Differentiation between borderline and benign ovarian tumors: combined analysis of MRI with tumor markers for large cystic masses (≥5 cm)

Imaging of a Pelvic Mass: Uterine

The most common origin of a non-uterine pelvic mass is from the ovary. Ultrasound is the initial imaging modality of choice, additional imaging with computed tomography (CT) and/or magnetic resonance imaging (MRI) is performed in selected cases. Adnexal masses are also encountered as incidental findings during ultrasound, CT or MRI. Many of the adnexal masses that are surgically removed are benign. For optimal outcome and cost effective management, noninvasive risk stratification of such adnexal masses is necessary when discovered incidentally or when identified in a patient with a clinically detected pelvic mass. The American College of Radiology Ovarian-Adnexal Reporting Data System is a pattern-based scoring system for adnexal masses imaged with ultrasound and MRI, which assists clinicians to guide in the appropriate management based on evidence-based risk categories. Non-ovarian and non-uterine pelvic masses include fallopian tube abnormalities, paraovarian cysts, peritoneal inclusion cysts, and rare causes include masses that arise from the gastrointestinal tract or the sacrum. To distinguish non-ovarian masses from an ovarian tumor, a critical step is to identify a normal appearing ovary separate from the pelvic mass. This may be challenging in the post-menopausal woman with an atrophic ovary. MRI is a useful adjunctive modality in such cases. Extraovarian masses typically displace pelvic side wall vasculature medially, compress, encase or medially displace the ureters.

Clinical Value of Combined Multi-Indicator Tests in Diagnosis of Benign Ovarian

Background

To investigate the existence and degree of correlation between benign ovarian tumors and physiological indicators such as reproductive hormones and tumor markers.

Methods

A total of 150 patients with benign ovarian tumors admitted to Jiaxing First Hospital between January 1, 2019, and May 30, 2021, were enrolled as research subjects, while 104 healthy women were enrolled in the control group. Comparative analysis of the correlation between the reproductive hormones LH, FSH, T, E2, and the tumor indicators AMH, AFP, CEA, CA125, and CA199 between the groups was performed.

Results

There was no statistical difference in LH, FSH, T, AMH, and CEA expression levels between the experimental and control groups (p≥0.05); E₂, CA125, and CA199 levels were higher significantly in the experimental group than in the control group (P<0.001); AFP levels were significantly lower in the experimental group than in the control group (P<0.05). CA125 (0.762) had the highest AUC when diagnosing the value of each index of E2, CA125, and CA199 for benign ovarian tumors. CA125 had the highest sensitivity (56.7%), followed by E₂ (50.0%); CA199 had the highest specificity (84.5%), followed by CA125 (83.7%). The combined diagnosis of benign ovarian tumors was performed using different combinations of the indicators. When the two indicators were combined for diagnosis, the combination of E2 + CA199 had the highest sensitivity (82.6%), whereas the combination of CA125 + CA199 had the largest AUC (0.783) and the highest specificity (86.4%). The combined diagnosis of E₂+CA125+CA199 had a higher AUC than the combined diagnosis of the two indicators (0.805), with a sensitivity of 77.2%, and a specificity of 70.9%.

Conclusion

The most relevant factors for benign ovarian tumors are E2, CA125, and CA199 and the combination of these three indicators has the highest AUC for disease prediction while increasing the detection rate of benign ovarian tumors.

T2-weighted MRI-based radiomics for discriminating between benign and borderline epithelial ovarian tumors: a multicenter study

Background

Preoperative differentiation between benign and borderline epithelial ovarian tumors (EOTs) is challenging and can significantly impact clinical decision making. The purpose was to investigate whether radiomics based on T2-weighted MRI can discriminate between benign and borderline EOTs preoperatively.

Methods

A total of 417 patients (309, 78, and 30 samples in the training and internal and external validation sets) with pathologically proven benign and borderline EOTs were included in this multicenter study. In total, 1130 radiomics features were extracted from manually delineated tumor volumes of interest on images. The following three different models were constructed and evaluated: radiomics features only (radiomics model); clinical and radiological characteristics only (clinic-radiological model); and a combination of them all (combined model). The diagnostic performances of models were assessed using receiver operating characteristic (ROC) analysis, and area under the ROC curves (AUCs) were compared using the DeLong test.

Results

The best machine learning algorithm to distinguish borderline from benign EOTs was the logistic regression. The combined model achieved the best performance in discriminating between benign and borderline EOTs, with an AUC of 0.86 ± 0.07. The radiomics model showed a moderate AUC of 0.82 ± 0.07, outperforming the clinic-radiological model (AUC of 0.79 ± 0.06). In the external validation set, the combined model performed significantly better than the clinic-radiological model (AUCs of 0.86 vs. 0.63, p = 0.021 [DeLong test]).

Conclusions

Radiomics, based on T2-weighted MRI, can provide critical diagnostic information for discriminating between benign and borderline EOTs, thus having the potential to aid personalized treatment options.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13244-022-01264-x.

Diagnosing Ovarian Cancer on MRI: A Preliminary Study Comparing Deep Learning and Radiologist Assessments

Simple Summary

As a preliminary experiment to explore the possibility of clinical application as a future reading assist, we present CNNs for the diagnosis of ovarian carcinomas and borderline tumors on MRI, including T2WI, DWI, ADC map, and CE-T1WI, and compare their diagnostic performance with interpretations by experienced radiologists. CNNs were trained using 1798 images from 146 patients and 1865 images from 219 patients with malignant tumors, including borderline tumors, and non-malignant lesions, respectively, for each MRI sequence and tested with 48 and 52 images of patients with malignant and non-malignant lesions. The CNN of each sequence had a sensitivity of 0.77–0.85, specificity of 0.77–0.92, accuracy of 0.81–0.87, and an AUC of 0.83–0.89, demonstrating diagnostic performances that were non-inferior to those of experienced radiologists, and the CNN showed the highest diagnostic performance on the ADC map for each sequence (specificity = 0.85; sensitivity = 0.77; accuracy = 0.81; AUC = 0.89).

Abstract

Background: This study aimed to compare deep learning with radiologists’ assessments for diagnosing ovarian carcinoma using MRI. Methods: This retrospective study included 194 patients with pathologically confirmed ovarian carcinomas or borderline tumors and 271 patients with non-malignant lesions who underwent MRI between January 2015 and December 2020. T2WI, DWI, ADC map, and fat-saturated contrast-enhanced T1WI were used for the analysis. A deep learning model based on a convolutional neural network (CNN) was trained using 1798 images from 146 patients with malignant tumors and 1865 images from 219 patients with non-malignant lesions for each sequence, and we tested with 48 and 52 images of patients with malignant and non-malignant lesions, respectively. The sensitivity, specificity, accuracy, and AUC were compared between the CNN and interpretations of three experienced radiologists. Results: The CNN of each sequence had a sensitivity of 0.77–0.85, specificity of 0.77–0.92, accuracy of 0.81–0.87, and an AUC of 0.83–0.89, and it achieved a diagnostic performance equivalent to the radiologists. The CNN showed the highest diagnostic performance on the ADC map among all sequences (specificity = 0.85; sensitivity = 0.77; accuracy = 0.81; AUC = 0.89). Conclusion: The CNNs provided a diagnostic performance that was non-inferior to the radiologists for diagnosing ovarian carcinomas on MRI.

Borderline epithelial ovarian tumors: what the radiologist should know

Ovarian borderline tumors are neoplasms of epithelial origin that are typically present in young patients and tend to have a less aggressive clinical course than malignant tumors. Accurate diagnosis and staging of borderline tumors has important prognostic and management implications (like fertility-sparing procedures) for women of child-bearing age. This article will review the sonographic, CT, and MRI features of borderline epithelial ovarian tumors with histopathologic correlation. Borderline tumors have less soft tissue and thinner walls/septations than malignant tumors. Serous borderline tumors more commonly have papillary projections, which can simulate the appearance of a sea anemone. Mucinous borderline tumors often are larger, multi-cystic, and more commonly unilateral. The borderline mucinous tumors may also present with pseudomyxoma peritonei, which can make it difficult to distinguish from malignant mucinous carcinoma. Ultrasound is usually the first-line modality for imaging these tumors with MRI reserved for further characterizing indeterminate cases. CT is best used to stage tumors for both locoregional and distant metastatic disease. Overall, however, the imaging features overlap with both benign and malignant ovarian tumors. Despite this, it is important for the radiologist to be familiar with the imaging appearances of borderline tumors because they can present in younger patients and may benefit from different clinical/surgical management.

MDCT-Based Radiomics Features for the Differentiation of Serous Borderline Ovarian Tumors and Serous Malignant Ovarian Tumors

Objective

To investigate whether multidetector computed tomography (MDCT)-based radiomics features can discriminate between serous borderline ovarian tumors (SBOTs) and serous malignant ovarian tumors (SMOTs).

Patients and Methods

Eighty patients with SBOTs and 102 patients with SMOTs, confirmed by pathology (training set: n = 127; validation set: n = 55) from December 2017 to June 2020, were enrolled in this study. The interclass correlation coefficient (ICC) and least absolute shrinkage and selection operator (LASSO) regression were applied to select radiomics parameters derived from MDCT images on the arterial phase (AP), venous phase (VP), and equilibrium phase (EP). Receiver operating characteristic (ROC) analysis of each selected parameter was carried out. Heat maps were created to illustrate the distribution of the radiomics parameters. Three models incorporating selected radiomics parameters generated by support vector machine (SVM) classifiers in each phase were analyzed by ROC and compared using the DeLong test.

Results

The most predictive features selected by ICC and LASSO regression between SBOTs and SMOTs included 9 radiomics parameters on AP, VP, and EP each. Three models on AP, VP, and EP incorporating the selected features generated by SVM classifiers produced AUCs of 0.80 (accuracy, 0.75; sensitivity, 0.74; specificity, 0.75), 0.86 (accuracy, 0.78; sensitivity, 0.80; specificity, 0.75), and 0.73 (accuracy, 0.69; sensitivity, 0.71; specificity, 0.67), respectively. There were no significant differences in the AUCs among the three models (AP vs. VP, P = 0.199; AP vs. EP, P = 0.260; VP vs. EP, P = 0.793).

Conclusion

MDCT-based radiomics features could be used as biomarkers for the differentiation of SBOTs and SMOTs.

Evaluation of a convolutional neural network for ovarian tumor differentiation based on magnetic resonance imaging

OBJECTIVES

There currently lacks a noninvasive and accurate method to distinguish benign and malignant ovarian lesion prior to treatment. This study developed a deep learning algorithm that distinguishes benign from malignant ovarian lesion by applying a convolutional neural network on routine MR imaging.

METHODS

Five hundred forty-five lesions (379 benign and 166 malignant) from 451 patients from a single institution were divided into training, validation, and testing set in a 7:2:1 ratio. Model performance was compared with four junior and three senior radiologists on the test set.

RESULTS

Compared with junior radiologists averaged, the final ensemble model combining MR imaging and clinical variables had a higher test accuracy (0.87 vs 0.64, p < 0.001) and specificity (0.92 vs 0.64, p < 0.001) with comparable sensitivity (0.75 vs 0.63, p = 0.407). Against the senior radiologists averaged, the final ensemble model also had a higher test accuracy (0.87 vs 0.74, p = 0.033) and specificity (0.92 vs 0.70, p < 0.001) with comparable sensitivity (0.75 vs 0.83, p = 0.557). Assisted by the model's probabilities, the junior radiologists achieved a higher average test accuracy (0.77 vs 0.64, Δ = 0.13, p < 0.001) and specificity (0.81 vs 0.64, Δ = 0.17, p < 0.001) with unchanged sensitivity (0.69 vs 0.63, Δ = 0.06, p = 0.302). With the AI probabilities, the junior radiologists had higher specificity (0.81 vs 0.70, Δ = 0.11, p = 0.005) but similar accuracy (0.77 vs 0.74, Δ = 0.03, p = 0.409) and sensitivity (0.69 vs 0.83, Δ = -0.146, p = 0.097) when compared with the senior radiologists.

CONCLUSIONS

These results demonstrate that artificial intelligence based on deep learning can assist radiologists in assessing the nature of ovarian lesions and improve their performance.

KEY POINTS

• Artificial Intelligence based on deep learning can assess the nature of ovarian lesions on routine MRI with higher accuracy and specificity than radiologists. • Assisted by the deep learning model's probabilities, junior radiologists achieved better performance that matched those of senior radiologists.

Evaluation of Ovarian Tumors with Multidetector Computed Tomography and Tumor Markers: Differentiation of Stage I Serous Borderline Tumors and Stage I Serous Malignant Tumors Presenting as Solid-Cystic Mass

Background

The aim of this study was to determine multidetector computed tomography (MDCT) features and tumor markers for differentiating stage I serous borderline ovarian tumors (SBOTs) from stage I serous malignant ovarian tumors (SMOTs).

Material/Methods

In total, 48 patients with stage I SBOTs and 54 patients with stage I SMOTs who underwent MDCT and tumor markers analysis were analyzed. MDCT features included location, shape, margins, texture, papillary projections, vascular abnormalities, size, and attenuation value. Tumor markers included serum cancer antigen 125 (CA125), carbohydrate antigen 19-9 (CA19-9), carcinoembryonic antigen (CEA), and human epididymis protein 4 (HE4). Parameters of clinical characteristic, MDCT features, and tumor markers were compared using a chi-square test and Mann-Whitney U tests. A binary logistic regression analysis was performed to detect predictors for SMOTs. A receiver operating characteristic (ROC) curve analysis was used to assess the potential diagnostic value of the quantitative parameters. Kappa and intraclass correlation coefficients were used to evaluate interobserver reproducibility for MDCT features.

Results

Median ages between patients with SBOTs and SMOTs were significantly different. Compared with SBOTs, vascular abnormalities were significantly more common in SMOTs. CA125, HE4, the maximum thickness of the wall, the maximum thickness of the septa, and the maximum diameter of the solid portions were significantly higher in patients with SMOTs. A binary logistic regression analysis revealed that age, vascular abnormalities, and the maximum diameter of the solid portion were independent factors of SMOTs. ROC analysis was used to assess the potential diagnostic value for predicting SMOTs. Moderate or good interobserver reproducibility for MDCT features were identified.

Conclusions

Age, vascular abnormalities, and the maximum diameter of the solid portion were independent factors for differentiating SBOTs from SMOTs. The combined analysis of age, vascular abnormalities, and the maximum diameter of the solid portion may allow better differentiation between SBOTs and SMOTs.

[Borderline Ovarian Tumours: CNGOF Guidelines for Clinical Practice - Value of Tumor Markers]

OBJECTIVES

To evaluate the diagnostic value of serum biomarkers in the management strategy of borderline ovarian tumours (BOT) to make management recommendations.

METHODS

English and French review of literature from 1990 to 2019 based on publications from Pubmed, Medline, Cochrane, with keywords: borderline ovarian tumors, tumour markers, CA125, CA19 9, ACE, CA72 4, TAG72, HE4, ROMA, mucinous, serous, mucinous, endometrioid ovarian tumours, peritoneal implants, recurrence, overall survival, follow-up. Among 1000 references, 400 were selected and only 30 were screened for this work.

RESULTS

Literature review: there is low evidence in literature concerning the discriminating value of serum tumour biomarkers (CA125, CA19-9, CEA, CA72-4, HE4) and specific score between presumed benign ovarian tumour/BOT/ovarian cancer (LE4). Serum CA125 antigen is higher in case of serous borderline ovarian tumour (LE4), increase with the tumor height, the FIGO stage, notably in case of serous borderline ovarian tumor. However, a normal value rate of serum CA125 antigen does not rule out a BOT (LE4). The preoperative positivity rate of CA19 9 in case of TFO is relatively lower than that of CA125 and is higher in mucinous TFO. The preoperative rate of serum CA19 9 antigen increases with the tumour height and the FIGO stage (LE4) and are higher in case of mucinous BOT (LE4). Preoperative rates of serum HE4 are not different between histologic type of BOT. A high level of serum biomarkers (CA125) is a predictive factor of peritoneal implants (LE4) and an independent predictive factor of recurrence (CA125) (LE4).

RECOMMENDATIONS

no recommendation can be made about the use of serum tumour biomarkers (CA125, CA19-9, CEA, CA72-4, HE4) or specific score in order to distinguish benign ovarian tumor/borderline ovarian tumor/ovarian cancer in case of indeterminate mass. In case of suspicion of mucinous ovarian tumour on imaging, the systematic dosage of serum CA19-9 antigen can be proposed (grade C). In case of an ovarian indeterminate mass on imaging; dosage of serum HE4 and C125 is recommended. If preoperative dosage of serum tumor biomarkers is normal, their systematic dosage is not recommended in the follow-up of BOT (grade C). If preoperative dosage of CA125 is high, the systematic dosage of CA125 is recommended in the follow-up of BOT with no precisions about the rhythm and the duration of the follow-up (grade B).

Imaging and Differential Diagnosis of Ovarian Cancer

Ovarian cancer is the seventh most common cancer affecting women. Despite advances in cancer control and healthcare in general, mortality from ovarian cancer remains unacceptably high due to diagnosis at an advanced stage of the disease. The 5-year survival rate is 47.4% because a majority of ovarian cancers are diagnosed when advanced. Only 14.9% of ovarian cancers are diagnosed when localized where the survival rate is 92.3%. Mortality rate reduction by screening has not been proven in women at an average risk for ovarian cancer. Ultrasound remains the primary modality for assessment of ovarian tumors. The need for standardizing terminology is critical for optimal assessment of the risk of malignancy in an ovarian tumor. The international ovarian tumor analysis group and more recently the American College of Radiology Ovarian - Adnexal Reporting and Data System Committee have published standardized lexicon for ovarian lesions and encourage ultrasound imagers to adopt this standardized terminology. The aim is to apply the lexicon for risk stratification to allow for consistent follow-up and management. Various methodologies have been tested for characterization of adnexal tumors and to assess risk of malignancy preoperatively. Risk assessment models have been studied against the gold standard of a pattern recognition approach and subjective assessment by an experienced imager. The morphologic patterns of ovarian tumors are detailed and features that are more discriminatory than others in suggesting an ovarian malignancy are described. The imaging pathologic correlation for different tumor types is presented. A brief summary of the ovarian cancer pathologic types and staging of cancer is presented. Finally, the current role of transvaginal sonography as a screening modality for ovarian cancer is discussed. Recently published data show encouraging results, that a multimodal approach of screening for ovarian cancer using transvaginal sonography in women with an elevated CA-125 may prove beneficial and cost effective.