Relationship Between Imaging Biomarkers of Stage I Cervical Cancer ...

Wo m e n ’s I m a g i n g • O r i g i n a l R e s e a r c h Downey et al. Diffusion-Weighted MRI of Cervical Cancer

FOCUS ON:

Downloaded from www.ajronline.org by Inst Of Cancer Research on 05/23/14 from IP address 194.81.84.99. Copyright ARRS. For personal use only; all rights reserved

Women’s Imaging Original Research

Kate Downey 1 Sophie F. Riches1 Veronica A. Morgan1 Sharon L. Giles1 Ayoma D. Attygalle2 Tom E. Ind 3 Desmond P. J. Barton 3 John H. Shepherd 3 Nandita M. deSouza1 Downey K, Riches SF, Morgan VA, et al.

Keywords: cervical cancer, diffusion-weighted MRI, histogram, lymphovascular space invasion, tumor grade, tumor heterogeneity, tumor type DOI:10.2214/AJR.12.9545 Received July 4, 2012; accepted after revision August 31, 2012. Supported by Medical Research Council grant G0802470, by the CRUK and EPSRC Cancer Imaging Centre in association with the MRC and Department of Health (England) grant C1060/A10334, and NHS funding to the NIHR Biomedical Research Centre. 1

CRUK/EPSRC Cancer Imaging Centre, MRI Unit, Institute of Cancer Research and Royal Marsden NHS Foundation Trust, Royal Marsden Hospital, Downs Rd, Sutton, Surrey SM2 5PT, UK. Address correspondence to K. Downey ([email protected]).

2 Department of Histopathology, Royal Marsden NHS Foundation Trust, Sutton, Surrey, United Kingdom. 3 Department of Gynecological Oncology, Royal Marsden NHS Foundation Trust, Sutton, Surrey, United Kingdom.

AJR 2013; 200:314–320 0361–803X/13/2002–314 © American Roentgen Ray Society

314

Relationship Between Imaging Biomarkers of Stage I Cervical Cancer and Poor-Prognosis Histologic Features: Quantitative Histogram Analysis of DiffusionWeighted MR Images OBJECTIVE. The purpose of this study was to determine whether histogram analysis of apparent diffusion coefficient (ADC) values from diffusion-weighted MRI can be used to differentiate cervical tumors according to their histologic characteristics. SUBJECTS AND METHODS. Sixty patients with International Federation of Gynecology stage I cervical cancer underwent MRI at 1.5 T with a 37-mm-diameter endovaginal coil. T2-weighted images (TR/TE, 2000–2368/90) followed by diffusion-weighted images (TR/TE, 2500/69; b values, 0, 100, 300, 500, and 800 s/mm 2) were acquired. An expert observer drew regions of interest around a histologically confirmed tumor on ADC maps by referring to the T2-weighted images. Pixel-by-pixel ADCs were calculated with a monoexponential fit of data from b values of 100–800 s/mm2, and ADC histograms were obtained from the entire tumor volume. An independent samples Student t test was used to compare differences in ADC percentile values, skew, and kurtosis between squamous cell carcinoma and adenocarcinoma, well or moderately differentiated and poorly differentiated tumors, and absence and presence of lymphovascular space invasion. RESULTS. There was no statistically significant difference in ADC percentiles between squamous cell carcinoma and adenocarcinoma, but the median was significantly higher in well or moderately differentiated tumors (50th percentile, 1113 ± 177 × 10−6 mm2/s) compared with poorly differentiated tumors (50th percentile, 996 ± 184 × 10−6 mm2/s) (p = 0.049). Histogram skew was significantly less positive for adenocarcinoma compared with squamous cell carcinoma (p = 0.016) but did not differ between tumor grades. There was no significant difference between any parameter with regard to lymphovascular space invasion. CONCLUSION. Median ADC is lower in poorly compared with well or moderately differentiated tumors, while lower histogram-positive skew in adenocarcinoma compared with squamous cell carcinoma is likely to reflect the glandular content of adenocarcinoma.

T

he prognosis of stage I cervical cancer (confined to the cervix) depends on tumor volume [1], tumor type and grade, absence or presence of lymphovascular space invasion (LVSI) [2], and locoregional lymph node status. T2-weighted MRI is highly accurate for assessing tumor volume [3], even in small volume disease, if an endovaginal technique is used [4–7]. Tumor type and grade and LVSI are evaluated at histologic examination, when assessment is subject to the effects of random sampling. Squamous carcinomas, which constitute 70% of tumors, have a better response rate and time to complete response than do adenocarcinomas (approximately 30%) when matched for tumor grade and stage [8]. Adenocarcinomas have an overall worse prognosis [9, 10]. In addition,

well-differentiated tumors have a better prognosis than poorly differentiated tumors [2]. The presence of LVSI, that is, tumor cells in the vascular channels in and surrounding the tumor, is also a poor prognostic sign because of a higher incidence of metastasis [10–16]. The use of MRI to differentiate histologic features over the entire tumor in stage I disease would be useful as a prognostic biomarker, particularly when fertility-sparing options such as radical trachelectomy [17, 18] are being considered. Because tumors with different histologic features are very similar in appearance on conventional T2-weighted images, the addition of other mechanisms of image contrast to differentiate these histologic features would be invaluable. Diffusion-weighted MRI (DWI) entails analysis of tissue structure by sensitization

AJR:200, February 2013

Diffusion-Weighted MRI of Cervical Cancer TABLE 1: Apparent Diffusion Coefficient Percentile Values and Histogram Skew and Kurtosis for Squamous Cell Carcinoma Versus Adenocarcinoma


Apparent Diffusion Coefficient (× 10 −6 mm2 /s) Tumor type

No.

10th Percentile

25th Percentile

50th Percentile

75th Percentile

90th Percentile

Skew

Kurtosis

Squamous cell carcinoma

26

826 ± 197

915 ± 188

1018 ± 189

1140 ± 196

1326 ± 206

0.97 ± 0.67

2.70 ± 2.62

Adenocarcinoma

13

885 ± 204

997 ± 185

1101 ± 187

1234 ± 188

1407 ± 243

0.30 ± 0.97

1.33 ± 1.63

0.384

0.209

0.203

0.162

0.282

0.016

0.09

p Note—Values are mean ± SD.

TABLE 2: Apparent Diffusion Coefficient Percentile Values and Histogram Skew and Kurtosis for Well and Moderately Versus Poorly Differentiated Tumors Apparent Diffusion Coefficient (× 10 −6 mm2 /s) Grade

No.

10th Percentile

25th Percentile

50th Percentile

75th Percentile

90th Percentile

Skew

Kurtosis

Well or moderately differentiated

18

911 ± 192

1010 ± 181

1113 ± 177

1234 ± 196

1418 ± 224

0.48 ± 0.94

1.65 ± 2.42

Poorly differentiated

22

797 ± 189

894 ± 180

996 ± 184

1127 ± 183

1302 ± 200

0.92 ± 0.71

2.60 ± 2.38

0.068

0.050

0.049

0.084

0.092

0.099

0.22


TABLE 3: Apparent Diffusion Coefficient Percentile Values and Histogram Skew and Kurtosis for Tumors Without and With Lymphovascular Space Invasion Lymphovascular Space Invasion

Apparent Diffusion Coefficient (× 10 −6 mm2 /s) No.

10th Percentile

25th Percentile

50th Percentile

75th Percentile

90th Percentile

Skew

Kurtosis

Present

18

839 ± 185

939 ± 163

1042 ± 152

1171 ± 155

1354 ± 181

0.72 ± 0.70

2.07 ± 1.78

Absent

22

856 ± 210

939 ± 163

1054 ± 216

1178 ± 226

1354 ± 246

0.72 ± 0.95

2.25 ± 2.87

0.780

0.834

0.843

0.908

0.999

0.988

0.82


of the sequence to water diffusion in tissues over distances of 1–20 μm [19]. A pair of diffusion-sensitizing gradients applied on either side of a 180° refocusing pulse enables detection of the random molecular motion of water. Multiple gradient pairs of different diffusion weighting (b value, measured in seconds per square millimeter) are required to derive the rate of signal decay with increasing b value. This rate of decay obtained for every pixel in the image represents an apparent diffusion coefficient (ADC) and can be displayed as an ADC map over the area of data acquisition. In tumor tissue, increased cellularity compared with nontumor tissue restricts water movement between cells so that the tumor retains signal intensity on DW images with a corresponding low value on ADC maps. ADC values in cervical cancer have been found to be significantly lower than in nonmalignant cervical epithelium and cervical intraepithelial neoplasia [4, 20–24]. Although most studies of cancer have been conducted with mean ADC values from

tumor regions of interest (ROIs), it is increasingly recognized that it is possible to examine the heterogeneity of diffusion in the tumor region by use of ADC histogram analysis. This method has proved reproducible in brain imaging [25] and ovarian cancer [26] but has not, to our knowledge, been exploited in the evaluation of cervical malignancies. The purpose of this study was to determine whether histogram analysis of ADCs from whole tumor regions by use of percentile and histogram skew and kurtosis can be used to differentiate cervical tumors according to the histologic characteristics of tumor type, grade, and absence or presence of LVSI.

cal cancer staged clinically and histologically as International Federation of Gynecology Ia, Ib1, or Ib2 were enrolled in this prospective study. On the basis of the imaging findings, patients were subsequently assigned to management with primary surgery or chemoradiotherapy. In the surgery group, to avoid inclusion of MRI false-positive lesions, only patients with residual invasive disease confirmed histologically after extended cone biopsy, radical trachelectomy, or radical hysterectomy after MRI were included in the final analysis. Of the 60 patients recruited, 20 were subsequently excluded. Nineteen of these patients had no visible tumor around which to draw an ROI, and in one patient, the b values used during data acquisition were wrongly assigned.

Subjects and Methods Patients

Imaging

The study was approved by local research ethics committee, and all patients gave written informed consent. From March 2006 to May 2009, 60 patients (age range, 24–74 years; mean age, 40 years) with the diagnostic biopsy finding of cervi-

Patients were supine for imaging with a 1.5-T MRI system (Intera, Philips Healthcare) with a 37-mm-diameter endovaginal coil positioned around the cervix [5]. After manual examination to evaluate the position of the cervix, the coil was inserted endovaginally and immobilized with an external clamp to minimize

AJR:200, February 2013 315


Downey et al.

A

B

C

Fig. 1—45-year-old woman with stage Ib squamous cell carcinoma of cervix. A–C, Sagittal (A), transverse (B), and coronal (C) T2-weighted (TR/TE, 2000/90, 11-cm FOV, 3-mm slice thickness) MR images through cervix show homogeneous mass (arrow) of intermediate signal intensity. D, Coronal apparent diffusion coefficient map shows homogeneous area of diffusion restriction in tumor region of interest (outline).

D its movement. Air introduced into the vagina during the examination and insertion of the receiver coil was aspirated with a Ryle tube (Pennine Healthcare) to reduce susceptibility artifacts on the images at the air-tissue interface. T2-weighted fast spin-echo images (TR/TE, 2000/90 (coronal and axial) and 2368/90 (sagittal); FOV, 11 cm; voxel size, 0.55 mm3; slice thickness, 3 mm) were acquired for assessment of the cervix in three planes (sagittal, coronal, and axial to the cervix) followed by single-shot echo-planar DW images (TR/TE, 2500/69; b values, 0, 100, 300, 500, and 800 s/mm2; FOV, 20 cm; voxel size, 17.6 mm3; number of signals averaged, 2) over the same volume in the coronal plane. The data acquisition matrix used for DWI was 96 × 95 (reconstructed to 128 × 128) with a slice thickness of 3 or 4 mm depending on the position of the individual lesion.

Data Analysis An expert observer with 20 years’ experience in endocavitary imaging calculated the tumor volume in each patient by summing areas defined by ROIs drawn around the tumor on every slice in which the tumor was found on the T2-weighted

316

images and multiplying by slice thickness. ADC maps were generated with a monoexponential fit of data from all b values by use of system software available on the MRI unit. Anonymous data were exported offline. On the ADC maps with reference to the T2-weighted images, the same observer, who was blinded to the histologic findings on the tumor, drew ROIs around the tumor in every slice in which the tumor was found. ADC values from tumor ROIs were calculated with in-house software (IDL 6.1, Research Systems), and ADC histograms were obtained from the entire tumor volume. The 10th, 25th, 50th, 75th, and 90th percentile pixel values were documented, and the skew and the kurtosis of the histogram were recorded. Skew and kurtosis reflect the shape of a histogram and were used to measure the asymmetry of the ADC value distribution around the mean. A positive skew had most of the values to the left of the mean and an asymmetric tail toward higher ADC values. A negative skew had most of the values to the right of the mean and an asymmetric tail toward lower ADC values. Kurtosis represented the concentration of values around the mean and reflected the peak of the distribution. In a normal distribution, skew is 0 and kurtosis is 3.

Histologic Analysis Histologic preparation and analysis of specimens were performed in a standard way for our institution with the cervix cut transversely from the internal os to approximately 1 cm from the ectocervix in 4-mm slices. The rest of the cervix was then sectioned longitudinally. Sections were embedded in paraffin and stained with H and E. Histologic analysis included classification of tumor type, differentiation, and

presence or absence of LVSI. Tumor for histologic analysis was acquired at radical hysterectomy, radical trachelectomy, or cold knife cone biopsy of 36 patients. The other four patients had no post-MRI histologic specimen for comparison because as a result of the MRI findings, the stage of disease was increased from I to IIB (two patients) or IV (two patients). These patients needed chemoradiotherapy rather than surgery and were therefore classified according to the diagnostic histologic result.

Statistical Analysis Nonparametric statistical calculations were performed with SPSS software (version 18 for Microsoft Windows, SPSS). An independent samples Student t test was used to compare differences in the histogram mean percentile ADC values and skew between good and poor prognosis histologic features, namely, squamous cell carcinoma versus adenocarcinoma (excluding the one patient with an adenosquamous histologic result), well or moderately differentiated versus poorly differentiated tumors, and the absence versus the presence of LVSI. A value of p < 0.05 was considered significant. In patients who underwent radical surgery, the histologic parameters used were those determined at the final histologic examination of the resected specimen. For those treated by chemoradiotherapy, the histologic parameters used were those from diagnostic biopsy. Receiver operating curve (ROC) analysis was used to establish the sensitivity and specificity of ADC cutoff values for separating tumors by type and grade when appropriate.

Results Of the 60 patients recruited, 20 were subsequently excluded. In 19 of these patients, there


A

B

C

Fig. 2—56-year-old woman with stage Ib adenocarcinoma of cervix. A–C, Sagittal (A), transverse (B), and coronal (C) T2-weighted (TR/TE, 2000/90, 11-cm FOV, 3-mm slice thickness) MR images through cervix show heterogeneous mass of intermediate signal intensity (arrow). D, Coronal apparent diffusion coefficient map shows heterogeneous area of diffusion restriction in tumor region of interest (outline).

was no visible tumor around which to draw an ROI, and in one patient the b values used during data acquisition were wrongly assigned. The median tumor volume calculated from the T2-weighted images of the 40 patients analyzed was 2.0 cm3 (range, 0.2–30 cm3). At histologic examination, 26 tumors were squamous cell carcinoma, 13 were adenocarcinoma, and one was adenosquamous carcinoma. Eighteen tumors were well or moderately differentiated, and 22 were poorly differentiated. Eighteen specimens had LVSI, and 22 did not. The ADC results are summarized in Tables 1–3. There was no statistically significant difference in the ADC percentiles for squamous cell carcinoma compared with adenocarcinoma (Figs. 1–3). However the 50th percentile ADC values were significantly higher for well or moderately differentiated (1113 ± 177 × 10−6 mm2/s) compared with poorly differentiated (996 ± 184 × 10−6 mm2/s) (p = 0.049)

Discussion Our preliminary findings suggest a potential role of ADC histogram analysis in the identification of adverse histologic characteristics of stage I cervical cancer. The significantly less positively skewed histogram for 12

Squamous cell carcinoma Adenocarcinoma

10 8 6 4 2 0

300 425 550 675 800 925 1050 1175 1300 1425 1550 1675 1800 1925 2050

D

tumors (Figs. 4–6). Differences in the 25th percentile values approached significance (well or moderately differentiated 25th percentile, 1010 ± 181 × 10−6 mm2/s; poorly differentiated 25th percentile, 894 ± 180 × 10−6 mm2/s) (p = 0.049). Histogram skew was significantly different between tumor types (p = 0.016) (Table 1). The skew was significantly less positive for adenocarcinomas but did not differ with degree of differentiation (Table 2). Kurtosis reflected that of a normal (squamous cell carcinomas and poorly differentiated tumors) or uniform (adenocarcinomas and well or moderately differentiated tumors) distribution but was not significantly different between groups (Tables 1 and 2). There was no significant difference between parameters with regard to absence or presence of LVSI (Table 3). A cutoff 50th percentile ADC value of 971 × 10 −6 mm 2 /s separated well or moderately from poorly differentiated tumors with a sensitivity of 78% and specificity of 46% (area under ROC curve [AUC], 0.66). A cutoff histogram skew of 0.45 separated squamous cell carcinomas from adenocarcinomas with a sensitivity of 77% and a specificity of 62% (AUC, 0.72). The numbers of patients in the well or moderately differentiated squamous cell cancer (n = 8) and poorly differentiated adenocarcinoma (n = 3) groups were too small for a formal comparison of these groups.

Percentage of Pixels


Diffusion-Weighted MRI of Cervical Cancer

ADC (× 10−6 mm2/s)

Fig. 3—Histogram distribution of apparent diffusion coefficient (ADC) values of squamous cell carcinoma (tumor volume, 22 cm 3) compared with adenocarcinoma (tumor volume, 2 cm3) of cervix shows skew of distribution of adenocarcinomas is less positively skewed than that of squamous cell carcinomas because of increase in proportion of pixels with high ADC values in adenocarcinoma compared with squamous cell carcinoma.



Downey et al.

A

B

C

Fig. 4—26-year-old woman with stage Ib well-differentiated carcinoma of cervix. A–C, Sagittal (A), transverse (B), and coronal (C) T2-weighted (TR/TE, 2000/90, 11-cm FOV, 3-mm slice thickness) MR images through cervix show homogeneous mass of intermediate signal intensity (arrow). D, Coronal apparent diffusion coefficient (ADC) map shows area of diffusion restriction that has relatively high ADC values in tumor region of interest (outline).

D adenocarcinoma compared with squamous cell carcinoma is likely to reflect the heterogeneous mix of glandular elements (high ADC values) with cellular tumor architecture (low ADC values) of adenocarcinoma, shifting ADC values to the right. In squamous lesions, the cell distribution is closer to homogeneous without extracellular spaces of glandular elements, making the distribution more strongly positively skewed. Kurtosis approached a normal distribution for squamous cell carcinomas and poorly differentiated tumors in which tumor cell density and distribution were homogeneous. For adenocarcinomas and well or moderately differentiated tumors in which glandular architecture was more prominent, the distribution of ADC values was closer to uniform across the range. However, because of large SDs in this small sample, differences between groups were not significant. The significantly lower ADC percentile values for poorly differentiated tumors are likely related to the microstructure of these tumors because the signal attenuation with increasing diffusion weighting reflects increased cellularity with hydrophobic cellular

318

membrane integrity, tissue disorganization, and extracellular space tortuosity. In these tissues, water can move less freely and therefore signal is retained; the increasing diffusion weighting results in a low slope of decay (low ADC value) compared with that of normal tissue. Studies of other tumor types have documented differences in mean ADC with histologic characteristics [27–30], but the investigators did not probe the heterogeneity of ADC values in the entire tumor volume. In our study, only use of the 50th percentile to differentiate grades of tumor was associated with a statistically significant difference. An AUC of 0.66 suggests a degree of overlap in ADC values between well or moderately and poorly differentiated tumors. This finding is likely due to the inclusion of moderately differentiated tumors in the well-differentiated group because their exclusion yielded an AUC of 0.83. Further studies with larger patient numbers or with higher field strengths are warranted to enable the use of ADC as a biomarker of tumor grade. In our cohort there was no significant difference between tumor types with respect to mean or percentile ADC values even though adenocarcinomas are composed of less solid and more glandular tissue than their squamous cell counterparts, which would theoretically result in a greater proportion of pixels with a higher ADC value. The mean ADC values obtained from both squamous cell carcinomas and adenocarcinomas are in keeping with values obtained from other tumor types in which the histologic features vary from glandular to more solid elements [27–30].

LVSI is a poor prognostic indicator but does not alter cellular architecture significantly, so it was not surprising to find no significant difference in ADC values between tumors with LVSI and those without. The incidence of LVSI was much greater in poorly differentiated tumors (64%, 14/22) compared with 22% (4/18) of well or moderately differentiated tumors, suggesting that identifying poorly differentiated tumors may identify most tumors with LVSI. ADC histogram analysis has been found to be a reproducible technique [25]. Its use in neuroradiology has enabled characterization of histologic differences between low-grade brain tumor subtypes. Tozer et al. [31] used ADC histograms to differentiate low-grade astrocytomas from oligodendrogliomas according to their different histologic characteristics. ADC histogram analysis has been used to predict response to treatment of brain tumors as measured by progression-free survival. ADC histograms also have been found predictive of response to antiangiogenic therapy in patients with recurrent high-grade gliomas [32, 33] and to help identify early response in ovarian cancer [26]. In the latter group, a shift in ADC percentile values and histogram skew occurred in patients with increased progression-free survival and CA-125 reduction. Accurate quantification of ADC values requires a good signal-to-noise ratio and an optimal range and choice of b values. In addition, fitting of the data with a monoexponential or multiexponential model [34] potentially significantly affects the final data. Previous studies of DWI in the analysis of cervical malignancies


A

B

C

Fig. 5—42-year-old woman with stage Ib poorly differentiated carcinoma of cervix. A–C, Sagittal (A), transverse (B), and coronal (C) T2-weighted (TR/TE, 2000/90, 11-cm FOV, 3-mm slice thickness) images through cervix show homogeneous mass of intermediate signal intensity (arrow). D, Coronal apparent diffusion coefficient map shows area of striking diffusion restriction in tumor region of interest (outline).

D

12

Poorly differentiated Well differentiated

10 Percentage of Pixels

8 6 4 2 0

300 425 550 675 800 925 1050 1175 1300 1425 1550 1675 1800 1925 2050


Diffusion-Weighted MRI of Cervical Cancer

ADC (× 10−6 mm2/s)

Fig. 6—Histogram distribution of apparent diffusion coefficient (ADC) values of well-differentiated (tumor volume, 2.8 cm3) compared with poorly differentiated (tumor volume, 22 cm3) tumor in cervix shows left shift of histogram to lower ADC in poorly differentiated tumor.

have been effectively conducted with a monoexponential model for signal decay from diffusion [24]. In several studies of kidney, brain, prostate, and rectal lesions, however, a biexponential or multiexponential model for fitting the data were more accurate than a monoexponential model [35–37] and allowed extraction of relative components of fast (perfusion) and slow (true) diffusion effects. However, fitting the data to a biexponential or multiexponential model requires a large range of b values, is time-consuming at acquisition, and cannot be achieved with the commercially available MRI system software. We used a monoexponential model and omitted the b0 value in the calculation of ADC to reduce the effects of any perfusion component. Because the instability of the gradients at low b values leads to significant error in measurement unless a large number of b values are used, we did not separately examine the perfusion component of diffusion in these tumors. There were several limitations to this study. First, the sample size was relatively small. Patients were often referred after diagnostic cone biopsy at which a small stage I cervical cancer was often completely excised. This occurred in 19 patients who were therefore excluded from the final analysis. Second, ROIs were outlined by a single radiologist, introducing observer bias. In a previous study of ADC values of invasive cervical carcinoma compared with nontumor cervical epithelium [4], two radiologists had good interobserver agreement with a kappa score of 0.8. Third, the distortion of the DW im-

ages meant that it was not possible to transfer ROIs from T2-weighted images to ADC maps. Therefore, bias was introduced by selection only of pixels indicative of tumor on ADC maps. However, the ROIs were drawn with visual reference to the T2-weighted images to reduce this bias. Finally, the presence of hemorrhage can alter the ADC measurement, although no hemorrhage was found on the T1-weighted images obtained with the external array coil for evaluation of pelvic lymphadenopathy. In summary, histogram analysis of ADC values in cervical tumors has potential for differentiating histologic subtypes and tumor grades, providing information about the entire tumor, and avoiding the limitation of data from randomly sampled histologic specimens. The identification of histologic features of cervical tumors associated with good or poor prognosis together with measurements of tumor volume may influence decisions on extent and timing of planned surgery, especially for women for whom preservation of fertility is a consideration. References 1. Wang JZ, Mayr NA, Zhang D, et al. Sequential magnetic resonance imaging of cervical cancer: the predictive value of absolute tumor volume and regression ratio measured before, during, and after radiation therapy. Cancer 2010; 116:5093–5101 2. Takeda N, Sakuragi N, Takeda M, et al. Multivariate analysis of histopathologic prognostic factors for invasive cervical cancer treated with radical hysterectomy and systematic retroperitoneal



Downey et al. lymphadenectomy. Acta Obstet Gynecol Scand 2002; 81:1144–1151 3. Jung DC, Ju W, Choi HJ, et al. The validity of tumour diameter assessed by magnetic resonance imaging and gross specimen with regard to tumour volume in cervical cancer patients. Eur J Cancer 2008; 44:1524–1528 4. Charles-Edwards EM, Messiou C, Morgan VA, et al. Diffusion-weighted imaging in cervical cancer with an endovaginal technique: potential value for improving tumor detection in stage Ia and Ib1 disease. Radiology 2008; 249:541–550 5. deSouza NM, Scoones D, Krausz T, Gilderdale DJ, Soutter WP. High-resolution MR imaging of stage I cervical neoplasia with a dedicated transvaginal coil: MR features and correlation of imaging and pathologic findings. AJR 1996; 166:553–559 6. deSouza NM, Whittle M, Williams AD, et al. Magnetic resonance imaging of the primary site in stage I cervical carcinoma: a comparison of endovaginal coil with external phased array coil techniques at 0.5T. J Magn Reson Imaging 2000; 12:1020–1026 7. deSouza NM, Dina R, McIndoe GA, Soutter WP. Cervical cancer: value of an endovaginal coil magnetic resonance imaging technique in detecting small volume disease and assessing parametrial extension. Gynecol Oncol 2006; 102:80–85 8. Katanyoo K, Sanguanrungsirikul S, Manusirivithaya S. Comparison of treatment outcomes between squamous cell carcinoma and adenocarcinoma in locally advanced cervical cancer. Gynecol Oncol 2012; 125:292–296 9. Galic V, Herzog TJ, Lewin SN, et al. Prognostic significance of adenocarcinoma histology in women with cervical cancer. Gynecol Oncol 2012; 125:287–291 10. Hopkins MP, Morley GW. A comparison of adenocarcinoma and squamous cell carcinoma of the cervix. Obstet Gynecol 1991; 77:912–917 11. Chung CK, Stryker JA, Ward SP, Nahhas WA, Mortel R. Histologic grade and prognosis of carcinoma of the cervix. Obstet Gynecol 1981; 57:636–642 12. Delgado G, Bundy B, Zaino R, Sevin BU, Creasman WT, Major F. Prospective surgical-pathological study of disease-free interval in patients with stage IB squamous cell carcinoma of the cervix: a Gynecologic Oncology Group study. Gynecol Oncol 1990; 38:352–357 13. Kosary CL. FIGO stage, histology, histologic grade, age and race as prognostic factors in determining survival for cancers of the female gynecological system: an analysis of 1973-87 SEER cases

of cancers of the endometrium, cervix, ovary, vulva, and vagina. Semin Surg Oncol 1994; 10:31–46 14. Kristensen GB, Abeler VM, Risberg B, Trop C, Bryne M. Tumor size, depth of invasion, and grading of the invasive tumor front are the main prognostic factors in early squamous cell cervical carcinoma. Gynecol Oncol 1999; 74:245–251 15. Morice P, Piovesan P, Rey A, et al. Prognostic value of lymphovascular space invasion determined with hematoxylin-eosin staining in early stage cervical carcinoma: results of a multivariate analysis. Ann Oncol 2003; 14:1511–1517 16. Sevin BU, Lu Y, Bloch DA, Nadji M, Koechli OR, Averette HE. Surgically defined prognostic parameters in patients with early cervical carcinoma: a multivariate survival tree analysis. Cancer 1996; 78:1438–1446 17. Gien LT, Covens A. Fertility-sparing options for early stage cervical cancer. Gynecol Oncol 2010; 117:350–357 18. Milliken DA, Shepherd JH. Fertility preserving surgery for carcinoma of the cervix. Curr Opin Oncol 2008; 20:575–580 19. Charles-Edwards EM, deSouza NM. Diffusionweighted magnetic resonance imaging and its application to cancer. Cancer Imaging 2006; 6:135– 143 20. Liu Y, Bai R, Sun H, Liu H, Wang D. Diffusionweighted magnetic resonance imaging of uterine cervical cancer. J Comput Assist Tomogr 2009; 33:858–862 21. McVeigh PZ, Syed AM, Milosevic M, Fyles A, Haider MA. Diffusion-weighted MRI in cervical cancer. Eur Radiol 2008; 18:1058–1064 22. Messiou C, Morgan VA, De Silva SS, Ind TE, deSouza NM. Diffusion weighted imaging of the uterus: regional ADC variation with oral contraceptive usage and comparison with cervical cancer. Acta Radiol 2009; 50:696–701 23. Naganawa S, Sato C, Kumada H, Ishigaki T, Miura S, Takizawa O. Apparent diffusion coefficient in cervical cancer of the uterus: comparison with the normal uterine cervix. Eur Radiol 2005; 15:71–78 24. Payne GS, Schmidt M, Morgan VA, et al. Evaluation of magnetic resonance diffusion and spectroscopy measurements as predictive biomarkers in stage 1 cervical cancer. Gynecol Oncol 2010; 116:246–252 25. Steens SC, Admiraal-Behloul F, Schaap JA, et al. Reproducibility of brain ADC histograms. Eur Radiol 2004; 14:425–430 26. Kyriazi S, Collins DJ, Messiou C, et al. Metastatic

ovarian and primary peritoneal cancer: assessing chemotherapy response with diffusion-weighted MR imaging–value of histogram analysis of apparent diffusion coefficients. Radiology 2011; 261:182–192 27. Avcu S, Koseoglu MN, Ceylan K, Bulut MD, Unal O. The value of diffusion-weighted MRI in the diagnosis of malignant and benign urinary bladder lesions. Br J Radiol 2011; 84:875–882 28. Hambrock T, Somford DM, Huisman HJ, et al. Relationship between apparent diffusion coefficients at 3.0-T MR imaging and Gleason grade in peripheral zone prostate cancer. Radiology 2011; 259:453–461 29. Razek AA, Fathy A, Gawad TA. Correlation of apparent diffusion coefficient value with prognostic parameters of lung cancer. J Comput Assist Tomogr 2011; 35:248–252 30. Zou QG, Xu HB, Liu F, Guo W, Kong XC, Wu Y. In the assessment of supratentorial glioma grade: the combined role of multivoxel proton MR spectroscopy and diffusion tensor imaging. Clin Radiol 2011; 66:953–960 31. Tozer DJ, Jager HR, Danchaivijitr N, et al. Apparent diffusion coefficient histograms may predict low-grade glioma subtype. NMR Biomed 2007; 20:49–57 32. Nowosielski M, Recheis W, Goebel G, et al. ADC histograms predict response to anti-angiogenic therapy in patients with recurrent high-grade glioma. Neuroradiology 2011; 53:291–302 33. Pope WB, Lai A, Mehta R, et al. Apparent diffusion coefficient histogram analysis stratifies progression-free survival in newly diagnosed bevacizumab-treated glioblastoma. AJNR 2011; 32:882–889 34. Le Bihan D, Breton E, Lallemand D, Aubin ML, Vignaud J, Laval-Jeantet M. Separation of diffusion and perfusion in intravoxel incoherent motion MR imaging. Radiology 1988; 168:497–505 35. Chandarana H, Lee VS, Hecht E, Taouli B, Sigmund EE. Comparison of biexponential and monoexponential model of diffusion weighted imaging in evaluation of renal lesions: preliminary experience. Invest Radiol 2011; 46:285–291 36. Mulkern RV, Gudbjartsson H, Westin CF, et al. Multi-component apparent diffusion coefficients in human brain. NMR Biomed 1999; 12:51–62 37. Riches SF, Hawtin K, Charles-Edwards EM, de Souza NM. Diffusion-weighted imaging of the prostate and rectal wall: comparison of biexponential and monoexponential modelled diffusion and associated perfusion coefficients. NMR Biomed 2009; 22:318–325

F O R YO U R I N F O R M AT I O N

The reader’s attention is directed to the commentary on this article, which appears on the preceding pages.

320
