RPS 2007 - Ovarian imaging today: what's new?

Author Block: M. Wang¹, J. perucho¹, L. Han², E. M. F. Wong¹, G. Ho¹, E. Y. P. Lee¹, P. Ip¹; ¹Hong Kong/CN, ²Guangzhou/CN
Purpose or Learning Objective: To evaluate the ability of contrast-enhanced CT (ceCT)-based radiomic features in differentiating high-grade serous carcinoma (HGSC) and non-HGSC of epithelial ovarian carcinoma (EOC).
Methods or Background: A total of 341 patients with EOC were enrolled in this multicentre retrospective study. Centre A with 282 patients was designated as the internal cohort and was divided into an internal training set (n = 225) and an internal test set (n = 57). Centre B with 59 patients, was designated as the external cohort and was used for external testing. To address the class imbalance between the distribution of HGSC and non-HGSC, synthetic balanced samples were generated using Random Over-Sampling Examples (ROSE). Tumours were manually delineated on each slice of ceCT images to encompass the entire tumour. Radiomic features with Laplacian of Gaussian (LoG) and wavelet transforms were extracted. Feature selection was performed using Mann-Whitney U tests and least absolute shrinkage and selection operator (LASSO) regression. Selected features were used to build the logistic regression (LR) models for differentiating HGSC and non-HGSC. The performances of the models were assessed by receiver operating characteristic curves (ROC) and the area under the curve (AUC).
Results or Findings: Thirty-four radiomic features were selected for modelling. The AUCs of LR models for differentiating HGSC and non-HGSC were 0.944 for the internal training set, 0.945 for the internal test set and 0.805 for the external test set.
Conclusion: Radiomic features extracted from ceCT were helpful in histological subtyping of EOC. The LR models offered outstanding diagnostic efficiency in the internal cohorts, and excellent in the external cohort.
Limitations: Centre B has a limited sample size. The distribution of HGSC and non-HGSC was different between Centre A and B.
Ethics committee approval: Institutional review board (No. UW 20-251).
Funding for this study: Health and Medical Research Fund (no. 08192106).

Author Block: S. Patra¹, S. B. Grover², H. Grover³, P. Mittal¹, G. Khanna¹; ¹New Delhi/IN, ²Greater Noida/IN, ³New York, NY/US
Purpose or Learning Objective: To compare the diagnostic performance of ORADS-MRI score vis-a-vis ADNEX-MR score for characterisation of ovarian-adnexal tumours as benign or malignant using histopathology as a reference standard. The secondary objective was to justify the transition from ADNEX-MR to ORADS-MRI.
Methods or Background: The ORADS-MRI algorithm was recently proposed for characterising ovarian tumours, directing the management of malignant tumours towards specialised oncology teams and benign tumours towards restricted intervention. This study was designed as a cross-sectional, observational study, with the retrospective application of ORADS-MRI and ADNEX-MR scores to a prospectively acquired data set during an IRB approved study on ovarian-adnexal masses conducted during the last 2 years. We recruited patients with ovarian-adnexal tumours, who had undergone 3T MRI with morphological, functional and dynamic contrast-enhanced sequences and who had final histopathology diagnosed. Major exclusion criteria at recruitment were age below 18 years, deranged renal functions, and metallic implants. Two trained radiologists initially blinded to histopathology diagnosis assessed ORADS-MRI and ADNEX-MR scores in each tumour. The final score and the diagnosis were assigned by consensus, ORADS 1-3 and ADNEX 1-3 as benign, ORADS 4/5 and ADNEX 4/5 as malignant, and subsequently correlated with histopathology.
Results or Findings: The cohort comprised 45 patients with 63 tumours, histopathology diagnosis was 36 benign and 27 malignant tumours. ORADS-MRI vis-a-vis ADNEX-MR had sensitivity 96.3% vis-a-vis 96.3%, specificity 72.2% vis-a-vis 63.9%, PPV 72.2% vis-a-vis 66.7%, NPV 96.3% vis-a-vis 95.8% and diagnostic accuracy 82.5% vis-a-vis 77.78%.
Conclusion: Performance of ORADS-MRI was superior in specificity, PPV and diagnostic accuracy and further, since ORADS-MRI is more definitive for risk assignment in tumours with score 5 (ORADS >90%, ADNEX >50%) the transition from ADNEX-MR score towards ORADS-MRI algorithms is adequately justified.
Limitations: Small sample size, single centre study.
Ethics committee approval: Ethics approval was obtained.
Funding for this study: No funding was received for this study.

Author Block: E. Y. P. Lee, J. A. U. Perucho, G. S. Kwok, K. Y. Tse, P. Ip; Hong Kong/CN
Purpose or Learning Objective: To evaluate the accuracy of PET/CT in monitoring treatment response in epithelial ovarian carcinoma (EOC) after neoadjuvant chemotherapy (NACT).
Methods or Background: Patients with advanced EOC (FIGO III-IV) were prospectively recruited to undergo PET/CT after 3-6 cycles of NACT before interval debulking surgery. PET/CT was evaluated using a 5-point Likert scale. Nineteen regions in the abdominopelvic cavity were evaluated on PET/CT (subdiaphragmatic spaces, perihepatic, liver serosa and hilum, gastric serosa, pancreas and lesser sac, splenic serosa, splenic hilum, paracolic gutters, omentum, bowel mesentery, paraaortic and pelvic lymph nodes, and central pelvis) and correlated with surgical evaluation and histopathological assessment. Histopathological specimens were taken as the gold standard. In the event where resection was not performed and biopsy could not be safely taken (e.g. miliary serosal disease), the intra-operative surgical findings were taken as the standard of reference. A region-based analysis was performed. The diagnostic characteristics of PET/CT were described by accuracy, sensitivity, specificity, positive (PPV) and negative predictive values (NPV).
Results or Findings: The average accuracy of region-based analysis was 0.88 (sensitivity 0.38, specificity 0.96, PPV 0.56 and NPV 0.91) based on 285 regions evaluated in 15 patients with EOC. Sensitivity in disease detection was low in all regions evaluated, but PET/CT showed high specificity and NPV, except for disease in the omentum (specificity 0.71, NPV 0.56) and central pelvis (specificity 0.71, NPV 0.63).
Conclusion: PET/CT was highly specific with a high NPV in evaluating the abdominopelvic cavity of advanced EOC following NACT. These metrics could be useful in surgical planning in stratifying patients for less aggressive and extensive debulking surgery when the disease was deemed to be absent on PET/CT.
Limitations: The study was limited by small a sample size.
Ethics committee approval: Institutional review board (No. UW 18-604).
Funding for this study: Health and Medical Research Fund (No. 06171706).

Author Block: P. N. Franco¹, S. Annibali², C. Cazzella¹, S. Viganò¹, A. Bonanomi¹, P. A. Bonaffini¹, S. Sironi¹; ¹Milan/IT, ²Bergamo/IT
Purpose or Learning Objective: To evaluate the diagnostic performance of T2*-weighted sequence compared to conventional MRI protocol, used as a reference standard, in the assessment of deep endometriosis.
Methods or Background: Patients who underwent a pelvic MRI on a 3T scan for clinical and/or ultrasound suspicion of deep endometriosis were prospectively enrolled (December 2020 - August 2021). Two radiologists (with 10 and 2 years of experience, respectively) qualitatively evaluated in consensus standard MRI sequences to detect endometriotic lesions. The most experienced radiologist assessed the presence of signal voids on T2*-weighted sequences: in case of discrepancy between standard and T2* sequences, MRI images were re-evaluated in consensus. Equivocal lesions were assessed through second-look ultrasound and clinical evaluation.
Results or Findings: Forty patients (mean age 35.5 years) with a total of 78 endometriosis foci were included in the study. 24,4% (19/78) of endometriotic lesions detected through morphologic standard sequences showed a signal loss in T2*-sequences. At radiologic re-evaluation, 13/19 lesions associated with signal voids (68.4%) have been related to artefacts (including air or previous surgery) or hemorrhagic foci not linked to endometriosis (i.e., hemorrhagic corpus luteum). Among lesions whose nature on T2* sequences was still unclear but not artefact-related (31.6%; 6/19), only 4 resulted consistent with endometriosis at the ultrasound and clinical follow-up and none was an isolated lesion.
Conclusion: T2*-weighted sequences, even if can allow the detection of hemosiderin deposits in endometriotic foci, do not seem to provide an added value in the assessment of deep lesions. Furthermore, artefacts caused by undesirable sources of magnetic signal voids may lead to diagnostic overestimation.
Limitations: The limited sample size. Lack of surgical correlation.
Ethics committee approval: Not applicable.
Funding for this study: No funding was provided for this study.

Author Block: K. H. Härmä, S. Imboden, F. Siegenthaler, M. Mueller, J. T. Heverhagen; Bern/CH
Purpose or Learning Objective: Diaphragmatic endometriosis on MRI.
Methods or Background: In this retrospective single-centre study from 2017 to 2021, nineteen patients with detected diaphragmatic endometriosis (DE) on MRI were consecutively included. Among them (1) patients without DE symptoms referred for pelvic endometriosis MRI (peMRI), including upper abdomen chemical shift sequence (uaCSS) and (2) symptomatic patients suspicious for DE referred for diaphragm endometriosis MRI (diMRI). The image analysis was followed-up by a gynecologic radiologist. Location, number, size, laterality and accuracy of DE findings, as clinical symptoms and correlation with pelvic/bowel deep infiltrating endometriosis (DIE), were investigated. The MR-graphic findings were intra-operative and histopathologically proven. Descriptive statistics, means, medians and ranges were calculated for continuous variables; frequencies for categorical variables. Confusion matrix/PPV/NPV (Excel 2016).
Results or Findings: MRI detected 49 DE suspicious lesions in 19 patients. 47% (9/19) of the patients were referred for diMRI, being suspicious for DE. 53% (10/19) patients were referred for peMRI with uaCSS without any suspicion on DE but pelvic endometriosis. 57% of DE lesions were located posteriorly, 33% laterally, and 10% anteriorly. The mean size of all DE lesions was 13.9 mm (range 3mm-38mm). Until now, nine patients (9/19) underwent a laparoscopic operation resulting in a PPV of 75% for DE detection on MRI. 5/9 operated patients were referred to pelvic MRI. DE was proven in 6/9 patients intra-operatively and histopathologically. False-positive findings in two patients were identified as a diaphragmatic post-operative scar. Both DE and pelvic or bowel DIE was diagnosed in 58%/53% of the women.
Conclusion: The current study demonstrated the importance of incorporating diaphragmatic imaging to pelvic endometriosis MRI. It allows the DE diagnosis in an early stage, offers optimal patient counselling, management of the surgery and anticipating its complexity.
Limitations: Small patient population.
Ethics committee approval: The ethics committee approvals was obtained.
Funding for this study: No funding was received for this study.

Author Block: A. Solopova, Y. Nosova, G. Khabas; Moscow/RU
Purpose or Learning Objective: To explore the correlation between quantitative parameters from DCE‐MRI and ADC with Ki‐67 proliferation status, p53 and p16 expression in patients with ovarian tumours.
Methods or Background: This prospective study, approved by the local ethical committee, enrolled 102 patients (26 benign (BOT), 24 borderline (BEOT) and 52 malignant epithelial ovarian tumours (MEOT)) from 2018-2020. Preoperative MRI was performed in all patients, DCE-MRI quantitative parameters (Ktrans, Kep, Ve) and ADC (mean) were measured, compared, and correlated with Ki-67, p53, p16 expression between the groups of BOT, BEOT, MEOT (according to the histologic verification). Statistical analyses were performed using the MannWhitney U-test, ROС-curves, and Spearman’s correlation by drawing ROIs.
Results or Findings: The Ktrans, Kep and Ve was significantly higher in MEOTs than in BOTs (0,39 (0,36-0,50), 0,25 (0,230-0,270) и 0,22 (0,210-0,240) vs 0,2 (0,16-0,22), 0,016 (0,016-0,051) and 0,1 (0,060-0,130), respectively). ADC values were significantly higher in BOTs than in MEOTs (1,277 (1,2-1,4) х10-3 vs 0,826 (0,7-0,9) х10-3) (p<0,0001). The Ktrans and ADC (mean) value were the most informative parameters for the differentiation between BOTs and MEOTs, providing the sensitivity, specificity, and accuracy of 88,9%, 94,7% and 90,6% for Ktrans and 85,1%, 87,5%, and 85,7% for ADC (mean), respectively; the combination of Ktrans and ADC mean value provides 91,1%, 94,7% and 92,2%, respectively. Ktrans and Kep were positively correlated with Ki-67 expression (r=0,6 and 0,671, respectively; p<0,01), ADC values were correlated with Ki-67 and p16 expression (r=0,624 and 0,6, respectively; p<0,01).
Conclusion: Preliminary findings demonstrate that DCE MRI and ADC are an effective additional tool for differentiating BOTs from MEOTs, and are correlated with Ki-67 and p16 expression.
Limitations: This is a single-centre study.
Ethics committee approval: All studies were approved by the local ethics committee.
Funding for this study: Grant of the President of Russian Federation (MD-130.2019.7).