As a common gynecological tumor, it is extremely important that ovarian cancer is accurately assessed preoperatively14. As a newly proposed US classification system, most studies are still in the process of validating the diagnostic efficacy of O-RADS and its observer agreement15,16,17,18,19,20,21. Currently, subjective assessment by senior sonologists is considered the most accurate method for diagnosing AMs22. In this study, we evaluated the performance of the O-RADS series of models in the preoperative identification of benign and malignant AMs and compared them separately with other commonly used clinical assessment methods. The overall results indicated that the O-RADS, particularly the O-RADS (v2022), has a high diagnostic efficacy and can assist the sonologists in the accurate preoperative assessment of benign and malignant AMs.
As in previous studies, there were statistically significant differences between benign and malignant AMs in this study in terms of patient age, lesion size, lesion type, and lesion blood flow score (all P < 0.001)17,21. Meanwhile, Di Legge et al.23 and Bruno et al.24 mentioned that even for lesions of small dimension, some ultrasound features such as irregular contour, absence of acoustic shadowing, vascularized solid areas, ≥ 1 papillae, vascularised septum and moderate-severe ascites, also play a role in the differentiation of benign and malignant lesions. The results of this study showed that the malignancy rates for O-RADS 2, O-RADS 3, O-RADS 4 and O-RADS 5 were 0%, 0%, 36% and 94.4% for O-RADS (v1) and 0%, 0%, 46.2% and 94.4% for O-RADS (v2022) respectively, with the malignancy rate of O-RADS 3 being less than the 1–10% provided in the guidelines7. Analysis of the reasons for this may be related to the small sample size included in this study and the small number of pathological types involved.
The low specificity of O-RADS (v1) has received a lot of attention10,19,21. Lan Cao et al.10 suggested that the diagnostic accuracy and specificity of O-RADS (v1) could be effectively improved if multilocular cysts and smooth solid masses in the 4 categories of O-RADS (v1) were classified as benign. Based on existing studies, O-RADS (v2022) provides a more specific classification of multilocular cysts and smooth solid masses in the O-RADS category 410,19,21. It also downgrades smooth bilocular cyst, which is ≥ 10 cm, and smooth solid lesion with acoustic shadow and color score (CS) of 2–3 to category 3. During the study, 11 benign lesions were successfully downgraded when classified using O-RADS (v2022), with significant improvements in diagnostic accuracy (84.4–89.4%) and specificity (79.5–86.1%) without altering sensitivity. These 11 lesions included three ovarian fibromas and one Brenner tumor (with acoustic shadow, CS = 2). Ovarian fibromas are the most common type of sex cord-stromal tumors and the lesions tend to present as smooth solid masses with acoustic shadowing and a small or moderate amount of blood flow (CS = 2–3)25. According to the O-RADS (v1) classification criteria, the lesions are mostly classified as O-RADS 47. When > O-RADS 3 is used as a predictor of malignancy, the lesions are often classified in the malignant category. The O-RADS (v2022) classification system classifies smooth solid lesions with acoustic shadowing and a 2–3 color score as category 3. When using this classification method, some ovarian fibromas and fibrothecomas with typical US features can be correctly classified as benign, effectively avoiding unnecessary surgery in some patients.
Lan Cao et al.10 proposed that the O-RADS (v1) category 4 of lesions are similar to the uncertain category in the IOTA SRs. To calculate the specific malignancy risk of lesions in the SRs model, the IOTA group developed the SRR assessment model in 201612. Numerous studies have confirmed that IOTA SRR assessment, ADNEX model and ORADS can help in the differentiation of benign and malignant masses15,16,17,18,19,20,21,26. In this study, the category 4 of O-RADS (v1) lesions were assessed for malignancy risk using SRR assessment and downgraded using 10% as the cutoff value, resulting in a combined assessment model with the largest AUC (0.976, 95% CI 0.946–0.992). However, the AUC of the combined model was not statistically significantly different from the AUC of O-RADS (v2022) (p = 0.1534). Thanks to its higher sensitivity, O-RADS (v1) is able to detect malignancies sensitively, minimising the occurrence of missed diagnoses, but its lower specificity may allow patients with AMs to be over-treated in the clinic15,19,21. Similar to the O-RADS (v2022) classification system, when assessed in combination with the SRR assessment, the specificity of the O-RADS (v1) was significantly improved (79.5–90.4, p = 0.006) without reducing diagnostic sensitivity. However, this is a single-centre study and much research is needed to determine the diagnostic efficacy of O-RADS (v1) combined with SRR assessment and how to further improve the specificity of O-RADS (v1) diagnosis.
A study by Moro et al.27 mentioned that serous borderline ovarian tumor showed an overlaping ultrasound appearance with non-invasive low-grade serous ovarian carcinoma, both presenting as cysts with papillary projections. However, unlike ovarian cancer, the prognosis for borderline tumors is relatively good, and women of fertile age can be treated with fertility-sparing surgery28. Therefore, it is extremely important and necessary to accurately distinguish borderline tumors from ovarian cancer before surgery. A total of 6 borderline tumors (4 Serous and 2 mucinous ovarian borderline tumors) were enrolled in the present study, and considering that the patients were in Stage I, and all were women of fertile age (range, 22–34 years), the surgical approach used for this group of patients was fertility-sparing surgery. Moro et al.27 proposed that the serous borderline ovarian tumor were described as unilocular-solid or as multilocular-solid with solid papillary projection. Meanwhile, another study by Moro et al29 suggested that a multilocular cyst with 2–10 locules is representative of a benign cystadenoma, whereas a multilocular cyst with > 10 locules is indicative of a gastrointestinal (GI)-type borderline tumor. The borderline tumors included in this study exhibited multilocular cyst (two cases, maximum diameter > 10 cm and > 10 locules) or multilocular cyst with solid component on ultrasound, and such lesions were classified as O-RADS categories 4 and 5 for both O-RADS (v1) and O-RADS (v2022) assessments, and lesions classified as category 4 failed to be downgraded for combined SRR assessment. Ludovisi et al.30 described the serous surface papillary borderline ovarian tumors (SSPBOTs) a rare morphologic variant of serous ovarian tumors that are typically confined to the ovarian surface, as irregular solid lesions surrounding normal ovarian parenchima. There were no SSPBOTs in the cases included in this study, but according to the O-RADS classification guidelines, such lesions met the classification criteria of O-RADS 5 in both O-RADS (v1) and O-RADS (v2022). Considering that the biological behavior of borderline tumors is intermediate between benign and malignant31, giving them a higher assessment of malignant risk can draw the attention of clinicians to avoid delaying patient treatment. However, the ability of the O-RADS classification system to identify borderline tumors is indeed limited, and which of the lesions assessed to be at moderate or high risk of malignancy are borderline tumors will have to be subjectively evaluated by experienced sonologists, which is a limitation of the O-RADS classification system that should be improved in subsequent studies.
The main strength of this study is that the results of subjective assessment and O-RADS (v1) assessment were collected prospectively and pathological results were available for all lesions. However, the O-RADS (v2022) classification results in this study were obtained from retrospective analysis of lesions, and the small sample size and single-centre nature of this study may lead to limitations in the wider application of the findings. In addition, all patients included in this study were those with obtainable pathology after surgery for AMs. Patients in both O-RADS 0 and O-RADS 1 categories were not included, which may result in selection bias and overestimation of PPV.
In summary, the O-RADS series models have good diagnostic performance for AMs. Among them, O-RADS (v2022) has higher diagnostic efficacy and diagnostic specificity than O-RADS (v1). However, when O-RADS (v1) is combined with SRR assessment, its diagnostic accuracy and specificity can be further improved.