500 s/mm2) b values on a 3.0-T scanner in the differentiation of bladder cancer from benign lesions and in assessing bladder tumors of different pathologic T stages and to evaluate the diagnostic performance of ADC-based histogram parameters. Materials and Methods: In all, 52 patients with bladder lesions, including benign lesions (n = 7) and malignant tumors (n = 45; T1 stage or less, 23; T2 stage, 7; T3 stage, 8; and T4 stage, 7), were retrospectively evaluated. Magnetic resonance examination at 3.0 T and diffusionweighted imaging were performed. ADC maps were obtained at two b values (b = 700 and 1500 s/mm2; ie, ADC-700 and ADC-1500). Parameters of histogram analysis included mean, kurtosis, skewness, and entropy. The correlations between these parameters and pathologic results were revealed. Receiver operating characteristic (ROC) curves were generated to determine the diagnostic value of histogram parameters. Results: Significant differences were found in mean ADC-700, mean ADC-1500, skewness ADC-1500, and kurtosis ADC-1500 between bladder cancer and benign lesions (P = .002–.032). There were also significant differences in mean ADC-700, mean ADC-1500, and kurtosis ADC-1500 among bladder tumors of different pathologic T stages (P = .000–.046). No significant differences were observed in other parameters. Mean ADC-1500 and kurtosis ADC-1500 were significantly correlated with T stage, respectively (r = 0.614, P < .001; r = 0.374, P = .011). ROC analysis showed that the combination of mean ADC-1500 and kurtosis ADC-1500 has the maximal area under the ROC curve (AUC, 0.894; P < .001) in the differentiation of benign lesions and malignant tumors, with a sensitivity of 77.78% and specificity of 100%. AUCs for differentiating low- and high-stage tumors were 0.840 for mean ADC-1500 (P < .001) and 0.696 for kurtosis ADC1500 (P = .015). Conclusions: Histogram analysis of ADC-1500 at 3.0 T can be useful in evaluation of bladder lesions. A combination of mean ADC-1500 and kurtosis ADC-1500 may be more beneficial in the differentiation of benign and malignant lesions. Mean ADC-1500 was the most promising parameter for differentiating low- from high-stage bladder cancer. Key Words: Histogram analysis; apparent diffusion coefficient; magnetic resonance imaging; bladder; pathology. ªAUR, 2014
B
ladder cancer is the most common type of malignant tumor in urinary tract, which is hazardous heavily to human health among both men and women (1). Preoperative assessment of the bladder cancer pathologic T stage,
Acad Radiol 2014; -:1–8 From the Department of Radiology, Ren Ji Hospital, School of Medicine, Shanghai Jiao Tong University, No. 160, Pujian Rd, Shanghai 200127, China (S.-T.S., X.-X.C., Y.F., L.-M.W., Q.-Y.Y., M.-Q.C., J.-R.X.) and Department of Pathology, Ren Ji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai, China (Q.L.). Received January 2, 2014; accepted March 4, 2014. S.-T.S. and X.-X.C. contributed equally to this work. The present research was supported by National Natural Science Foundation of China (No. 81371622) and by Medical Engineering Cross Research Foundation of Shanghai Jiao Tong University (No. YG2013MS37). Address correspondence to: J.-R.X. e-mail: [email protected] ªAUR, 2014 http://dx.doi.org/10.1016/j.acra.2014.03.004
which is a measure of clinical aggressiveness, is the most primary factor in choosing the most appropriate treatment method. As a result, disease prognosis can be different among patients depending on different tumor stages. Low-stage superficial tumors (T1 stage or lower) are associated with low risk of progression and can be effectively treated by local endoscopic resection with a favorable survival rate. On the other hand, high-stage invasive tumors (T2 stage or higher) often develop metastatic disease and are treated either by curative cystectomy or by radiation therapy or chemotherapy (2). Diffusion-weighted imaging (DWI) as a functional magnetic resonance imaging (MRI) technique has shown its ability to diagnose bladder cancer and distinguish tumors of different stages (3–6). DWI is based on the microscopic diffusion movements in the protons of the tissues’ water molecules and can reveal information about microstructure complexity, such 1
SUO ET AL
as, cellularity, integrity of the cellular membranes, and aggregation of macromolecules (7). Apparent diffusion coefficient (ADC), which is a quantitative parameter derived from DWI, can reflect water mobility within various tissues according to the pathophysiologic state. In recent years, DWI has been used as part of a routine MRI protocol for bladder examination. Although DWI has proven to be a useful imaging tool to distinguish among benign and malignant lesions and evaluate tumor stage (8), especially to reduce overstaging rate compared to gadolinium-enhanced imaging (5), the accuracy in differentiating superficial from invasive tumors is relatively low (4,9). In addition, ADC values showed a substantial overlap in lowand high-stage tumors, thus limiting the usefulness of DWI for the individual patient (10). As an additional technique to analyze ADC maps, histogram analysis based on pixel distribution yields more diffusion metrics and provides more statistical information over the mean ADC values, such as the heterogeneity of water diffusivity within tissue. Histogram analysis has shown its potential to contribute to clinical diagnosis in a complementary or more effective way (11–13). To the best of our knowledge, quantitative analysis of bladder tumors has mainly be assessed by determined the ADC values within one region of interest (ROI) at 1.5 T, but studies have rarely reported the usefulness of histogram analysis of ADC in assessment of bladder benign lesions and malignant tumors. In addition, few quantitative studies on bladder DWI have been conducted using high–field strength (3.0-T) MRI that enables increased signal-to-noise ratio and increased spatial resolution. Moreover, various studies show that different b values result in different ADC values (8,14–16). No evidence has indicated the ideal b value for ADC histogram analysis. Therefore, the purpose of the study was to retrospectively investigate the potential value of histogram analysis of ADC obtained at standard (700 s/mm2) and high (1500 s/mm2) b values on a 3.0-T scanner in the differentiation of bladder cancer from benign lesions and in distinguishing bladder tumors of different pathologic T stages and to evaluate the diagnostic performance of ADC-based histogram parameters. MATERIALS AND METHODS Study Population
This retrospective study was approved by our institutional review board, and written informed consent was obtained from all patients. Between October 2010 and December 2011, 97 patients suspected of bladder cancer underwent MRI examination including DWI. From these patients, 45 were excluded for the following reasons: 1) no pathologic records of diagnosis were established within 30 days after MRI examination (n = 22), 2) DWI was not obtained at b values of 700 and 1500 s/mm2 (n = 18), 3) no lesion was detected on magnetic resonance (MR) images (n = 3); and 4) low image quality caused by artifacts was difficult for observers to interpret (n = 2). 2
Academic Radiology, Vol -, No -, - 2014
Consequently, a final cohort of 52 patients (age range, 44–89 years; mean, 66 years) was included in this study. There were 43 men (82.7%) and 9 women (17.3%). Pathologic diagnosis for bladder benign lesions was obtained via cystoscopy biopsy and for bladder cancer via transurethral resection (n = 31) or via radical cystectomy (n = 14). Pathologic T stage was classified according to the 2002 tumor-node-metastasis (TNM) system (17). Pathologic outcomes were obtained within 22 days (mean, 8 days) after MRI. MRI Examination
Before MRI examination, patients were instructed not to urinate for at least 1 hour to achieve adequate distension of the urinary bladder. All patients underwent routine pelvic MRI and DWI on a 3.0-T MR scanner (HDxt; GE Medical Systems, Waukesha, WI) using a phased-array eight-channel cardiac coil. Routine pelvic MR images were acquired as follows: axial T1weighted spin-echo images (repetition time [TR]/echo time [TE], 420 ms/7 ms; 352 192 matrix; 28-cm field of view; 5-mm slice thickness; 1-mm intersection gap; number of excitation, 2), axial and sagittal turbo spin-echo T2-weighted images (TR/TE, 3920 ms/131 ms; 320 386 matrix; 28-cm field of view; 5-mm slice thickness; 1-mm intersection gap; number of excitation, 2), axial and sagittal fat-suppressed T2-weighted images (TR/TE, 3920 ms/131 ms; 320 386 matrix; 28-cm field of view; 5-mm slice thickness; 1-mm intersection gap; number of excitation, 2). Parameters for T1-weighted imaging and T2-weighted imaging were slightly floating because of individual difference. Subsequently, axial DWI images were obtained using a single-shot spin-echo echo-planar sequence (TR/TE, 4000 ms/66 ms for b value = 700 s/mm2; TR/TE, 5200 ms/75 ms for b value = 1500 s/mm2; 96 130 matrix; 28-cm field of view; 3-mm slice thickness; 1-mm intersection gap; number of excitation, 6). Image Analysis
All Digital Imaging and Communications in Medicine (DICOM) data were transferred from the picture archiving and communication system workstation to a personal computer for further analysis. Postprocessing codes were written in Matlab version R2011b (MathWorks, Natick, MA). ADC was calculated on a pixel-by-pixel basis using the following equation: ADC = ln (Sb/S0)/b, where b is the diffusion-sensitizing factor (b value) and Sb and S0 the signal intensity at a nonzero b value and zero b value, respectively. All ROIs were determined in consensus by two radiologists (X.-X.C. and L.-M.W., with 2 and 5 years of experience in pelvic MRI, respectively) who were blind to the intraoperative clinical findings and the postoperative pathologic outcomes. For each case, the observers were presented with T2weighted images, DWI images, and ADC maps (ADC-700 and ADC-1500). All measurements were performed on ADC maps and the other images were used to help locate the lesion. A freehand ROI was placed over the entire lesion
Academic Radiology, Vol -, No -, - 2014
HISTOGRAM ANALYSIS OF ADC IN BLADDER LESIONS
Figure 1. A representative case of histogram analysis from a 71-year-old male patient with urinary bladder carcinoma pathologically diagnosed as stage T3. (a) Axial T2-weighted image shows a soft tissue mass invading the perivesical fat on the right lateral bladder wall (arrow). (b,c) Apparent diffusion coefficient (ADC) maps obtained at standard (700 s/mm2) and high (1500 s/mm2) b values, respectively, show the mass with a hypointense signal corresponding to a restriction in diffusion. (d,e) Histograms of the mass on ADC maps of (b) and (c), respectively.
on the ADC map slice with the maximum diameter of lesion. In patients with multiple lesions, only the largest lesion was included in the analysis. ROIs were placed carefully so as to avoid normal bladder wall, urine, or the areas of tumor necrosis. Therefore, a total number of 52 ROIs were obtained with areas in the range of 16–220 mm2. Once the ROI was determined, histogram analysis was done. Parameters derived from the histogram were listed as follows: mean ADC, kurtosis (a measure of peakedness of the distribution, equals three if the histogram is Gaussian, larger than three if the histogram has a sharper peak, and smaller than three if the histogram is less peaked), skewness (a measure of asymmetry of the distribution, is positive if more points lie to the left of the mean and negative if the opposite), and entropy (a measure of variation in a histogram, equals zero if all data are the same and increases as the data distribution becomes more irregular). Statistical Analysis
Statistical analyses were performed with MedCalc version 12.7.0.0 for Windows (MedCalc Software, Mariakerke, Belgium). Data are presented as medians with interquartile ranges. P < .05 was considered indicative of a statistically significant difference. The nonparametric Mann–Whitney U test was used to compare histogram parameters between bladder benign and
malignant lesions. To assess the differences in histogram parameters among bladder cancer of different T stages, the Kruskal– Wallis test was used, followed by the Mann–Whitney U test with Bonferroni correction for pairwise comparisons. Spearman correlation coefficients (r) were calculated to examine the correlations between histogram parameters and T stage. Receiver operating characteristic (ROC) curves for histogram parameters that were significantly different and the combination of significant measures were generated to help differentiate benign and malignant lesions and differentiate low- and high-stage tumors. For ROC analyses, the combination of significant measures were first calculated with logistic regression models and then used for the construction of the ROC curves. Diagnostic performance was determined by calculating the area under the ROC curve (AUC). Sensitivity and specificity for the classification were calculated. Cutoff values were established by calculating the maximal Youden index (Youden index = sensitivity + specificity 1). RESULTS Histopathologic Findings
Of the 52 patients, 7 were diagnosed with benign lesions (glandular cystitis, 2; eosinophilic cystitis, 3; and granulomatous inflammation, 2). In the 45 patients with bladder cancer, the 3
SUO ET AL
4 TABLE 1. Histogram Parameters of ADC-700 and ADC-1500 for Bladder Benign and Malignant Lesions b = 700 s/mm2
Benign Malignant P value
b = 1500 s/mm2
Mean
Skewness
Kurtosis
Entropy
Mean
Skewness
Kurtosis
Entropy
1.74 (1.50–2.02) 1.30 (1.03–1.51) .002
0.38 (0.11–0.61) 0.62 (0.21–1.06) .198
3.14 (2.17–4.11) 3.02 (2.71–4.50) .520
4.08 (3.73–4.52) 4.06 (3.62–4.35) .592
1.38 (1.30–1.55) 1.07 (0.88–1.25) .002
0.30 ( 0.17–0.65) 0.76 (0.36–1.28) .032
2.65 (2.20–3.64) 4.23 (3.13–5.63) .004
3.97 (3.77–4.32) 3.78 (3.39–4.04) .217
ADC, apparent diffusion coefficient. Data are expressed as medians with interquartile ranges. ADC mean values are expressed as 10 ADC skewness, kurtosis, and entropy are dimensionless.
3
mm2/s.
TABLE 2. Histogram Parameters of ADC-700 and ADC-1500 for Bladder Cancer of Different T Stages
#T1 T2 T3 T4 P value
b = 1500 s/mm2
Mean
Skewness
Kurtosis
Entropy
Mean
Skewness
Kurtosis
Entropy
1.38 (1.16–1.62) 1.39 (1.07–1.49) 0.97 (0.85–1.11) 0.95 (0.88–1.30) .001
0.64 ( 0.09–1.21) 0.59 (0.52–0.63) 0.66 (0.31–1.82) 0.69 (0.16–1.05) .892
3.06 (2.64–4.51) 2.82 (2.65–4.13) 3.44 (2.88–7.74) 2.89 (2.72–4.07) .548
4.06 (3.62–4.22) 4.33 (3.88–4.42) 3.74 (3.42–4.27) 4.13 (3.36–4.50) .277
1.23 (0.99–1.32) 1.15 (0.87–1.21) 0.81 (0.73–0.93) 0.78 (0.76–1.13)
