We retrospectively reviewed the medical records of 135 patients with suspected ovarian cancer or recurrence between February 2016 and May 2019 (supplementary Table 1). Of these, 103 patients (mean age, 55.5 years; age range, 11–80 years) who had undergone [18F]FDG PET/MRI, ceCT and pelvic dynamic ceMRI with obtained informed consent for the characterization, initial staging, and determination of the presence of residual disease after NAC and detection of recurrence based on the Japanese Imaging Guideline from Japan Radiological Society were included in the present study. Patients had completed [18F]FDG PET/MRI, ceCT and ceMRI within four4 months (mean, 29.1 days; range, 1–103 days) prior to treatment. The maximum interval among [18F]FDG PET/MRI, ceCT and ceMRI was 121 days (mean, 14.3 days; range, 0–121 days). Of the 103 patients, 67 patients with suspected ovarian cancer were characterized using [18F]FDG PET/MRI and ceMRI. The 56 patients with pathologically or cytologically proven diagnoses of ovarian cancer underwent initial staging with [18F]FDG PET/MRI, ceCT and ceMRI. Seven patients with pathologically or cytologically confirmed diagnoses of ovarian cancer were evaluated for the presence of residual disease after NAC and 11 with pathologically proven diagnosis of ovarian cancer were evaluated for recurrence with [18F]FDG PET/MRI and ceCT. This was a multi-center study, as 54 patients with data from ceCT and/or ceMRI were referred from other institutions, although all patients underwent [18F]FDG PET/MRI in our institution.
Patients fasted for at least 4 h prior to intravenous injection of 200 MBq of [18F]FDG. Fifty minutes after injection, patients were transferred to a whole-body 3.0-T PET/MR scanner (Signa PET/MR; GE Healthcare, Waukesha, WI). Anatomical coverage was from the vertex to the mid-thigh. PET acquisition was performed in 3-dimensional (3D) mode with 5.5 min/bed position (89 slices/bed) in 5–6 beds with a 24-slice overlap. A 2-point Dixon 3D volumetric interpolated T1-weighted fast spoiled gradient echo sequence was acquired at each table position and was used to generate MR attenuation correction (MR-AC) maps. Dixon-based MR-AC classifies body tissues into soft tissue, fat, and air. PET data were reconstructed by ordered subset expectation maximization (OSEM), selecting 14 subsets and 3 iterations, and post-smoothing with a 3-mm Gaussian filter. Reconstructed images were then converted to semiquantitative images corrected by the injected dose and the body weight of the subject as the standardized uptake value (SUV).
After whole-body scanning and a brief break for urination, the patient was repositioned in the PET/MR scanner. The pelvic PET scan was performed as a 3D acquisition in list mode with 15 min/bed position (89 slices/bed) in 1–2 beds with a 24-slice overlap. Regional PET data were reconstructed with OSEM selecting 16 subsets and 4 iterations, and post-smoothing with a 4-mm Gaussian filter. Reconstructed images were then converted to SUV images. For pelvic MRI, T2-weighted images were acquired in the sagittal, transaxial and coronal planes, using the following T2-weighted image parameters: TR, 4000–7000 ms; TE, 146 ms; section thickness, 4 mm; section overlap, 0 mm; flip angle, 100°; FOV, 240 × 240 mm; matrix, 384 × 384; two excitations; and bandwidth, 83.3 kHz.
Dynamic contrast-enhanced (DCE) MRI
Pelvic MRI was performed using a 3-T clinical scanner (Discovery MR750; GE Healthcare, Waukesha, WI) in 27 patients. To delineate the anatomy of the pelvis prior to pelvic DCE-MRI, T2-weighted imaging was performed in the sagittal, transaxial, and coronal planes. The following T2-weighted image parameters were used: TR, 3200–6000 ms; TE, 60–85 ms; section thickness, 4 mm; interval, 1 mm; flip angle, 111°; FOV, 240 × 240 mm; matrix, 320 × 224; two excitations; echo train length, 10; and bandwidth, 62.5 kHz. For DCE-MRI, a sagittal 3D fast spoiled-gradient-recalled T1-weighted sequence using the Dixon method with fat suppression (LAVA Flex; GE Healthcare) was used with the following parameters: TR, 5.0 ms; TE, 1.3 ms; section thickness, 3 mm; flip angle, 12°; FOV, 260 × 260 mm; matrix, 320 × 192; 1 excitation; and bandwidth, 166.7 kHz. After non-contrast images were acquired, 0.2 ml/kg of gadolinium-based contrast agent was injected at a rate of 2 ml/s using a contrast injector, followed by a 20-ml saline flush. Image sets were acquired at multiple phases, at 45, 80 and 120 s after initiation of injection. In 40 patients, DCE-MRI was performed at other institutes using 1.5-T clinical scanners (Magnetom Aera; Siemens Healthineers, or Signa HDe; GE Healthcare).
CT examinations covering the chest, abdomen and pelvis were performed using a 64-slice multidetector CT scanner (Discovery CT 750HD; GE Medical Systems, Milwaukee, WI) before and after intravenous administration of nonionic iodinated contrast material (iopamidol, Iopamiron 300; Schering, Berlin, Germany).
Images were analyzed on a dedicated workstation (Advantage Workstation 4.6; GE). Two board-certificated radiologists/nuclear medicine physicians, each with double certifications and specializing in gynecological imaging, evaluated the [18F]FDG PET/MRI, ceCT and ceMRI images retrospectively and reached consensus decisions. Images were evaluated for the following: a) characterization; b) tumor extension into the uterus, fallopian tubes, or ovaries (T2a); c) tumor extension into other nearby pelvic organs such as the bladder, sigmoid colon, or rectum (T2b); d) tumor extension into organs outside the pelvis, no bigger than 2 cm in extent (T3b); e) tumor extension into organs outside the pelvis, larger than 2 cm in extent (T3c); f) pelvic or para-aortic lymph nodes (N); g) distant metastasis (M); h) residual disease for IDS after NAC; and i) recurrence. The present study applied the TNM classification to evaluate the diagnostic value of the imaging modalities, because this anatomically based system separately records the primary and regional nodal extent of the tumor and the absence or presence of metastases. Diagnostic performance of [18F]FDG PET/MRI and ceMRI for assessing the characterization and extent of the primary tumor and [18F]FDG PET/MRI and ceCT for assessing nodal and distant metastases was evaluated. Both readers were blinded to the results of other imaging studies, histopathologic findings and clinical data. Each dataset was reviewed as the consensus decisions of the two readers after a minimum interval of three weeks to avoid any decision threshold bias due to reading-order effects. For CT and MRI interpretation, several previous standard criteria related to primary tumor and nodal or distant metastatic staging of ovarian cancer were used as the reference criteria (16). Swollen lymph nodes larger than 1 cm in short-axis diameter were graded as malignant. For [18F]FDG PET/MRI interpretations, the classification of lymph nodes as cancer-positive was based on the presence of focally appreciable metabolic activity above that of normal muscle; or asymmetric metabolic activity greater than that of normal-appearing lymph nodes at the same level in the contralateral pelvis, in a location corresponding to the lymph node chains on CT or MRI images, with reference to previous reports (12, 13). Furthermore, the presence of a central unenhanced area suggesting central necrosis or peripheral low attenuation suggesting a fatty hilum within lymph nodes was considered a benign sign. Tumor invasion of neighboring structures was decided primarily on the basis of CT or MRI findings, with reference to the [18F]FDG PET findings.
Histopathological results were used as the standard of reference for the characterization, T, N, and M staging, determination of residual disease after NAC and determination of recurrence. Because clinical and ethical standards of patient management do not require surgery or sampling of all detected lesions, a modified reference standard was used for lesions without histopathological sampling to take into account all prior and follow-up imaging. A decrease in size and/or SUVmax under therapy or an increase in size and/or SUVmax without therapy was regarded as a sign of malignancy. PET-negative and inconspicuous lesions with constant size were rated as benign.
The McNemar test was used to determine the statistical significance of differences in the accuracy of T, N and M staging as determined by PET/MRI, ceCT and ceMRI. Statistical analysis was performed using PRISM version 6.0 software (GraphPad, San Diego, CA). Differences at the level of p<0.05 were considered statistically significant.