The value of dynamic contrast–enhanced MRI in characterizing complex ovarian tumors

The institutional review board approved this retrospective study, and informed consent was waived. From February 2011 to October 2014, a total of 131 consecutive patients who were suspected of having complex ovarian tumors underwent DCE-MRI scanning. We excluded 15 patients who had not undergone surgery, 5 patients who received prior chemotherapy before MRI scanning, 4 patients with a poor imaging quality, and 5 patients without obvious solid components. Finally, 102 patients with 15 benign, 16 borderline, and 71 malignant ovarian tumors were included. All tumors were proven by surgery and histopathology within 2 weeks after the MR scan. The histopathological types of ovarian tumors are listed in Table 1. Patients’ ages ranged from 24 to 88 years (mean, 57 ± 15 years; median age, 57 years) in benign tumors, 16 to 58 years (mean, 37 ± 13 years; median age, 37 years) in borderline tumors, and 15 to 83 years (mean, 54 ± 12 years; median, 56 years) in malignant tumors, with a statistically significant difference between borderline and benign (P = 0.001) and between borderline and malignant tumors (P < 0.001).

MRI was performed using a 1.5-T scanner (Avanto or Espree; Siemens, Erlangen, Germany) with a phased-array abdominal coil. The patients lay in a supine position and breathed freely during acquisition. The parameters of conventional sequences are listed in Table 2. After completing four phases of scans, the axial DCE flashes 2D, T1-weighted imaging (T1WI) with fat saturation focusing on the solid components (solid portions, vegetations, thickened septa) was performed after the intravenous administration of 0.2 mmol/kg Gadopentetate dimeglumine (Gd-DTPA, Magnevist; Bayer Schering, Berlin, Germany) at a rate of 3.0 ml/s, followed by injection of 20 ml of normal saline to flush the tube. The scanning parameters were as follows: TR/TE = 5.6/2.38 ms, slice thickness 4 mm; gap 1.0 mm; matrix 256 × 256; field of view 280–340 mm, flip angle 10°, NEX = 1. A total of 30 phases of scans were obtained sequentially at 7-s intervals for 3 min 20 s. Every phase consists of 20 images.

MR images were reviewed independently by two radiologists (H.M.L. and J.W.Q., with 7 and 31 years of experience in gynecological imaging, respectively) who were blinded to the original reports (radiology, clinic, and histopathology). Any discrepancies were resolved by consensus. The MR features (tumor maximal diameters, bilaterality, shape, boundary, mass configuration, signal intensity [SI] on T2-weighted imaging (T2WI), ascites/pelvic fluid, and peritoneal implants/lymph nodes) were assessed. The time-signal intensity curve (TIC) was generated using the Mean Curve software package (Siemens). The semi-quantitative DCE-MRI parameters, including enhancement amplitude (EA), maximal slope (MS), and time of half rising (THR), were calculated. A round region of interest (ROI) of at least 1 cm² was placed at targeted areas referring to T2WI and contrast-enhanced images and avoiding areas such as hemorrhage, necrosis, and major vascular structures. TIC was classified into three types referring to study of Thomassin-Naggara et al. Type I curve showed a gradual slow rising; type III had a rapid rising followed by a plateau or slow-out curve. Type II curve had a moderate rising (lower than the type III and higher than the type I). The EA is the maximal value on the y-axis, MS is the maximal ratio of signal intensity increasing in one scanning period, and THR is the time from the injection of Gd-DTPA to reach half of the EA. In patients with bilateral masses (25 patients with ovarian malignant tumors, 6 patients with serous borderline ovarian tumors), only the most complex mass was evaluated to reduce intra-individual influence.

Statistical analysis was performed with SPSS 23.0 for Windows (SPSS Inc., Chicago, IL, USA). The differences in laterality, shape, boundary, mass configuration, SI on T2WI, ascites/pelvic fluid, peritoneal implants/lymph nodes, and TIC types among benign, borderline, and malignant tumors were compared using the Pearson Chi-square test or Fisher’s exact test. The differences in the age, diameters, and the parameters of EA, MS, and THR among the three groups of tumors were compared using one-way analysis of variance (ANOVA) or the nonparametric Kruskal-Wallis test. Receiver operating characteristic (ROC) curves were used to determine the cutoff values of semi-quantitative DCE-MRI parameters (MS and THR) for distinguishing between benign and malignant (borderline and malignant) tumors. A p-value less than 0.05 was considered statistically significant, and when the two-two comparisons were done by Pearson Chi-square test, a p-value less than 0.017 was considered statistically significant.

The MRI morphological features are summarized in Table 3. A significant difference was found in the bilaterality (P = 0.021, 0.006), shape (P = 0.001, 0.001), mass configuration (P = 0.009, 0.001), and peritoneal implants/lymph nodes (P = 0.006, 0.003) among the three groups and between benign and malignant tumours, respectively. A significant difference was also found in mass configuration between benign and borderline tumors (P = 0.003). However, there was no statistically significant difference between borderline and malignant tumors.

There were 59 (83%) type III, 12 (17%) type II, and no type I of TIC in malignant ovarian tumors; 4 (25%) type III, 9 (56%) type II, and 3 (19%) type I of TIC in borderline tumors; and 5 (33%) type I, 10 (67%) type II, and no type III of TIC in benign tumors, with a statistically significant difference between malignant and benign tumors (P < 0.001) and between malignant and borderline tumors (P < 0.001) (Figs. 1, 2 and 3). There was no statistically significant difference between borderline and benign tumors (P = 0.104). If type III TIC indicated a malignant tumor, type II and I TIC represented a benign or borderline tumor; it would generate a sensitivity of 83% (59/71), a specificity of 100% (15/15), and an accuracy of 86% (74/86) between malignant and benign tumors and a sensitivity, specificity, and accuracy of 83% (59/71), 75% (12/16), and 82% (71/87) between malignant and borderline tumors, respectively.

The mean values of EA, MS, and THR in the three groups of tumors are shown in Table 4. In benign, borderline, and malignant tumors, the mean value of EA was 220.2 ± 90.5, 269.3 ± 70.9, and 267.4 ± 86.2, respectively, without significant difference in overall and in two-two comparisons (all P > 0.05). The mean value of MS was higher in malignant than in borderline tumors and higher in borderline tumors than in benign tumors, whereas the mean value of THR in malignant tumors was lower than in borderline tumors and significantly lower in borderline tumors than in benign tumors. There was a significant difference overall (P = 0.001, < 0.001, respectively) between benign and malignant (P < 0,001, 0.001, respectively), borderline tumors (P = 0.013, 0.007, respectively). However, no significant difference was observed between borderline and malignant tumors (P = 0.19, 0.53, respectively).

Receiver operating characteristic (ROC) analysis was performed for the two statistically significant parameters of MS and THR to evaluate diagnostic ability in distinguishing malignant from benign tumors and borderline from benign tumors. The area under the curve (AUC), the cutoff value, and the diagnostic performance are listed in Table 5.