[MR Imaging of primary ovarian tumors with pathologic correlation]

MRI-Based Multiple Instance Convolutional Neural Network for Increased Accuracy in the Differentiation of Borderline and Malignant Epithelial Ovarian Tumors

BACKGROUND

Preoperative differentiation of borderline from malignant epithelial ovarian tumors (BEOT vs. MEOT) is challenging and can significantly impact surgical management.

PURPOSE

To develop a multiple instance convolutional neural network (MICNN) that can differentiate BEOT from MEOT, and to compare its diagnostic performance with that of radiologists.

STUDY TYPE

Retrospective study of eight clinical centers.

SUBJECTS

Between January 2010 and June 2018, a total of 501 women (mean age, 48.93 ± 14.05 years) with histopathologically confirmed BEOT (N = 165) or MEOT (N = 336) were divided into the training (N = 342) and validation cohorts (N = 159).

FIELD STRENGTH/SEQUENCE

Three axial sequences from 1.5 or 3 T scanner were used: fast spin echo T2-weighted imaging with fat saturation (T2WI FS), echo planar diffusion-weighted imaging, and 2D volumetric interpolated breath-hold examination of contrast-enhanced T1-weighted imaging (CE-T1WI) with FS.

ASSESSMENT

Three monoparametric MICNN models were built based on T2WI FS, apparent diffusion coefficient map, and CE-T1WI. Based on these monoparametric models, we constructed an early multiparametric (EMP) model and a late multiparametric (LMP) model using early and late information fusion methods, respectively. The diagnostic performance of the models was evaluated using the receiver operating characteristic (ROC) curve and compared to the performance of six radiologists with varying levels of experience.

STATISTICAL TESTS

We used DeLong test, chi-square test, Mann-Whitney U-test, and t-test, with significance level of 0.05.

RESULTS

Both EMP and LMP models differentiated BEOT from MEOT, with an area under the ROC curve (AUC) of 0.855 (95% CI, 0.795-0.915) and 0.884 (95% CI, 0.831-0.938), respectively. The AUC of the LMP model was significantly higher than the radiologists' pooled AUC (0.884 vs. 0.797).

DATA CONCLUSION

The developed MICNN models can effectively differentiate BEOT from MEOT and the diagnostic performances (AUCs) were more superior than that of the radiologists' assessments.

LEVEL OF EVIDENCE

3 TECHNICAL EFFICACY STAGE: 2.

Computer Tomography in the Diagnosis of Ovarian Cysts: The Role of Fluid Attenuation Values

Pathological analysis of ovarian cysts shows specific fluid characteristics that cannot be standardly evaluated on computer tomography (CT) examinations. This study aimed to assess the ovarian cysts’ fluid attenuation values on the native (Np), arterial (Ap), and venous (Vp) contrast phases of seventy patients with ovarian cysts who underwent CT examinations and were retrospectively included in this study. Patients were divided according to their final diagnosis into the benign group (n = 32) and malignant group (n = 38; of which 27 were primary and 11 were secondary lesions). Two radiologists measured the fluid attenuation values on each contrast phase, and the average values were used to discriminate between benign and malignant groups and primary tumors and metastases via univariate, multivariate, multiple regression, and receiver operating characteristics analyses. The Ap densities (p = 0.0002) were independently associated with malignant cysts. Based on the densities measured on all three phases, neoplastic lesions could be diagnosed with 89.47% sensitivity and 62.5% specificity. The Np densities (p = 0.0005) were able to identify metastases with 90.91% sensitivity and 70.37% specificity, while the combined densities of all three phases diagnosed secondary lesions with 72.73% sensitivity and 92.59% specificity. The ovarian cysts’ fluid densities could function as an adjuvant criterion to the classic CT evaluation of ovarian cysts.

Ovarian serous surface papillary borderline tumor: characteristic imaging features with clinicopathological correlation

OBJECTIVE

To retrospectively evaluate the clinical, pathological, and imaging features of ovarian serous surface papillary borderline tumors (SSPBTs).

METHODS

Imaging features were analyzed, including the tumor size, laterality, tumor spread and the presence of ascites. Morphological nature (encased normal ovary, characteristics on MRI) and contrast enhancement (increased flow on Doppler ultrasound) were also evaluated. Clinical and pathological features, such as tumor markers (CA 125), treatment methods, outcomes on follow-up, and surgical staging, were analyzed.

RESULTS

10 tumors in 5 patients were evaluated. Mean largest tumor diameter was 45 mm. All patients had bilateral involvement and ascites. 40% of all patients showed peritoneal implants. 20% of patients evaluated had lymph node metastasis. These patients showed grossly normal ovaries that were encased in or surrounded by irregular solid tumors. They had a mostly hyperintense papillary architecture with hypointense internal branching on T₂ weighted MR images (90%). Contrast enhancement and serum levels of CA 125 were elevated in all cases evaluated. All patients underwent radical surgery, and 80% of patients evaluated were of low stage. No recurrence occurred, during follow-up.

CONCLUSION

SSPBT of the ovary, which has a good prognosis, should be considered as a diagnosis for patients who have bilateral enhancing irregular solid masses with papillary architecture and internal branching, and encasing normal-appearing ovaries. Advances in knowledge: Serous surface papillary borderline tumor of the ovary is unique and has characteristic features. Knowledge of this specific ovarian tumor and radiological suspicion can have important implications for the patient to facilitate management including fertility-preserving surgery.

MRI features and score for differentiating borderline from malignant epithelial ovarian tumors

PURPOSE

To identify the MRI features of borderline epithelial ovarian tumors (BEOTs) and to differentiate BEOTs from malignant epithelial ovarian tumors (MEOTs).

MATERIALS AND METHODS

The clinical and MRI data of 89 patients with a BEOT and 109 patients with a MEOT proven by surgery and histopathology were retrospectively reviewed. MRI features, including bilaterality, size, shape, margin, cystic-solid interface, configuration, papillae or nodules, signal intensity, enhancement, presence of an ipsilateral ovary, peritoneal implants and ascites were analyzed and compared. Based on the odds ratio (OR) values, the significant risk features for BEOTs were scored as 3 (OR≈∞), 2 (5≤OR<∞) or 1 (OR<5).

RESULTS

There were 89 BEOT patients with 113 tumors [mean size of (13±6.7)cm], with bilateral ovary involvement in 24 cases. There were 109 MEOT patients with 142 tumors [(9.3±4.2)cm] with bilateral ovary involvement in 33 cases. There were eight significant risk factors for BEOTs, including round or oval shape (OR=2.714), well-defined margins (OR=3.318), clear cystic-solid interfaces (OR=5.593), purely cystic (OR=15.206), predominantly cystic with papillae or nodules (OR=2.579), exophytic papillae or nodules (OR=5.351), branching papilla (OR≈∞) and the presence of an ipsilateral ovary (OR≈∞). Based on the scoring of the eight risk factors, a cut-off score of 3.5 yielded a differential sensitivity, specificity, and accuracy of 82%, 85% and 84%, respectively.

CONCLUSION

In contrast to MEOTs, BEOTs frequently had the following features on MRI: round or oval, with well-defined margins and clear cystic-solid interfaces, purely cystic or predominantly cystic with papillae or nodules, branching or exophytic papillae, with the presence of an ipsilateral ovary. MRI can reveal the distinct morphological features of BEOTs and MEOTs and facilitate their discrimination.

The MRI features of histologically proven ovarian cystadenofibromas-an assessment of the morphological and enhancement patterns

OBJECTIVES

To assess the morphological and enhancement features of histologically proven cystadenofibromas (CAFs) on magnetic resonance imaging (MRI).

METHODS

Forty-seven histologically proven CAFs (42 benign, five borderline) were retrospectively reviewed. One benign CAF had a synchronous adenocarcinoma in the same ovary. The morphological, signal and enhancement characteristics on MRI were recorded.

RESULTS

The mean long axis diameter of the CAFs was 80 mm. The contralateral ovary was abnormal in 45 % of cases. A solid component was seen in 85 %, which returned low T2-weighted signal in 75 % of CAFs. Septa were seen in 74 % and one CAF was purely cystic. The majority of solid components and septa demonstrated enhancement that was less than the myometrium. Wash-in rates (WIR) of the solid tissue were available for measurement in nine patients with an average WIR of 3.2 l/s.

CONCLUSION

This is the largest series describing MRI appearances of histologically proven CAFs. They are typically complex adnexal lesions containing septa, cystic components and solid tissue. The majority of solid components demonstrate low T2 signal and minimal enhancement. Almost half of the cases have an abnormal contralateral ovary.

Benign and Suspicious Ovarian Masses-MR Imaging Criteria for Characterization: Pictorial Review

Ovarian masses present a special diagnostic challenge when imaging findings cannot be categorized into benign or malignant pathology. Ultrasonography (US), Computed Tomography (CT), and Magnetic Resonance Imaging (MRI) are currently used to evaluate ovarian tumors. US is the first-line imaging investigation for suspected adnexal masses. Color Doppler US helps the diagnosis identifying vascularized components within the mass. CT is commonly performed in preoperative evaluation of a suspected ovarian malignancy, but it exposes patients to radiation. When US findings are nondiagnostic or equivocal, MRI can be a valuable problem solving tool, useful to give also surgical planning information. MRI is well known to provide accurate information about hemorrhage, fat, and collagen. It is able to identify different types of tissue contained in pelvic masses, distinguishing benign from malignant ovarian tumors. The knowledge of clinical syndromes and MRI features of these conditions is crucial in establishing an accurate diagnosis and determining appropriate treatment. The purpose of this paper is to illustrate MRI findings in neoplastic and non-neoplastic ovarian masses, which were assessed into three groups: cystic, solid, and solid/cystic lesions. MRI criteria for the correct diagnosis and characteristics for differentiating benign from malignant conditions are shown in this paper.

MR-PET fusion imaging in evaluating adnexal lesions: a preliminary study

PURPOSE

The objective of this preliminary study was to examine the effects of combined magnetic resonance/positron emission tomography (MR-PET) evaluation in the morphofunctional characterisation of ovarian lesions.

MATERIALS AND METHODS

From June 2008 to September 2010, we evaluated 24 patients (mean age 44±10 years; range 24-74) with ovarian lesions incidentally detected on ultrasonography (US) and/or multislice computed tomography (CT). All patients underwent MR imaging of the pelvis and total-body CT-PET. PET and MR images were subsequently fused at postprocessing using specific anatomical criteria. Results were compared with the histological examination.

RESULTS

Of the 24 examined lesions, 19 were malignant and five were benign on histological examination. MR, CT-PET and MR-PET sensitivity was 84%, 74% and 94%, respectively and specificity 60%, 80% and 100%, respectively. Positive (PPV) and negative predictive (NPV) values were 93% and 44% for CT-PET, 89% and 50% for MR and 100% and 83% for MR-PET, respectively.

CONCLUSIONS

Pelvic MR-PET fusion imaging provides advantages in terms of sensitivity and especially specificity compared with MR imaging or CT-PET alone. The added value of this fusion imaging modality lies in combining the benefits of the morphological evaluation provided by MR imaging and the metabolic assessment provided by PET.

MRI of the female pelvis: a possible pitfall in the differentiation of haemorrhagic vs. fatty lesions using fat saturated sequences with inversion recovery

The use of fat-saturated techniques should be an integral part of the work-up of any T1-hyperintense structure in the female pelvis for tissue characterization and for differentiation of a fat-containing ovarian mature teratoma from a haemorrhagic lesion. Two cases with haematocolpos and haematometra are presented, respectively. The haemorrhagic content showed high signal both on T1- and T2-weighted images, whereas an unexpected signal decrease in the fat-saturated T2-weighted inversion-recovery sequence was encountered. This unspecific suppression of signal in tissues with similar T1 relaxation times as fat can lead to a diagnostic pitfall both in T1- and T2-weighted STIR pulse sequences. Furthermore, a loss of signal on T2-weighting may also be due to the phenomenon of "T2-shading" in T1-bright ovarian endometrioma. Therefore, the fat-specific spectral fat-saturation of T1-weighted images is strongly recommended for tissue characterization in gynaecological disease.

[Multiple ovarian fibromas in a patient with Gorlin syndrome: US and MR imaging features with pathological correlation]

We report a case of multiple ovarian fibromas in a 23 year old woman with Gorlin syndrome. We describe the US and MR imaging features with pathological correlation. The fibrous component of the tumors were hypoechoic and attenuating on US with corresponding T2W hypointensity whereas myxoid components were hypoechoic with increased through transmission on US with corresponding T2W hyperintensity.