Abstract
Background/Aim: Magnetic resonance (MR) and ultrasound (US) fusion imaging (MR-US fusion) is already used to guide prostate biopsies and has been proven accurate for diagnosing cervical cancer. In this study, we aimed to evaluate the feasibility and performance of MR-US fusion for characterizing adnexal masses. Patients and Methods: A retrospective study was conducted between 2014 and 2018 including women referred to our Gynaecological Oncology Department for characterization of an adnexal mass (n=106). Performance of MR-US fusion was evaluated in a subgroup of patients who underwent surgery (n=26). Two readers, blinded to final histology, performed and rated US findings according to the International Ovarian Tumor Analysis simple rules score, MR according to Ovarian-Adnexal Reporting Data System Magnetic score, and MR-US fusion through a tailored score. The reference outcome was the final pathology. Results: MR-US fusion had a sensitivity of 100% (95%CI=80-100), specificity of 89% (95%CI=52-99), positive likelihood ratio of 9 (95%CI=1.4-57), and accuracy of 96% (95%CI=80-99). Conclusion: MR-US fusion is feasible for characterizing adnexal masses to predict ovarian cancer.
- Magnetic resonance
- ultrasound
- MR-US fusion
- IOTA simple rules
- O-RADS MRI scoring system
- adnexal masses
- ovarian cancer
Ovarian cancer is the deadliest gynaecological cancer with a 5-year survival rate of 46% (1). This poor prognosis is largely due to the lack of specific clinical symptoms meaning that it is commonly diagnosed at an advanced stage (1). Early diagnosis and specialized care in high-volume Institutions with trained expert surgeons is known to improve the survival rate (2), therefore characterising an adnexal mass as benign, borderline or malignant is crucial to decide on optimal management (3, 4). According to international guidelines (5, 6), ultrasonography (US) is the first-line technique for characterising an adnexal mass with an accuracy between 65-94% (5) depending on the expertise of the ultrasonographer (7, 8). In this setting, the “simple rules score” developed by the International Ovarian Tumor Analysis (IOTA) group (9) may help the less expert reader and has been proven to be the best performing and reproducible method among the various US scoring systems for describing adnexal masses (10). Nevertheless, between 20-25% of adnexal masses remain indeterminate and require a second-line technique. In this respect, magnetic resonance imaging (MR) is the most accurate technique for distinguishing malignant from benign complex adnexal masses with values ranging from 83% to 93% (11, 12). An MR scoring system has been developed to improve communication between the radiologist and the gynaecologist (13). External validation of the score confirmed the accuracy of MR at 90% (14, 15) and it has recently been prospectively validated as the Ovarian-Adnexal Reporting Data System Magnetic (O-RADS) MRI score (16). However, some limitations were reported due to a low spatial resolution of MR imaging to characterize small solid tissue components.
The MR and US fusion imaging (MR-US fusion) technique consists of livestream merging of US and MR images. It has already been used to guide prostate biopsies (17), and more recently to improve the diagnosis of cervical cancer (18). The first objective of this study was to test the feasibility of the MR-US fusion in imaging ovarian suspicious masses. Secondary objectives were to estimate fusion performance to predict malignancy when compared to final histology and confront those findings with ultrasound and MRI.
Patients and Methods
Patients. We conducted a retrospective, single-centre study in the Gynaecological Oncology Unit of Creteil Hospital [Centre Hospitalier Intercommunal de Creteil (CHIC)], France. The Institutional Ethics Committee approved the study as the patients were included in the PELVIMASS cohort (n°ID RCB: 2016-A01381-42) and had provided written informed consent allowing us to anonymously and prospectively record all imaging and medical data. The PELVIMASS cohort consisted of adult women referred to the centre for a pelvic mass from January 2014 to July 2018. We recorded the age at diagnosis, body mass index (BMI), menopausal status, hormonal treatment, and parity. A pelvic US, using the IOTA simple rules scoring system, is routinely performed in our centre for all patients referred for a suspected pelvic mass. If the patient had undergone a pelvic MRI in another centre, the images were re-assessed to determine the O-RADS MRI score. If the patient was referred without an MRI, and if indicated, the assessment was completed with a pelvic MRI according to the O-RADS MRI protocol in our centre. Finally, an MR-US fusion was performed. The feasibility of MR-US fusion was evaluated in the entire population. The performance of MR-US fusion compared to US and MR was evaluated in the subgroup of patients who underwent surgery. The reference standard was the final histology report. Borderline tumours were considered malignant.
Imaging techniques. The US findings were rated by a sonographer (JML) using the IOTA simple rules scoring system (scale ranged from 1: very low risk, to 5: very high risk) (19). MRI was rated by a senior radiologist (ITN) (with more than 15 years of experience) using the O-RADS MRI score based on T2 weighted sequence, T1 weighted sequence with and without fat saturation, diffusion weighted sequence and dynamic contrast enhanced sequences (16, 20). Time-intensity curves were performed if a solid tissue was detected. Both the raters were blinded to final histology.
MR-US fusion acquisition was then performed by two experts (JML and CT) before surgery. Three MR sequences – T2W axial and sagittal and T1W axial with gadolinium – were imported. More sequences were added when requested by the radiologist. The patient was placed on her back and a magnetic field generator was set close to the area of interest. A sensor was fixed to the probe so it could be spotted in the magnetic field. The probes were the same as in a classic US: 4-8 MHz for an abdominal probe and 5-9 MHz for a transvaginal probe. The MRI sequences and US were aligned by selecting 2 or 3 points of interest. For an abdominal MR-US fusion procedure the points are the anterior edge of the L5-S1 discus, the pubal symphysis and the cervix. For a vaginal MR-US fusion the points are the anterior edge of the L5-S1 discus or the cervix so that the movements of the probe are replicated on the MR imaging. In some cases, readjustment was necessary by selecting a point on the solid tumour tissue if present. The examination systematically began with an abdominal US, then an abdominal MR-US fusion if possible, and finally a vaginal MR-US fusion was performed. Systematic Doppler colour acquisition was performed on solid tissue. All the examinations were recorded to be subsequently consensually reviewed by the two experts (JML and ITN). The IOTA items were applied in addition to MR features: for example, the presence of peritoneal implants on MR imaging corresponded to a score 5.
Patient characteristics.
Statistical analysis. Reference outcome for the assessment of adnexal masses was final histology. Descriptive analysis was performed using a non-parametric Mann–Whitney test for continuous variables and Fisher's exact test for categorical or nominal variables. Scores of 4 or 5 were considered positive and scores of 1 or 2 or 3, negative. We calculated the sensitivity (Se), specificity (Sp), positive and negative predictive Value (PPV and NPV), positive and negative likelihood ratio (PLR and NLR) for each technique using the Excel Software (Microsoft, Dublin, Ireland). All tests were two-sided. A p-value of less than 0.05 was considered to indicate a statistically significant difference. Statistical analyses were performed using the MedCalc software (www.medcalc.be, Belgium, medcalc version 9.3.0.0).
Results
Feasibility of the study. Within the study period, 106 patients were referred to our centre for MR-US fusion for a suspicious or undetermined mass on US examination. MR-US fusion was technically feasible in 103 patients (97%). The three remaining patients could not have MR-US fusion for the following reasons: severe obesity (BMI 53 kg.m−2); orthopaedic issues; bedridden patient. The mean age of the 103 patients who underwent MR-US fusion was 51 years and the median BMI was 23.2 kg.m−2 (range=16-38). Fifty-one of the patients (49.5%) were menopausal (Table I). Mean (±SD) duration of examination was 43 (±22) min.
Study flow chart. IOTA, International Ovarian Tumor Analysis; US, ultrasound; O-RADS, Ovarian-Adnexal Reporting Data System; MRI, magnetic resonance imaging.
Performance analysis. Seventy-seven patients were excluded for the performance analysis: 33 (41.5%) did not undergo surgery (32 because of absence of indication and one who was ineligible due to contraindications for anaesthesia), 32 (41.5%) had undergone MRI outside of our centre and did not have sufficient sequences to evaluate the O-RADS MRI score; a total of 12 (15.6%) did not respect the protocol - 8 of them because the data could not allow a second lecture by our expert, 4 patients were lost to follow-up (Figure 1). The mean (±SD) age of the 26 patients in this group was 57 (±18) years and the median BMI was 22(±6) kg/m2. Seventeen (65%) patients were menopausal and five (19%) received hormonal replacement therapy. They underwent MR-US fusion in a mean (±SD) time of 15 (±6.7) days after MR imaging. The mean (min-max) time between MR-US fusion and surgery was 19.5 days (4-119).
Histology results revealed 14 (54%) cancer cases, three (12%) borderline tumours and nine (35%) benign tumours. Among the cancer cases, we found seven (50%) serous carcinomas, one (7%) undifferentiated, one intraepithelial carcinoma developed within a borderline tumour, one mucinous, one clear-cell, one granulosa, two (14%) ovarian metastases: one colon adenocarcinoma, and one neuroendocrine tumour. Three (11.5%) patients presented a serous borderline tumour. Eight patients presented benign lesions (Table II). Using the IOTA simple rules score, US had a sensitivity of 94% (95%CI=73-99), a specificity of 44% (95%CI=14-79), and a PLR of 1.7 (95%CI=0.9-3.1) with an accuracy of 77% (95CI=56-91). O-RADS MRI scoring system had a sensitivity of 94% (95%CI=71-99), a specificity of 89% (95%CI=52-99), a PLR of 8.5 (95%CI=1.3-54) and an accuracy of 92% (95%CI=75-99). Compared to histology, MR-US fusion had a sensitivity of 100% (95%CI=80-100), a specificity of 89% (95%CI=52-99) with a PLR of 9 (95%CI=1.4-57) and an accuracy of 96% (95%CI=80-99) (Table III).
Histological subtypes in the subgroup analysis (N=26).
MR-US fusion (with energy doppler) appearance of a right borderline ovarian tumour.
Sensitivity, specificity, positive likelihood ratios and accuracy of ultrasound, MRI and MR-US FUSIONFUSION compared to histology (performance analysis) (N=26).
Detailed results. One mass was correctly reclassified as malignant by MR-US fusion. The 51-year-old woman had a history of a pancreas neuroendocrine tumour and was considered to be in clinical and radiological remission. A score of 3 was allocated on US because of an absence of malignant features and one benign feature that was a smooth multilocular tumour of less than 10 cm. On MR, this ovarian mass was detected as a 5 cm multilocular cyst without solid tissue, and was thus rated with a score of 3. MR-US fusion finally revealed peritoneal implants and effusion in the pouch of Douglas. The rapid worsening of her clinical condition indicated explorative laparoscopy, which revealed recurrence of ovarian and peritoneal neuroendocrine tumours.
Three patients (12%) presented borderline lesions on final histology. Two of them had been classified as score 4 on US, and one as score 5. On MRI, they were all classified as score 4 and on MR-US fusion, they were all classified as score 5. One patient was reclassified as score 5 due to better visualization of solid vegetations within the cyst (Figure 2 and Movie clip 1).
Four masses were classified as suspicious or malignant by US and reclassified as benign by MRI: one myoma of 60 mm of the broad ligament, one endometrioma of 140 mm, one fibrothecoma of 80 mm and one mucinous cystadenoma. One mass was misclassified as malignant by MR-US fusion: an 80 mm fibrothecoma described as suspicious on US but reclassified as benign on MR as a solid tissue with TIC type 1 and a score 3.
Discussion
To the best of our knowledge this is the first study to evaluate MR-US fusion technology for the characterization of adnexal masses. We demonstrated that MR-US fusion is feasible in this setting, even among obese women, as we were able to explore 103 adnexal masses of the 106 patients referred to our centre for suspicious or undetermined mass after US examination. In addition, in a performance analysis on 26 patients, MR-US fusion was accurate in assessing pelvic masses with a sensitivity of 100% (95%CI=80-100), a specificity of 89% (95%CI=52-99), PLR of 9 (95%CI=1.4-57) and an accuracy of 96% (95%CI=80-99) when compared to final histology. The patients underwent the MR-US fusion in a mean (SD) time of 15 (6.7) days after MRI. Moreover, MR-US fusion was able to correctly reclassify as malignant an ovarian metastasis of a neuroendocrine tumour misdiagnosed on both US and MRI.
The MR-US fusion is an accessible tool that does not delay surgical management. Moreover, it is perfectly accepted by patients and we have had no refusal to perform this examination. Fusion MR-US is subject to the same constraints as a gynaecological ultrasound, and these are essentially mechanical. One morbidly obese patient was unable to climb onto the examination table, one patient with a hip prosthesis was unable to settle in the gynaecological position, while the third was bedridden. In our centre, we use MR-US fusion as an educational tool to help recognise anatomical structures and target flagged lesions which optimizes the learning curves. Multidisciplinary meetings, merging the experience of gynaecologists and radiologists, improves interpretation of the MR-US fusion loops and encourages dialogue. As for the limitations of the technique, these mainly lie in the additional equipment and the training required to use it. It is obvious that this technique remains a technological innovation that is still being developed and perfected and was therefore carried out in our centre by experts with experience in gynaecological ultrasound. In addition to the ultrasound machines used daily in the department, the engineers provided us with a magnetic field generator as well as sensors adapted to each probe and installed fusion software. Engineers from each company (i.e. HITACHI and General Electrics) spent 4 days in the department to train the teams. Another limitation of the use is the difficulty of synchronizing the volumes. This problem may have been encountered when the time between the MRI and the date of the examination was too far apart. We advise users to define the constraints in accordance with the conditions for performing the MR (i.e. empty or full bladder at MRI) and if possible, to perform the MR-US fusion on the same day as the MR.
The feasibility of MR-US fusion has already been reported in the gynaecological field. In endometriosis, MR-US fusion can be used to identify deep pelvic endometriotic lesions, with better performance than MRI and US alone, because it improves visualization of the main pelvic anatomical sites of endometriotic lesions (21). Moreover, our team previously assessed the use of MR-US fusion in cervical cancers to improve the diagnosis of extension to the vagina, bladder or parametrium (18). In the current study we propose the use of MR-US fusion in the exploration of indeterminate masses of the ovary. Currently, all undetermined pelvic masses after US and MR, are considered for surgery (4). Thus, the objective of a pre-surgical third-line examination must be to avoid unnecessary surgery in patients with benign masses. The examination, therefore, requires a good NPV. For MR-US fusion the NPV is 100%. This implies that for a patient presenting an undetermined mass with the ORADS score and an MR-US fusion score ruling out malignancy, it would be legitimate to recommend monitoring alone, the modalities of which remain to be defined.
Limitations that could be raised about this technique concern the expertise required to interpret the images and loops of MR-US fusion. The technology requires a senior in sonography with training in MRI or, ideally, a team specialised in pelvic mass imaging. We, therefore, believe that patients should be referred to expert centres for MR-US fusion.
Some limitations to this study deserve to be mentioned. We only had histological correlation for part of our cohort, mostly because there was no indication of surgery, but also because a lot of the MR imaging performed outside of our reference centre did not contain sufficient sequences to evaluate the O-RADS MRI score. By the time we conducted this study, the use of O-RADS score had become the standard for risk stratification in adnexal masses on MRI (4, 16). Therefore, the power of our analysis of diagnostic performance was limited, and the calculation of performance on such a small cohort is debatable. However, this study was designed to evaluate the feasibility of MR-US fusion and further work is required to study performance and determine the place of MR-US fusion in the care pathway.
In conclusion, we demonstrated the feasibility of MR-US fusion in the characterization of adnexal masses. The study supports the use of MR-US fusion as a promising multidisciplinary imaging technique in onco-gynaecology. More studies are required to determine if MR-US fusion is more than just a simple addition of two imaging techniques, and define its indications.
Footnotes
Authors' Contributions
Adrien Crestani, Claire Theodore, Jean-Marc Levaillant, Isabelle Thomassin-Naggara: Manuscript composition. Adrien Crestani: Statistical analysis.
Isabelle Thomassin-Naggara, Dounia Skalli, Grégoire Miaihle, Yohann Dabi, Bassam Haddad, Cyril Touboul: Manuscript corrections.
Conflicts of Interest
The Authors declare no conflicts of interest with regard to the present study.
Supporting Material
Movie clip 1. MR-US Fusion appearance of a borderline ovarian tumour.
Available at: https://www.youtube.com/watch?v=M8vudlF5f3E&t
- Received May 13, 2020.
- Revision received June 1, 2020.
- Accepted June 10, 2020.
- Copyright© 2020, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.