Histogram analysis of apparent diffusion coefficient map of diffusion-weighted MRI in endometrial cancer: a preliminary correlation study with histological grade

Abstract

Background

Until now, several investigators have explored the value of diffusion-weighted magnetic resonance imaging (DWI) for the preoperative tumor grading of endometrial cancer. However, the diagnostic value of DWI with quantitative analysis of apparent diffusion coefficient (ADC) has been controversial.

Purpose

To explore the role of histogram analysis of ADC maps based on entire tumor volume in determining the grade of endometrial cancer.

Material and Methods

This study was IRB-approved with waiver of informed consent. Thirty-three patients with endometrial cancer underwent DWI (b = 0, 600, 1000 s/mm²), and corresponding ADC maps were acquired. Regions of interest (ROIs) were drawn on all slices of the ADC map in which the tumor was visualized including areas of necrosis to derive volume-based histographic ADC data. Histogram parameters (5th–95th percentiles, mean, standard deviation, skewness, kurtosis) were correlated with histological grade using one-way ANOVA with Tukey-Kramer test for post hoc comparisons, and were compared between high (grade 3) and low (grades 1/2) grade using Student t-test. ROC curve analysis was performed to determine the optimum threshold value for each parameter, and their corresponding sensitivity and specificity.

Results

The standard deviation, quartile, 75th, 90th, and 95th percentiles of ADC showed significant differences between grades (P ≤ 0.03 for all) and between high and low grades (P ≤ 0.024 for all). There were no significant correlations between tumor grade and other parameters. ROC curve analysis yielded sensitivities and specificities of 75% and 96%, 62.5% and 92%, 100% and 52%, 100% and 72%, and 100% and 88%, using standard deviation, quartile, 75th, 90th, and 95th percentiles for determining high grade with corresponding areas under the curve (AUCs) of 0.787, 0.792, 0.765, 0.880, and 0.925, respectively.

Conclusion

Histogram analysis of ADC maps based on entire tumor volume can be useful for predicting the histological grade of endometrial cancer. The 90th and 95th percentiles of ADC were the most promising parameters for differentiating high from low grade.

Keywords

Genital reproductive MR diffusion/perfusion uterus histogram apparent diffusion coefficient histological grade

Introduction

Endometrial cancer is the most common female gynecologic cancer, and also the fourth most common cancer in women (1). The prognosis of endometrial cancer depends on several parameters including the tumor stage, involvement of pelvic lymph nodes, depth of myometrial invasion, histological grade, cell type, and patient age (2). In particular, the histological grade of the tumor is well-known as one of the most important prognostic factors regarding the metastasis to the lymph nodes and overall survival of the patient (3). Therefore, accurate preoperative assessment of the histological grade of endometrial cancer would be invaluable as it would affect the surgical approach regarding lymph node dissection (4).

Until now, several investigators have explored the value of diffusion-weighted magnetic resonance imaging (DWI) for preoperative tumor grading of endometrial cancer. Initially, Tamai et al. (5) reported that not only were apparent diffusion coefficient (ADC) values able to differentiate between endometrial cancer and normal endometrium, but were also helpful in differentiating between grades 1 and 3 endometrial cancers. However, the diagnostic value of DWI with quantitative ADC has been controversial. A few investigators agreed with the previous report in that ADC values were able to differentiate between the tumor grades (grade 1 versus grade 3 or grade 1 versus grades 2 and 3) (6 –8). On the other hand, contradictory findings have been reported (9 –12). Rechichi et al. (12) observed that the ADC value in endometrial cancers did not show a significant relationship with tumor grade, depth of myometrial invasion, or the presence of lymph node metastasis.

Most of the studies (5,6,8 –10,12) to date were based on selected regions of interest (ROIs) placed on a representative image of the tumor for analysis. Furthermore, the methods for placing the ROI were different among the investigators. While some placed the largest possible ROI that covered the tumor while avoiding artifacts at the border, most excluded the areas of necrosis (5,9,10,12). Another investigator (6,8) manually placed the ROI to include only the five voxels with minimum ADC values. The selection of an area within the representative section of the tumor can be subjective and prone to sampling bias (9). Moreover, whole tumor analysis has been demonstrated to be more representative of tumor heterogeneity compared with the largest cross-sectional area (13). We believe that this may have attributed to the discrepant results among previous reports regarding the diagnostic value of DWI to differentiate between the grades of endometrial cancer. It would be more objective to analyze the ADC of the entire tumor, which would be able to largely eliminate sampling bias. Furthermore, if histogram analysis is performed on the entire tumor, even including the areas of necrosis, additional information about the distribution of ADCs, and in turn, the heterogeneity of the tumor may be acquired.

In this context, investigators have previously performed histogram analysis of ADC maps to evaluate tumor grade in brain glioma, head and neck cancer, and cervical cancer (14 –16). However, to our knowledge, the value of volume-based histogram analysis of ADC maps to differentiate between histological grades of endometrial cancer has not been well demonstrated in the literature. The purpose of our study was to explore the role of histogram analysis of ADC maps based on entire tumor volume in determining the grade of endometrial cancer.

Material and Methods

Patient selection

Institutional Review Board approval was obtained for this retrospective study; requirement for informed consent was waived. Through a computerized search of our medical database from January 2007 to May 2013, we identified 60 patients with pathologically proven endometrial cancer who underwent preoperative MR evaluation at our institution. Exclusion criteria were as follows: (a) magnetic resonance imaging (MRI) performed without DWI (n = 17); and (b) no clear detection of the tumor on ADC maps (n = 10). As a result, a total of 33 patients were included. The diagnoses of the endometrial cancers included 30 endometrioid, two clear cell, and one serous adenocarcinoma. The endometrial cancers in our study consisted of American Joint Committee on Cancer (AJCC) grades 1 (n = 14, 42.4%), 2 (n = 11, 33.3%), and 3 (n = 8, 24.2%). The mean age of the study population was 52.1 years ± 11.3 (range, 30–84 years). All patients received total hysterectomy. The mean interval between MRI and surgery was 13 days ± 9.9 (range, 1–48 days).

MR protocol

The patients underwent MR examinations using a 1.5-T system (Signa Excite; GE Healthcare, Waukesha, WI, USA, or Sonata; Siemens, Erlangen, Germany). An eight-channel body array coil was used. A total of 20 mg of hyoscine butyl bromide (Buscopan, Boehringer Ingelheim Pharma, Ingelheim am Rhein, Germany) was injected intramuscularly before imaging in order to suppress bowel motion. The imaging protocol included: T2W-weighted (T2W) fast spin-echo imaging in the axial, coronal, and sagittal planes (repetition time [TR], 4500 ms; echo time [TE], 110 ms; flip angle [FA], 90°; field of view [FOV], 22 cm; slice thickness, 5 mm; intersection gap, 1 mm; matrix, 384 × 224; number of excitations [NEX], 2); and axial T1-weighted (T1W) fast spin-echo imaging (TR, 650 ms; TE, 9.2 ms; FA, 90°; FOV, 22 cm; slice thickness, 5 mm; intersection gap, 1 mm; matrix, 320 × 192; and NEX, 2. Echo-planar DWI with three b-values (b = 0, 600, 1000 s/mm²) were acquired in the axial plane before contrast injection (TR, 8000 ms; TE, 120 ms; FA, 90°; FOV, 30 cm; slice thickness, 5 mm; intersection gap, 1 mm; matrix, 160 × 160; and NEX, 3). ADC maps were generated on a voxel-by-voxel basis with the software built into the MR unit. In addition dynamic contrast-enhanced MRI (DCE-MRI) was acquired in the axial plane (TR, 4.2 ms; TE, 2.1 ms; FA, 12°; FOV, 30 cm; slice thickness, 5 mm; intersection gap, 2.5 mm; matrix, 256 × 160; and one signal acquisition obtained before and after intravenous bolus administration of 0.1 mmol/kg of gadopentetate dimeglumine [Magnevist; Berlex Laboratories, Wayne, NJ, USA] at a rate of 2 mL/s, and followed by a 20-mL saline flush by a power injector). Scanning delay times were 1, 3, and 5 min after contrast media injection.

Image analysis

After digital transfer of the data for the ADC maps from the picture archiving and communication system (PACS) workstation to a personal computer, analysis was performed with ImageJ (http://rsb.info.nih.gov/ij/) and a software program developed in-house (YJK and KGK) by using Visual C++ (Microsoft, Redmond, WA, USA). ROIs were drawn in each slice of the ADC map by one author (SW) along the margin of the tumor including all slices in which the tumor was visualized. Tumor boundaries were identified with reference to location and extensions on T2W images and DCE-MRI in consensus by two authors (SW and JYC) who were aware of the diagnosis of endometrial cancer but blinded to the histologic grade. Unlike the methods of the previous studies (5,9,10,12) which excluded the necrotic areas in the tumor on the basis of T1W and T2W MR images, we included these areas in the ROI (Fig. 1) to assess the heterogeneity of the tumor in its entirety (14). The data acquired from each slice were summated in order to obtain an ADC histogram for the entire tumor with the in-house-developed software.

Fig. 1.

A 54-year-old woman with grade 3 endometrial cancer showing how ROIs were drawn. ROIs were drawn on ADC maps (a) using ImageJ program with reference to (b) axial T2W MR image (TR/TE, 4500/110 ms) including all slices in which the tumor was visualized. Note that the necrotic areas within the tumor with high SI on (a) ADC map and (b) T2W MR image, and the lack of enhancement on (c) axial 1-min postcontrast T1W MR image (TR/TE, 4.2/2.1 ms) are included in the ROI.

The following parameters were calculated from the ADC histograms: (a) mean; (b) standard deviation; (c) 5th, 10th, 25th, 50th (or median), 75th, 90th, and 95th percentiles; (d) quartile (difference between the 25th and 75th percentiles); (e) kurtosis; and (f) skewness. Kurtosis and skewness reflect the shape of a histogram and were used to measure the asymmetry of the ADC value distribution around the mean. Kurtosis, which is a measure of the peakedness of the histogram: the values is 3 when the histogram is Gaussian, >3 with a sharper peak, and <3 with a flatter top. Skewness, which is a measure of the asymmetry of the histogram, is positive if the majority of the data is concentrated on the left of the histogram and negative if the majority of data is concentrated on the right.

Statistical analysis

All statistical analyses were done with PASW statistical software (version 18.0; SPSS, Chicago, IL, USA) and MedCalc version 11.1.1.0 for Windows (MedCalc Software, Mariakierke, Belgium). A two-tailed P value of <0.05 was considered to indicate a statistically significant difference. One-way analysis of variance with Tukey-Kramer post hoc comparisons was used to correlate the histogram parameters with the histologic grade of the tumors. The unpaired Student t-test was used to compare the histogram parameters between high grade (grade 3) and low grade (grades 1 and 2) (17,18). For the parameters that were significantly different between high- and low-grade tumors, receiver-operating characteristic (ROC) curve analysis was performed to calculate the sensitivity and specificity according to the threshold which yielded the greatest Youden index for differentiating low- and high-grade tumors. In order to compare the diagnostic capacity among the histogram parameters, the areas under the curves (Az) were compared by using the method of DeLong et al. (19).

Results

Differentiation of histological grade of endometrial cancer using ADC histogram parameters

The standard deviation for ADC was found to be significantly different between the tumor grades (P = 0.03). The standard deviation was significantly different between grades 1 and 3 (P = 0.03), and between grades 2 and 3 (P = 0.018), but not between grades 1 and 2 (P = 0.818). Moreover, the quartile values were significantly different between tumor grades (P = 0.017). These values were significantly different between grades 1 and 3 (P = 0.014), but not between grades 2 and 3 (P = 0.066) or grades 1 and 2 (P = 0.819).

In addition, the 75th, 90th, and 95th percentiles for ADC were significantly different between the tumor grades (P = 0.030, P = 0.002, and P < 0.001 for the 75th, 90th, and 95th percentiles, respectively). The 75th percentile was significantly different between grades 2 and 3 (P = 0.025), but not between grades 1 and 2 (P = 0.629), and grades 1 and 3 (P = 0.116). The 90th and 95th percentiles were significantly different between grades 1 and 3 (P = 0.009 and 0.001 for 90th and 95th percentiles, respectively) and between grades 2 and 3 (P = 0.02 and <0.001 for 90th and 95th percentiles, respectively), but not between grades 1 and 2 (P = 0.604 and 0.666 for 90th and 95th percentiles, respectively). Other histogram parameters, including the mean, percentiles (5th–50th) other than the 75th–95th percentiles, quartile, kurtosis, and skewness for ADC were not significantly difference between the grades (Table 1).

Table 1.

ADC histogram parameters in terms of histological grade of endometrial cancer.

Parameter	Grade 1 (n = 14)	Grade 2 (n = 11)	Grade 3 (n = 8)	P value
Mean	948 ± 177	877 ± 141	997 ± 435	0.584
Standard deviation*	142 ± 33	153 ± 44	216 ± 68	0.003
5th percentile	733 ± 182	646 ± 171	731 ± 208	0.458
10th percentile	781 ± 182	692 ± 161	858 ± 306	0.251
25th percentile	858 ± 182	776 ± 154	940 ± 298	0.247
50th percentile	944 ± 175	864 ± 147	1062 ± 280	0.113
75th percentile^†	1039 ± 168	972 ± 146	1205 ± 240	0.030
90th percentile*	1135 ± 167	1071 ± 146	1367 ± 184	0.002
95th percentile*	1194 ± 174	1141 ± 146	1470 ± 1117	<0.001
Quartile^†	181 ± 50	196 ± 54	265 ± 91	0.017
Kurtosis	3.29 ±0.69	3.74 ± 0.90	3.93 ± 1.14	0.221
Skewness	0.17 ± 0.48	0.51 ± 0.45	0.55 ± 0.63	0.153

Data are mean ± standard deviation (×10^–6 mm²/s).

Significant difference between grades 1 and 3 and between grades 2 and 3.

†

Significant difference between grade 1 and grade 3.

Differentiation of high- and low-grade endometrial cancer using ADC histogram parameters

The standard deviation (P = 0.024), quartile (P = 0.005), 75th percentile (P = 0.012), 90th percentile (P < 0.001), and 95th percentile (P < 0.001) for ADC proved to be significant histogram parameters for differentiating high- from low-grade endometrial cancers. The T2W images, ADC maps, and histograms of representative cases of low- and high-grade endometrial cancers are shown in Fig. 2. Other histogram parameters, including the mean, percentiles (5th–50th) other than the 75th–95th percentiles, kurtosis, and skewness for ADC were not significantly difference between high- and low-grade endometrial cancers (Table 2).

Fig. 2.

A 59-year-old woman with grade 3 endometrial cancer (a, b, c) and 52-year-old woman with grade 1 endometrial cancer (d, e, f). ADC histograms (a, d), axial (b, e) T2W MR images (TR/TE, 4500/110 ms), and (c, f) ADC maps (TR/TE, 8000/120 ms; b = 0, 600, 1000 s/mm²) of (a, b, c) grade 3 endometrial cancer of and (d, e, f) grade 1 endometrial cancer both demonstrate intermediate high T2W signal intensity and diffusion restriction. ADC histogram of grade 3 endometrial cancer (a) shows a higher relative frequency at high percentile ADCs compared with grade 1 endometrial cancer (d). This indicates that high-grade endometrial cancers contain more voxels with high ADCs, which suggests necrotic component. Note that there was no macroscopic necrosis in the high-grade endometrial cancer on (b) T2W MR image and on the (c) ADC map.

Table 2.

ADC histogram parameters of high- and low-grade endometrial cancer.

Parameter	Low grade (n = 25)	High grade (n = 8)	P value*
Mean	917 ± 163	997 ± 435	0.436
Standard deviation^†	147 ± 3	216 ± 68	0.024
5th percentile	695 ± 179	731 ± 208	0.641
10th percentile	742 ± 176	858 ± 306	0.188
25th percentile	822 ± 173	940 ± 280	0.172
50th percentile	908 ± 165	1062 ± 280	0.064
75th percentile^†	1009 ± 159	1204 ± 240	0.012
90th percentile^†	1106 ± 158	1367 ± 184	<0.001
95th percentile^†	1171 ± 161	1470 ± 117	<0.001
Quartile^†	188 ± 52	265 ± 91	0.005
Kurtosis	3.49 ± 0.80	3.93 ± 1.14	0.228
Skewness	0.32 ± 0.49	0.55 ± 0.63	0.294

Data are mean ± standard deviation (×10^–6 mm²/s).

The difference between grades was evaluated by using Student t-test.

†

Significant difference between low- (grades 1 and 2) and high-grade.

ROC analyses of ADC histogram parameters in predicting high-grade endometrial cancer

Table 3 summarizes the results of the ROC analyses of the ADC histogram parameters used to differentiate between high- and low-grade endometrial cancers. The Az for 95th (>1320 × 10^–6 mm²/s) and 90th percentiles (>1194 × 10^–6 mm²/s) for ADC showed the highest area under the curve (Az = 0.925 and 0.880 for 95th and 90th percentiles, respectively) among the histogram parameters with a sensitivity and specificity of 100% and 88% for 95th percentiles, and 100% and 72% for 90th percentiles (Fig. 3). These were significantly higher than that of the 75th percentile (Az = 0.765) using the criteria of >967 × 10^–6 mm²/s. However, the differences with that of the standard deviation (Az = 0.792) and quartile (Az = 0.787) were not significantly different (P >0.05 for all comparisons).

Fig. 3.

ROC curves for the 75th, 90th, and 95th percentiles, standard deviation, and quartile of the ADC histogram based on entire tumor volume for a prediction of high-grade endometrial cancer (grade 3). The area under the curve (Az) were highest for the 90th (Az = 0.880) and 95th percentiles (Az = 0.925), which was significantly higher than that of the 75th percentile (Az = 0.765). The Az for standard deviation and quartile were 0.792 and 0.787, respectively.

Table 3.

Sensitivity, specificity, and accuracy for differentiating high- and low-grade endometrial cancer

Parameter	Cut-off value*	Sensitivity^†	Specificity^†	Area under the curve^‡	P value
75th percentile	>967	100	52	0.765 (0.586, 0.894)	0.0022
90th percentile	>1194	100	72	0.880 (0.719, 0.967)	<0.0001
95th percentile	<1320	100	88	0.925 (0.778, 0.987)	<0.0001
Standard deviation	>194	75	96	0.792 (0.616, 0.913)	0.0202
Quartile value	>236	62.5	92	0.787 (0.610, 0.910)	0.0045

Data are in units of ×10^–6 mm²/s.

†

Sensitivity and specificity for identifying high-grade tumors.

‡

Data in parentheses are 95% confidence intervals.

Discussion

The results of our study suggest that the high percentile values (90th or 95th percentiles) as well as the standard deviation and quartile value of ADC histograms based on the entire tumor volume could be used to differentiate high- from low-grade endometrial cancer. As histological grade is one of the main prognostic factors in patients with endometrial cancer, preoperative prediction of the grade using histographic analysis of ADC maps may provide necessary clinical information and guide clinicians with more accurate staging and subsequent management. The histological grade of endometrial cancer is strongly predictive of the presence of nodal invasion. In patients with FIGO stage 1, <10% of the patients with low-grade (grades 1 and 2) endometrial cancer are expected to have nodal metastases, whereas high-grade (grade 3) histology carries a higher risk of up to 18% (3,20). Compared with other malignancy, it is relatively less difficult to preoperatively obtain pathologic specimens in patients with endometrial cancer. Therefore, preoperative cytology via endometrial sampling is used to diagnose endometrial cancer and to assess the histological grade; however, these specimens only contain a portion of the whole tumor and may not represent the true histological grade (21,22). It has been reported that 19% of patients diagnosed with grade 1 endometrial cancer were upgraded following surgical resection (23). Compared with preoperative cytology, ADC histogram based on the entire tumor volume used in our study aims to assess the whole tumor rather than a representative portion. Therefore, although more laborious, it may be regarded as a more sophisticated and accurate method to predict the histological grade of endometrial cancer.

To date, results of previous studies (5 –12) have been controversial over the value of ADC in the differentiation of high- from low-grade endometrial cancer. Especially, the minimum ADC within the tumor has demonstrated variable results regarding correlation with tumor grade. While Bharwani et al. (9) reported that minimum ADC was not useful in differentiating histological grade, Nakamura et al. (6,8) observed that it was helpful in discriminating grade 3 from grade 1 endometrial cancer. In our study, we correlated the 5th percentile, which may be representative of the minimum ADC while being less affected by artifacts (14), with histological grade and observed that there was no significant difference between high and low grades. The minimum ADC, which is known to be well correlated with highly cellular components within the tumor, is theoretically expected to decrease along with higher grade, since high-grade tumors have higher cellularity and subsequently decreased extracellular space and diffusivity of water molecules. We believe that several factors may be the cause of discrepant results. First, in previous studies, ROIs were placed at the single representative slice (6,8,9) and minimum ADC was obtained by manual selection of the five voxels containing the minimum ADC values (6,8). Compared with our histographic approach based on whole tumor volume, these methods may be prone to sampling bias. In addition, there is a possibility that the use of minimum values may result in inclusion of areas with artifacts. Another explanation could be that cellularity is only one of the factors that are used to determine histological grade. Other features such as nuclear atypia are not assessable with DWI (24).

In our study, the higher end values, especially the 90th and 95th percentiles of ADC histograms, were significantly higher in high-grade endometrial cancers. Furthermore, there was no significant difference in the mean ADC values between low- and high-grade endometrial cancers. We believe that the necrotic component more frequently seen in higher grade endometrial cancers may have been attributed to the higher frequency of voxels with high ADC values (25). While the expected decrease in ADC due to higher cellularity in higher-grade cancers, the combined effect of high ADCs from necrotic portions may have resulted in even higher (although not statistically significant) mean ADC values in the high than in low-grade endometrial cancers in our study. Until now, most studies have excluded macroscopic cystic or necrotic areas within the tumor (5,9 –11). However, considering the spatial resolution of DWI, we cannot exclude the possibility that microscopic necrotic portions could have been included in their selected ROIs. Contrary to the previous reports, we included the whole tumor without excluding the areas of necrosis. Our study results indicate that inclusion of the necrotic portions may be helpful in discriminating high- and low-grade endometrial cancers based on histographic ADC analysis based on the entire tumor.

Our study results demonstrate that the standard deviation and quartile value of ADC histograms were significantly different between the high- and low-grade endometrial cancers. The standard deviation and quartile value of the ADC histogram based on the entire tumor volume can be regarded to reflect the heterogeneity of the water diffusivity within the tumor. Considering that the higher end values, especially the 90th and 95th percentiles, showed significant difference between high- and low-grade tumors, while the lower end values did not show such difference, we speculate that the increased standard deviation and quartile values in higher grade endometrial cancers possibly reflect the higher necrotic component (25).

Apart from the intrinsic limitation of the retrospective nature of our study, several other limitations should be mentioned. First, the study population was relatively small. Although this did not limit us from deriving criteria using the high percentile values of ADC histograms for differentiating high- from low-grade endometrial cancer, further investigation that includes a larger population is warranted to strengthen the statistical power. Second, our study population did not include patients with normal endometrium or benign endometrial pathology. However, it was not the aim of our study to compare ADC parameters of endometrial cancer to that of normal or benign, as it has been demonstrated in previous studies (5,9). Third, there was a methodological challenge in defining the tumor boundary. Therefore, as in previous studies, it was necessary to refer to the T2W and DCE-MR images in order to properly place the ROI (10,12). However, it has been suggested that imperfect tumor delineation and subsequent variation among observers is negligible given the large data included in histogram analysis based on the entire tumor volume. Fourth, our study was performed using 1.5-T MRI systems with three b-values (b = 0, 600, 1000) to acquire DWI. The possibility of differing results using a higher magnetic field (3-T) and different number and magnitude of b-values cannot be excluded. However, regarding the comparison of our results and those of the aforementioned studies, we speculate that the different results may have been more attributed to the selection and positioning of the ROIs rather than the MR protocol itself, as most studies have been performed with the same magnetic field (1.5-T) and similar choice of b-values (b = 0–1000 s/mm²) (5,6,8,11,12). Fifth, when performing ROI measurements on ADC maps, there is a possibility that extreme ADCs resulting from DWI and ADC map misregistration artifacts could have been included. It can be presumed that the 5th or 95th percentiles obtained from ADC histograms are less affected by these artifacts, and therefore be more reliable for histographic analysis than the minimum or maximum ADC values which were not used in our study (15).

In conclusion, histogram analysis of ADC maps based on entire tumor volume can be useful for predicting the histological grade of endometrial cancer. The 90th and 95th percentiles of ADC were the most promising parameters for differentiating high- from low-grade endometrial cancers.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

References

Amant

Moerman

Neven

. Endometrial cancer. Lancet 2005; 366: 491–505.

Berman

Ballon

Lagasse

. Prognosis and treatment of endometrial cancer. Am J Obstet Gynecol 1980; 136: 679–688.

Creasman

Morrow

Bundy

. Surgical pathologic spread patterns of endometrial cancer. A gynecologic oncology group study. Cancer 1987; 60: 2035–2041.

Goff

Rice

. Assessment of depth of myometrial invasion in endometrial adenocarcinoma. Gynecol Oncol 1990; 38: 46–48.

Tamai

Koyama

Saga

. Diffusion-weighted MR imaging of uterine endometrial cancer. J Magn Reson Imaging 2007; 26: 682–687.

Nakamura

Imafuku

Nishida

. Measurement of the minimum apparent diffusion coefficient (adcmin) of the primary tumor and ca125 are predictive of disease recurrence for patients with endometrial cancer. Gynecol Oncol 2012; 124: 335–339.

Seo

Kim

Choi

. Endometrial cancer: utility of diffusion-weighted magnetic resonance imaging with background body signal suppression at 3T. J Magn Reson Imaging 2012; 37: 1151–1159.

Nakamura

Joja

Fukushima

. The preoperative suvmax is superior to adcmin of the primary tumour as a predictor of disease recurrence and survival in patients with endometrial cancer. Eur J Nucl Med Mol Imaging 2013; 40: 52–60.

Bharwani

Miquel

Sahdev

. Diffusion-weighted imaging in the assessment of tumour grade in endometrial cancer. Br J Radiol 2011; 84: 997–1004.

10.

Lin

Chang

. Myometrial invasion in endometrial cancer: diagnostic accuracy of diffusion-weighted 3.0-T MR imaging--initial experience. Radiology 2009; 250: 784–792.

11.

Cao

Gao

Sun

. Apparent diffusion coefficient of diffusion weighted MRI in endometrial carcinoma-relationship with local invasiveness. Eur J Radiol 2012; 81: 1926–1930.

12.

Rechichi

Galimberti

Signorelli

. Endometrial cancer: correlation of apparent diffusion coefficient with tumor grade, depth of myometrial invasion, and presence of lymph node metastases. Am J Roentgenol 2011; 197: 256–262.

13.

Kozarski

Ganeshan

. Assessment of tumor heterogeneity by CT texture analysis: can the largest cross-sectional area be used as an alternative to whole tumor analysis? Eur J Radiol 2013; 82: 342–348.

14.

Kang

Choi

Kim

. Gliomas: Histogram analysis of apparent diffusion coefficient maps with standard- or high-b-value diffusion-weighted MR imaging--correlation with tumor grade. Radiology 2011; 261: 882–890.

15.

Ahn

Choi

Kim

. Histogram analysis of apparent diffusion coefficient map of standard and high b-value diffusion MR imaging in head and neck squamous cell carcinoma: a correlation study with histological grade. Acad Radiol 2012; 19: 1233–1240.

16.

Downey

Riches

Morgan

. Relationship between imaging biomarkers of stage I cervical cancer and poor-prognosis histologic features: quantitative histogram analysis of diffusion-weighted MR images. Am J Roentgenol 2013; 200: 314–320.

17.

Creutzberg

van Putten

Warlam-Rodenhuis

. Outcome of high-risk stage Ic, grade 3, compared with stage I endometrial carcinoma patients: the postoperative radiation therapy in endometrial carcinoma trial. J Clin Oncol 2004; 22: 1234–1241.

18.

Alkushi

Kobel

Kalloger

. High-grade endometrial carcinoma: serous and grade 3 endometrioid carcinomas have different immunophenotypes and outcomes. Int J Gynecol Pathol 2010; 29: 343–350.

19.

DeLong

Clarke-Pearson

. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988; 44: 837–845.

20.

Boronow

Morrow

Creasman

. Surgical staging in endometrial cancer: clinical-pathologic findings of a prospective study. Obstet Gynecol 1984; 63: 825–832.

21.

Dubinsky

Parvey

Maklad

. The role of transvaginal sonography and endometrial biopsy in the evaluation of peri- and postmenopausal bleeding. Am J Roentgenol 1997; 169: 145–149.

22.

Park

Kim

Park

. Differentiation of the various lesions causing an abnormality of the endometrial cavity using MR imaging: emphasis on enhancement patterns on dynamic studies and late contrast-enhanced T1-weighted images. Eur Radiol 2006; 16: 1591–1598.

23.

Ben-Shachar

Pavelka

Cohn

. Surgical staging for patients presenting with grade 1 endometrial carcinoma. Obstet Gynecol 2005; 105: 487–493.

24.

Whittaker

Coady

Culver

. Diffusion-weighted MR imaging of female pelvic tumors: a pictorial review. Radiographics 2009; 29: 759–774.

25.

Jimenez-Ayala

Jimenez-Ayala Portillo

Iglesias

. Endometrial adenocarcinoma: prevention and early diagnosis. Monogr Clin Cytol 2008; 17: 1–91.