Ultrasound evaluation of ovarian masses and assessment of the extension of ovarian malignancy

Abstract

The current review sums up the literature on the diagnostic performance of models to predict malignancy in adnexal masses and the ability of ultrasound to make a specific diagnosis in adnexal masses. A summary of the role of ultrasound in assessing the extension of malignant ovarian disease is also provided.

The ADNEX model and the O-RADS classification system are accurate ultrasound based-models that can be used in clinical practice. The International Ovarian Tumor Analysis group suggests using the International Ovarian Tumor Analysis Simple Descriptors as a first step, followed by ADNEX if the Simple Descriptors do not apply, in all females with an adnexal tumor.

We emphasize that the ultrasound image reflects the macroscopic appearance of tumors so that a specific diagnosis (dermoid, endometrioma, etcetera) can be assigned to an adnexal mass, and we describe the accuracy of ultrasound in detecting the extension of disease in patients with advanced ovarian cancer.

Introduction

Transvaginal ultrasound examination is an excellent diagnostic tool for discrimination between benign and malignant ovarian masses in the hands of experienced examiners using subjective assessment.^1–3 The Risk of Malignancy Index (RMI) (based on ultrasound parameters, menopausal status, and CA125 levels)⁴ and the ultrasound-based risk prediction models developed by the International Ovarian Tumor Analysis (IOTA) group may help less-experienced clinicians to discriminate between benign and malignant adnexal masses.^5–7 Adnexal masses judged to be benign can be safely managed with follow-up,⁸ whereas if there is a strong suspicion of malignancy the mass should be managed in an oncological referral centre.^9–12

The most used IOTA models are Simple Rules (SRs), simple rules risk (SRRisk) model, and Assessment of Different NEoplasias in the adneXa (ADNEX). The SRs are based on five ultrasound features suggestive of a benign lesion (B-features) and five ultrasound features suggestive of a malignant lesion (M-features). The IOTA SRs are popular because they are easy to use, without the need of a computer. However, in cases where neither M- nor B-rules apply, or where both are present, the mass is classified as ‘inconclusive’ and further investigations are required.⁵ Therefore, the SRRisk model was developed. It uses the number of benign or malignant ultrasound features present in a mass to calculate a risk of malignancy.¹³

The IOTA ADNEX model calculates not only the percentage risk of an adnexal mass being benign or malignant but the likelihood that the mass is benign, borderline, Stage I primary invasive malignancy, Stage II-IV primary invasive malignancy or a metastasis in the ovary from another primary tumor.⁷

European and North American gynecologists and radiologists have developed a management system for adnexal masses based on either specific ultrasound features or on the risk of malignancy calculated by ADNEX. This system classifies masses into different risk groups of malignancy and is called Ovarian-Adnexal Reporting and Data System (O-RADS). It provides a management recommendation for each risk category.¹⁴

The IOTA group has also described the typical ultrasound appearance of different adnexal pathologies, including that of various histological types of malignancy.^15–30

Ultrasound can also be used to assess the extension of malignant disease in the pelvis.^31,32 It has diagnostic performance similar to CT for assessing pelvic and abdominal tumor spread in females with epithelial ovarian cancer.³³

The current review sums up the literature on the diagnostic performance of models to predict malignancy in adnexal masses and the ability of ultrasound to make a specific diagnosis in adnexal masses. A summary of the role of ultrasound in assessing the extension of malignant ovarian disease is also provided.

IOTA ultrasound-based models to discriminate between benign and malignant ovarian masses

In all IOTA studies, the IOTA examination technique, measurement technique and terminology are used.³⁴ The aim of the IOTA studies is to develop and validate diagnostic models to discriminate between benign and malignant adnexal tumors in the hands of experts,^5,13,35 and non-expert ultrasound operators.^36–38

Logistic regression models (LR1-LR2)

Data from IOTA studies Phase 1, 1b and 2 were used to develop and validate two logistic regression models, LR1 and LR2, including 12 and 6 variables respectively.³⁹ The performance of these models has also been compared with that of RMI on the same patients.⁴⁰ Using a risk threshold of 10%, LR1 showed sensitivity of 92% and a specificity of 87%; LR2 showed a sensitivity of 92% and specificity of 86%; whereas RMI demonstrated a sensitivity of 67%, and specificity of 95%. LR1 and LR2 have been replaced by a more recent IOTA model (ADNEX) which has the advantage of using relatively few ultrasound parameters without considering color score.

Simple rules

SRs have been developed from patients collected in the IOTA study Phase 1 including 1066 patients from 9 centers in 5 countries. SRs are based on five benign (B) and five malignant (M) ultrasound features.⁵ The B-features are: unilocular cyst, presence of solid components with largest diameter <7 mm, acoustic shadowing, smooth multilocular tumor with largest diameter <100 mm, and no blood flow on color or power Doppler. The five M-features are: irregular solid tumor, ascites, four or more papillary projections, irregular multilocular solid tumor with largest diameter >100 mm, and very strong blood flow on color or power Doppler. The SR are applicable when at least one of the B-features is present without any M-feature (benign mass) or when at least one M-feature is present without any B-feature (malignant mass).

The performance of the SRs was compared with that of LR1 and LR2 using the IOTA 1b and IOTA two dataset, and their performance was similar. SRs were also temporally and externally validated using patients in IOTA Phase 1b, 2 and 3.^13,41 The SRs are not applicable in up to 77% of cases. If neither M- nor B-features apply, or if both M-and B-features are present, the diagnosis is ‘inconclusive’. In these inconclusive cases, subjective assessment by an experienced examiner can be used as a second-stage test. When externally validated, this “two-step strategy” (SRs followed by subjective assessment when SRs are inconclusive) had a sensitivity 90–92% and specificity 89–93%.^41,42 Alternatively, inconclusive cases can all be managed as malignant given the prevalence of malignancy is about 40%, or they can be triaged to perfusion MRI.⁴³

A small percentage (8–10%) of ovarian masses are difficult to classify as malignant or benign. For these, the use of 3D ultrasound^44–46 or contrast media^47–49 has been proposed but with limited additional value. MRI could play a role in classifying masses with unclear ultrasound diagnosis. A multicenter prospective study (IOTA-MRI) is ongoing to estimate the ability of pelvic MRI with diffusion- and perfusion-weighted sequences to correctly discriminate between benign and malignant masses unclassifiable by the SRs. Other ongoing prospective studies aim to explore the role of radiomics, artificial intelligence and proteomics in the preoperative assessment of adnexal masses.

Simple descriptors

The IOTA group defined six simple descriptors that can be used to make an instant diagnosis in an adnexal mass. They include four benign simple descriptors and two malignant simple descriptors. The benign simple descriptors are 1) unilocular cyst with ground glass echogenicity in a premenopausal female; 2) unilocular cyst with mixed echogenicity and acoustic shadows in a pre-menopausal female; 3) unilocular anechoic cyst with regular walls and maximum diameter of lesion <10 cm; 4) remaining unilocular cysts with regular walls. The malignant simple descriptors are: 1) tumor with ascites and at least moderate amount of color Doppler signals in a post-menopausal female; 2) age >50 years and CA 125 > 100 U ml⁻¹. The simple descriptors are applicable when at least one benign descriptor is present without any malignant descriptor (benign mass) and when at least one malignant descriptor is present without any benign descriptor (malignant mass). When applicable, the descriptors are very accurate (sensitivity 98%, specificity 97%).⁶ If none of the six simple descriptors can be applied, or if both a benign and malignant descriptor are applicable, the diagnosis is “non-instant”. For “non-instant masses”, a three-step strategy has been proposed, using SRs as second step and subjective assessment by an expert examiner as a third step when SRs are inconclusive.⁶ This three-step strategy had sensitivity 92% and specificity 92% when applied on 1938 patients examined in IOTA Phase 2 and sensitivity 92.5% and specificity 87.6% when applied on 2403 females examined in IOTA Phase 3.⁴² The three-step strategy has also been externally validated in the hand of examiners with different background, training and experience, showing a sensitivity 93% and specificity 92%.⁵⁰

ADNEX model

The IOTA model Assessment of Different NEoplasias in the adneXa (ADNEX) differentiates between benign tumors and four types of malignant ovarian tumors: borderline, Stage I ovarian malignancy, Stage II-IV malignancy, and metastases in the adnexa from other primary tumors. This approach is novel compared to existing tools that only differentiate between benign and malignant tumors. The model was developed from data of patients included in IOTA Phase 1,1b and 2 (n = 3506) and validated on those included in IOTA Phase 3 (n = 2403).⁷ On the validation data, ADNEX had sensitivity 96.5% and specificity 71.3% when using the 10% risk cut-off for total risk of malignancy. The ADNEX model discriminated well between benign tumors and each of the four types of malignancy (validation area under the receiver operating characteristic curve, AUC, between 0.85 and 0.99). The model distinguished Stage II-IV cancer from other malignancies (validation AUCs between 0.82 and 0.95) and showed fair discrimination between stage I malignancy and borderline tumors (AUC 0.75) and between Stage I malignancy and secondary metastatic cancer (AUC 0.71). ADNEX includes three clinical variables: age (years), type of center (oncology or not), Ca125 (U/mL) (optional), and six ultrasound variables: number of papillary projections (0, 1, 2, 3 or more than 3), number of cyst locules more than 10 (yes/no), acoustic shadowing (yes/no), largest diameter of the largest solid component (mm), largest diameter of the lesion (mm). ADNEX has been externally validated in the hands of examiners with varied training and experience in a multicenter cross-sectional cohort study including data from 610 females in 3 centees in Europe.³⁷ The AUC for ADNEX to differentiate between benign and malignant masses was 0.937 (95% CI: 0.915–0.954) when CA125 was included, and 0.925 (95% CI: 0.902–0.943) when CA125 was excluded. The model also showed good discrimination between the different subtypes of tumor.

Simple rules risk model

The SRs have the important disadvantage that they do not give a predicted risk, and hence not a level of confidence in the classification. This was overcome by the development of the SRRisk calculation tool. The SRRisk model uses the 10 ultrasound features in the SR in a mathematical model to calculate the likelihood of malignancy in all masses.¹³ The model was developed and validated using 5020 patients from IOTA Phases 1b, 2 and 3.

O-RADS classification

ADNEX can be used to classify patients in the O-RADS risk Groups 2 to 5,¹⁴ a classification system for pre-operative assessment of adnexal masses. The O-RADS system includes six categories (O-RADS 0–5). O-RADS 0 means incomplete evaluation, and a repeat ultrasound examination is recommended; O-RADS 1 means a normal ovary, and follow-up is not necessary; O-RADS 2 includes almost certainly benign masses with <1% risk of malignancy, no follow-up is recommended, but further characterization by an ultrasound specialist or an MRI study may be advised in some subgroups; O-RADS 3 includes adnexal masses with 1% to <10% risk of malignancy and management by a general gynecologist is recommended; O-RADS 4 includes adnexal masses with 10% to <50% risk of malignancy, and consultation with gynecologic oncologist is recommended; O-RADS 5 includes adnexal masses with 50–100% risk of malignancy and direct referral to a gynecologic oncologist is recommended.

This classification system was developed by both radiologists from North America and gynecologists of the IOTA group.

Comparison of the performance of different models

The performance of LR1, LR2, SRs and Simple descriptors have been summarized in a narrative review by Kaijser et al⁵⁰ (Figure 1).

Figure 1.

The image illustrates three strategies recommended by the IOTA group to be used to discriminate between benign and malignant adnexal masses on the basis of published work. IOTA, International Ovarian Tumor Analysis.

The strategy recommended by Kaijser et al in a narrative review proposes to use LR1, LR2, Simple rules or Simple Descriptors as a first step, to use subjective assessment (performed by an experienced examiner) as a second step if Simple Rules are inconclusive, or to use subjective assessment as a third step if Simple descriptors are not applicable and Simple rules are inconclusive.⁵⁰. The approach recommended by Meys et al in a systematic review includes either Simple rules with referral for subjective assessment by an expert ultrasound examiner if the Simple rules are not applicable, or use of the LR2 model if an expert examiner is not available.⁵¹ The strategy recommended by Van Calster et al in a multicenter cohort study, proposes to use ADNEX without CA125 as a first step if CA125 results are not available, and ADNEX with CA125 as a second step when there is a suspicion of malignancy.⁵²

Meys and colleagues performed a systematic review and metaanalysis comparing the diagnostic accuracy of subjective assessment, simple rules, LR2 and RMI for differentiating benign from malignant adnexal masses prior to surgery.⁵¹ They analyzed 47 articles, enrolling 19,674 adnexal tumors (13,953 benign and 5721 malignant), showing that subjective assessment by an experienced examiner was the best method to classify adnexal masses before surgery. SRs should be used by less-experienced ultrasound examiners. They proposed an approach using SRs with referral for subjective assessment by an expert examiner if the result of SRs is inconclusive (Figure 1).

ADNEX, SRs and SRRisk have undergone temporal validation on IOTA Phase 3 data (2009–2012) to assess their ability to discriminate between benign disease and Stage I–II primary ovarian malignancy.³⁵ This analysis included 1653 females from 18 centers, 230 of which had Stage I–II invasive ovarian malignancy. The results showed that the ability of the IOTA methods to differentiate benign disease from Stage I–II primary ovarian malignancy was similar to their ability to distinguish benign tumors from all malignant subtypes grouped together.

The IOTA group analyzed the performance of IOTA LR2, SRs, SRRisk, ADNEX without CA125 and ADNEX with CA125 and RMI when applied on all adnexal masses irrespective of management (conservative or surgical).⁵² The analysis was carried out on interim data from the IOTA Phase 5 study (2012–2017, 8519 patients from 36 centers in 14 countries).⁸ The authors concluded that ADNEX with or without CA 125 and SRRisk were the best models to distinguish between benign and malignant adnexal masses. They recommended the use of ADNEX rather than SRRisk in clinical practice because ADNEX uses variables that are less subjective (color score is not included) (Figure 1). In addition, if a mass is classified as malignant, ADNEX can calculate the likelihood of four subtypes of malignancy.

We want to emphasize that the diagnostic performance reported for each IOTA model is generalizable only to ultrasound examiners that use the standardized examination technique, measurement technique and terminology described in the IOTA consensus statement³⁴ (http://www.iotaeducation.org).

The practical application of IOTA models is illustrated in Figures 2–5.

Figure 2.

Ultrasound image (a) and clinical information (b) of a 39-year-old patient, examined at an oncological center for a unilocular-solid mass with a largest diameter of 68 mm in size and maximal diameter of the largest solid component 46 mm. The mass was moderately vascularized, there were more than three papillary projections and no acoustic shadows. No ascites was observed and serum level of CA 125 was 22 U ml⁻¹. Using Simple descriptors, the mass was classified as “non-instant” (c), because no descriptor applied. Using Simple Rules, the mass was classified as malignant (d), because one malignant feature (“at least four papillary projections”) and no benign feature was present. After including all the variables in the ADNEX model, the mass was classified as malignant (e), because the risk of malignancy was above the cut-off indicating malignancy (>10%). The highest relative risk corresponded to borderline ovarian tumor. The risk calculated by ADNEX classified the patient into O-RADS 5 (f). Final histology was serous borderline ovarian tumor.

Figure 3.

Ultrasound image (a) and clinical information (b) of a 57-year-old patient, examined at an oncological center for a multilocular tumor (more than 10 locules) with a largest diameter of 205 mm in size. Neither papillary projection nor acoustic shadow was observed. The mass was not vascularized. No ascites was observed and serum level of CA 125 was 22 U ml⁻¹. Using Simpe descriptors, the mass was classified as “non-instant” (c), because no descriptor applied. Using Simple Rules, the mass was classified as benign (d), because one benign feature (“no blood flow”) and no malignant feature was present. After including all the variables in the ADNEX model, the mass was classified as malignant (e), because the risk of malignancy was above the cut-off indicating malignancy (>10%). The highest relative risk corresponded to borderline ovarian tumor. The risk calculated by ADNEX classified the patient into O-RADS 4 (f). Final histology was mucinous borderline ovarian tumor.

Figure 4.

Ultrasound image (a) and clinical information (b) of a 51-year-old patient, examined at an oncological center for a solid ovarian mass with a largest diameter 42 mm in size. Neither papillary projection nor acoustic shadow was observed. The mass was richly vascularized. Ascites was reported and serum level of CA 125 was 423 U ml⁻¹. Using Simple descriptors, the mass was classified as “malignant” (c), because one malignant descriptor (“age >50 years and CA125 >100 U ml⁻¹”) and no benign descriptor was present. Using Simple Rules, the mass was classified as malignant (d), because three malignant features (“irregular solid tumor”, “presence of ascites”, “very strong blood flow”) and no malignant feature were present. After including all the variables in the ADNEX model, the mass was classified as malignant (e), because the risk of malignancy was above the cut-off indicating malignancy (>10%). The highest relative risk corresponded to II-IV ovarian cancer. The risk calculated by ADNEX classified the patient into O-RADS 5 (f). Final histology was high-grade serous ovarian carcinoma.

Figure 5.

Ultrasound image (a) and clinical information (b) of 68-year-old patient, examined at an oncological center for a solid ovarian mass with largest diameter 60 mm in size. Neither papillary projection nor acoustic shadow was observed. The mass was richly vascularized. No ascites was observed and serum level of CA 125 was 4.9 U ml⁻¹. Using Simple descriptors, the mass was classified as “non-instant” (c), because no descriptor applied. Using Simple Rules, the mass was classified as malignant (d), because one malignant feature (“very strong blood flow”) and no malignant feature was present. After including all the variables in the ADNEX model, the mass was classified as malignant (e), because the risk of malignancy was above the cut-off indicating malignancy (>10%). The highest relative risk corresponded to metastatic cancer to the adnexa. The risk calculated by ADNEX classified the patient into O-RADS 5 (f). Final histology was ovarian metastasis from gastric cancer.

Benefits and disadvantages of the models

LR1 and LR2 models include 12 and 6 variables, respectively. They are very accurate in discriminating between benign and malignant masses. However, some of the variables in these models (color score, and cyst-wall irregularity) are subjective, and some require ultrasound expertise (detection of color Doppler signals in a papillary projection). Therefore, the risk estimates calculated by LR1 or LR2 may be difficult to reproduce,⁵³ and the models may be less suitable for less-experienced ultrasound examiners.

A simple form using tick boxes that can be easily used in clinical practice to help less-experienced ultrasound operators is provided by the SRs. SRs have performance similar to that of LR1 and LR2. The limitation of the SRs in clinical practice is the high percentage of tumors in which results are inconclusive (almost 25% of the tumors). The SRs work well for endometriomas, dermoid cysts, simple cysts and advanced invasive malignancies, but they work less well for hydrosalpinx, peritoneal cysts, abscesses, fibromas, rare benign tumors, Stage I borderline tumors and Stage I primary invasive malignancies.⁵

The simple descriptors are very easy to use in clinical practice, but they are applicable only in about 40% of ovarian masses. The simple descriptors may be useful for teaching purposes, particularly for clinicians with limited experience of scanning ovarian tumors.⁶

ADNEX has the advantage of using few and simple ultrasound variables and of being able to calculate not only the risk of malignancy in general but also the likelihood of different types of malignancy. ADNEX can be used to classify patients into the O-RADS risk groups.

O-RADS has been compared with GI-RADS (Gynecologic Imaging Reporting and Data System) and with SRs in a retrospective multicenter study including 647 adnexal masses. In this study, five experienced consultant radiologists independently categorized each adnexal mass according to O-RADS (using the originally suggested system based on pattern recognition and on a combination of a large number of ultrasound features), GI-RADS and SRs; pathology and follow-up were used as reference standards. The O-RADS had significantly higher sensitivity for malignancy than GI-RADS and SRs (96.6 vs 92.7% and 92.1%; p = 0.003 and 0.0007, respectively) and non-significant slightly lower specificity (92.8 vs 93.6% and 93.2%; p > 0.05).⁵⁴

Ultrasound for making a specific diagnosis in ovarian tumors

The typical ultrasound appearance of different adnexal pathologies, including various histological types of malignancy, have been described in the “Imaging in gynecological disease” series in Ultrasound in Obstetrics and Gynecology. Ovarian tumors can be grouped into four histological categories: epithelial tumors, germ cell tumors, stromal tumors and metastases from other primary tumors. Each type of ovarian tumor has some typical macroscopic features, as described in text-books of pathology. Because the ultrasound image reflects the macroscopic appearance of tumors, an ultrasound examiner can often suggest a correct histopathological diagnosis. The typical ultrasound features of some specific types of ovarian tumor are shown in Table 1.(Table 1)^15–30

Table 1.

Summary of retrospective studies “Imaging in gynecological disease” reporting the typical ultrasound features of each subtype of ovarian tumor (Table 1)

Author, year	Pathology	Participants	Ultrasound findings	Pattern recognition
Testa, 2007	Ovarian metastases	67	Metastases from stomach, breast, lymphoma, uterus - median largest diameter of 71 (range 27–170) mm, solid morphology, moderate or rich vascularization. Metastases from colon, rectum, appendix or biliary tract median of the largest diameter of 122 (range 16–200) mm, multilocular or multilocular-solid morphology, moderate or rich vascularization.	-
Demidov, 2008	Sertoli cell tumor Sertoli-Leydig cell tumor, Leydig cell tumor	23	median largest diameter of 61 (13–175) mm, multilocular-solid or solid morphology, no papillary projection.	Leydig cell tumors: small solid tumors (1–3 cm). Sertoli cell tumors: solid tumors medium-sized (4–7 cm). Sertoli–Leydig cell tumors: either small (3–4 cm) or medium-sized (6–7 cm) solid tumors, or multilocular-solid tumors of any size (3–18 cm) with purely solid areas mixed with areas of innumerable closely packed small cyst locules.
Holsbeke, 2008	Granulosa cell tumors	22	median largest diameter of 102 (range 37–242) mm. multilocular-solid with ≥10 locules or solid morphology, moderate or rich vascularization.	“Swiss cheese aspect”.
Savelli, 2008	Struma ovarii	31	Benign pure struma ovarii (15) median largest diameter of 94 (range 25–198) mm, multilocular-solid morphology, moderate vascularization. Benign impure struma ovarii (15): median largest diameter of 58 (range 25–152) mm, multilocular-solid morphology, no or minimal vascularization, presence of shadowing. Malignant struma ovarii (1): largest diameter of 74 mm, solid tumor no vascularization.	“Struma pearl” (i.e., a smooth roundish solid area, similar but not identical to the ‘round white ball’ seen in dermoid cysts).
Paladini, 2009	Fibroma and fibrothecoma	68	Fibroma (53) median largest diameter of 78 (range 19–258) mm, solid morphology, minimal or moderate vascularization. Fibrothecoma (15) median largest diameter of 64 (range 27–143) mm, solid morphology, minimal or moderate vascularization.	Solid tumors with regular or slightly irregular internal echogenicity with stripy shadows, some contained cystic spaces.
Guerriero, 2011	Dysgerminomas	21	median largest diameter of 118 (range 40–210) mm, solid morphology, moderate or rich vascularization.	Solid tumor with lobules
Dierickx, 2012	Brenner tumors	28	Benign (23) median largest diameter of 70 (range 20–278) mm, solid morphology, no or minimal vascularization. BOT (2) median largest diameter of 90 (range 60–120) mm, unilocular or multilocular-solid morphology, minimal or moderate vascularization. Malignant (3) median largest diameter of 66 (range 16–143) mm, unilocular, multilocular or solid morphology, moderate or rich vascularization, presence of calcifications.
Franchi, 2013	Borderline ovarian recurrences	68	Serous BOT recurrence (62) median largest diameter of 31 (range 11–106) mm, unilocular-solid morphology, moderate vascularization. Mucinous BOT recurrence (6) Endocervical type: median largest diameter of 37.5 (range 27–48) mm, Intestinal type: median largest diameter of 58 (range 47–80) mm, multilocular or multilocular-solid cyst, moderate vascularization.	-
Ludovisi, 2014	Tubal carcinomas	79	median largest diameter of 58 (range 18–141) mm, multilocular-solid or solid morphology, moderate or rich vascularization.	Three types of ultrasound appearance: a) sausage-shaped cystic structure with solid tissue protruding into it like a papillary projection, b) sausage-shaped cystic structure with a large solid component filling part of the cyst cavity, c) ovoid or oblong completely solid mass.
Moro, 2017	Mucinous tumors	123	Mucinous cystadenoma (57) median largest diameter of 112.5 (range 31–301) mm, multilocular cyst with 2–10 locules, no or minimal vascularization. GI type BOT (24) median largest diameter of 195 (range 32–387) mm, multilocular cyst with >10 locules, minimal or moderate vascularization. Endocervical type BOT (10) median largest diameter of 37 (range 20–154) mm, unilocular-solid or multilocular-solid morphology, minimal or moderate vascularization. Mucinous invasive tumors (22) median largest diameter of 197 (range 49–300) mm, multilocular-solid morphology, moderate or rich vascularization.	-
Moro, 2017	Serous malignant tumors	406	Serous BOT lesions (64) median of largest diameter 59 (range 16–230) mm, unilocular-solid morphology, minimal or moderate vascularization. Non-invasive LGSCs (11) median largest diameter of 77 (range 22–201) mm, multilocular-solid morphology, minimal or moderate vascularization. Invasive LGSCs (31) median largest diameter of 73 (range 13–179) mm, multilocular-solid morphology, minimal to rich vascularization. HGSCs (300) median largest diameter of 87.5 (range 22–290) mm, multilocular-solid or solid masses, minimal to rich vascularization.	-
Moro, 2018	Endometrioid carcinomas	239	median largest tumor diameter of 102.5 mm (range 20–300 mm), multilocular-solid morphology, low-level or ground-glass content, moderate-richly vascularized.	“Cockade-like” appearance (large central solid component entrapped within locules giving the tumor a cockade-like appearance).
Pozzati, 2018	Clear cell carcinomas	152	median largest diameter of lesion 117 (25–310) mm, unilocular-solid/multilocular-solid/solid morphology, papillary projections minimal or moderate vascularization	_
Virgilio, 2019	Cystadenofibromas	233	median largest diameter of 76 (range 21–350) mm, unilocular-solid or multilocular- solid morphology, papillary projections, no or minimal vascularization.	Five patterns: unilocular or multilocular cysts with no solid components cysts with papillary projections but no other solid components cysts with solid components other than papillary projections, but no papillary projections cysts with both papillary projections and other solid components purely solid tumors Shadowing Microcystic appearance of papillary projections.
Anfelter, 2020	Yolk sac tumors	21	median largest diameter of 157 (range 107–181) mm multilocular‐solid or solid tumor rich vascularization.	Solid or multilocular-solid with inhomogeneous but still fine-textured and slightly hyperechoic solid tissue, in solid tumors this giving rise to a ‘Lunar-surface’ appearance.
Moro, 2020	Embryonal carcinoma, non-gestational choriocarcinomas, mixed germ cell tumors	12	median largest diameter of 129 (range 38‐216) mm, solid morphology, rich vascularization.	Large solid tumor with inhomogeneous echogenicity of the solid tissue and small and irregular dispersed cysts.

Ultrasound assessment of the extension of malignant disease

Accurate imaging of peritoneal carcinomatosis is key to providing appropriate management for patients with ovarian masses suspicious for malignancy. Historically, CT scan has been the most widely used imaging modality for ovarian cancer staging. Recently, MRI performed with standard and functional imaging sequences has been used with good results in terms of detection and evaluation of peritoneal carcinomatosis in ovarian cancer patients.⁵⁵

The performance of ultrasonography for assessing the extension of malignant disease has improved over the years. The potential role of transvaginal ultrasonography for the detection of peritoneal carcinomatosis in the pelvis was first shown by Savelli et al, who described its nodular and sheet-like appearance in 53 patients.⁵⁶ Later, Testa et al described the ability of transabdominal ultrasound to detect metastatic omental involvement and the typical ultrasound features of metastatic omental disease.⁵⁷ In a prospective study of 147 patients with ovarian cancer, the same group examined the ability of ultrasound to correctly determine the extent of intra-abdominal cancer spread. Ultrasound findings of peritoneal carcinomatosis, bowel mesentery, omental, or massive pelvic involvement, ascites and liver and/or spleen metastases were compared with surgical findings. Ultrasound demonstrated high sensitivity and specificity for diagnosing pelvic involvement (sensitivity 94%, specificity 97%), parenchymal liver metastases (sensitivity 93%, specificity 98%), ascites (sensitivity 98%, specificity 97%), peritoneal carcinomatosis (sensitivity 91%, specificity 88%) and omental involvement (sensitivity 94%, specificity 90%). The sensitivity for detection of spleen metastases and bowel mesentery involvement was low (75 and 67% respectively) but the specificity high (98 and 88% respectively).³² In the same study, a score to predict optimal cytoreduction was developed, but its performance was unsatisfactory.

The high accuracy of ultrasound examination in detecting the extension of disease was confirmed in a larger prospective single center study by Fischerova et al including 394 patients with primary ovarian cancer.³¹ The study showed high accuracy of pre-operative ultrasound for detecting pelvic carcinomatosis (AUC, 0.892; sensitivity, 81.4%; specificity, 97.0%) and rectosigmoid wall infiltration (AUC, 0.898; sensitivity, 83.1%; specificity 96.6%) when histology was used as reference standard. For the evaluation of peritoneal involvement in the upper and middle abdomen (diaphragm, liver or splenic surface, anterior wall/paracolic gutters, supracolic omentum, infracolic omentum, surface of the small bowel or colon, mesentery of the small bowel and/or colon), the sensitivity of ultrasound was poor. However, the specificity was high (>90%). The authors concluded that ultrasonography could play a key role in the pre-operative evaluation of rectosigmoid wall infiltration, and therefore for planning rectosigmoid resection. On the other hand, ultrasound examination failed to predict miliary mesenteric and intestinal involvement. This poor ability of ultrasound is similar to that observed for other imaging modalities (e.g. MRI and CT).^33,58,59

Espada et al⁵⁸ analyzed the diagnostic accuracy of diffusion-weighted MRI (DWMRI) compared to exploratory laparotomy in assessing the extension of disease in 34 patients with ovarian cancer. They found that DWMRI had sensitivity 75% and specificity 76% in assessing small and/or large bowel mesentery, sensitivity 75% and specificity 76% in assessing hepatic parenchyma/hepatic hilum or surface implant >2 cm, sensitivity 37% and specificity 92% for diaphragm involvement, and sensitivity 75% and specificity 76% for miliary visceral peritoneum implants.⁵⁸ Similarly, Michielsen et al⁵⁹ explored the diagnostic value of DWI/MRI in comparison with CT and FDG-PET/CT in staging 32 patients with suspicion of ovarian cancer. The authors demonstrated that DWI/MRI had high specificity for all parameters evaluated of the peritoneal cavity except for small bowel mesentery (71%) and high sensitivity for right diaphragm (92%), omentum (96%), and small bowel mesentery (100%). The sensitivity was low for left diaphragm (67%), hepatic surface (75%) and small bowel serosa (50%).⁵⁹

A study published by Alcazar et al demonstrated that ultrasound and CT had similar ability to detect disease in rectosigmoid, pelvic peritoneum, major omentum, abdominal peritoneum, bowel, root of mesentery, mesogastrium, hepatic hilum, liver and spleen parenchyma.³³ Their study is retrospective and includes a relatively small number of patients with ovarian cancer. Large prospective studies using standardized terminology are needed to evaluate the role of all imaging methods in ovarian tumor staging. It is also worth exploring the ability of different imaging methods to predict optimal cytoreduction. Another innovative research area is represented by the possibility to assess BRCA mutation status⁶⁰ by using radiogenomics applied to CT scan MRI or ultrasound examination.

Conclusion

In this review, we have summarized the achievement of 20 years of ultrasound research aiming at developing simple rules and mathematical models for discrimination between benign and malignant ovarian masses. This research resulted in the development of the ADNEX model and the O-RADS classification system suitable for use in clinical practice. The IOTA group suggests using the Simple Descriptors as a first step, followed by ADNEX if the Simple Descriptors do not apply, in all females with an adnexal tumor.

In this review, we have also summarized the role of ultrasound examination in assessing disease extension in patients with ovarian cancer. Ultrasound examination can be used for initial selection of patients to either primary debulking or neoadjuvant chemotherapy, provided that an experienced examiner and high-end ultrasound equipment are available.