From January 2019 to December 2022, 210 patients were referred to our department to perform an MRI for the characterisation of a US indeterminate adnexal mass. We excluded 27 patients from the study. Fifteen due to the whole MRI protocol for characterisation of an ovarian mass was not completed, 11 were lost to follow-up, and one who presented acute pelvic pain at the time of MRI acquisition (Fig. 1).

Population flowchart. US, ultrasound. Note that in 25 patients two ovarian masses were found; in these cases, only the mass with the greatest O-RAD score was included. In the 57 patients with O-RADS 1, 40 were benign, 8 malignant, and in 9, no mass was found. Borderline and malignant were considered malignant for statistical purposes

A total of 183 women (mean age 51; standard deviation 16) with 208 US indeterminate or solid adnexal masses were included in the study. Twenty-five patients (13,7%) had two adnexal masses, and in these cases, only the mass with the highest score was considered for the statistical analysis. Thus, the final number of masses analysed was 183. The clinical characteristics of the patients are summarised in Table 2. All the patients in the study were discussed at the oncologic gynaecology multidisciplinary committee of our hospital, as is the standard procedure for all patients with indeterminate or suspicious adnexal masses in our centre.

In relation to the US characteristics, 125 out of 183 masses (68.3%) were classified as indeterminate according to “IOTA-SR”; 37 (20.2%) met criteria for category B although seemed suspicious by the expert gynaecologist’s opinion (classified as indeterminate according to “IOTA-SR modified by expert”), and 21 (11.47%) met criteria for category M but were completely solid and hypervascular on Doppler study (Table 2).

Of the 183 masses, 57 (31%) were extraovarian pelvic masses or physiological ovarian findings classified as O-RADS 1 (Fig. 2). Of these 57 masses, 40 were benign (70.2%), with 21 (36.8%) of these masses corresponding to fibroids. Less frequent extraovarian benign masses included peritoneal inclusion cysts (5, 8.8%), deep endometriosis (4, 7%), and diverticulitis (3, 5.3%). In isolated cases, we found a uterine malformation, an extramedullary haematopoiesis focus, pelvic gross calcifications related to a mesenteric granuloma, endometrial polyp, and a non-gynaecological abscess. In 8 cases (14.1%), malignant extraovarian pelvic masses, such as appendiceal neoplasms (3 patients, 5.3%), neurogenic tumours (2 patients, 3.5%), peritoneal implants (2 patients, 3.5%), or sigmoid neoplasms (1 patient, 1.8%), were found. In 9 (15.7%) patients, no pelvic mass was found.

Distribution of non-adnexal masses and examples. a O-RADS score 1 distribution. b Parametrial fibroid (arrow). c Uterine adenomyosis (arrow). d Appendiceal gastrointestinal stromal tumour (GIST) (arrow) e Sigmoid adenocarcinoma (arrow)

Of the 126 patients with true adnexal masses, 62 were menopausal. The series comprised 77 benign masses (61.1%), 17 borderline masses (13.5%), and 32 malignant masses (25.4%). Borderline and malignant masses were both considered malignant for the statistical analysis. A total of 80 (63.5%) patients underwent surgery and 46 (36.5%) were followed for at least 1 year. The final diagnoses (histopathology results for the operated adnexal masses and clinical diagnosis after 1 year of follow-up) are summarised in Table 3. None of the 46 patients followed for 1 year presented progression of the disease during the study duration.

The JR assigned a score 2 or 3 to 77 of the 126 masses (61.1%). Of these, 4 (5.2%) were finally malignant lesions in the histopathological analysis (a borderline serous tumour, an immature teratoma, an ovarian metastasis from a mucinous appendiceal neoplasm, and a clear cell carcinoma arising from a cystadenofibroma). The SR assigned a score of 2 or 3 to 74 masses (58.7%), and all were benign in the histopathological analysis or remained stable after 1 year of follow-up.

The JR assigned a score 4 or 5 to 49 of the 126 masses (38.9%). Of these, 3 (6.1%) were finally benign lesions (two fibrothecomas and one mucinous cystadenoma mixed with a Brenner tumour). The SR assigned a score 4 or 5 to 52 of the 126 masses (41.2%). Of these, 2 (3.8%) were benign lesions (two fibrothecomas, one with luteinisation).

In the case of the JR, the percentage of malignancy in O-RADS MRI scores 2, 3, 4, and 5 was 1.20%, 10.80%, 93.30%, and 93.80%, respectively. The percentage of malignancy according to O-RADS MRI scores 2, 3, 4, and 5 by the SR was 0%, 1%, 100%, and 94.10%, respectively.

Overall, there were seven misclassified cases, five by the JR and two by both the JR and the SR. The cases misclassified by both radiologists (Fig. 3) corresponded to fibrothecomas that showed a solid component with a high-risk TIC. The other cases misclassified by the JR are shown in Figs. 4, 5, and 6 and correspond to a metastasis of a mucinous appendiceal tumour with a small solid component that was classified with a score of 3 because it had a low-risk TIC, a borderline serous tumour with a low-risk TIC that was classified with a score of 3, a solid-cystic mass classified as score 3 because of a misinterpretation of the TIC and was finally a clear cell carcinoma arising from a cystadenoma, a solid mass with macroscopic fat content that was classified as score 2 and was actually an immature teratoma, and finally, a solid-cystic mass with an intermediate-risk TIC classified as with a score of 4 that was a mucinous cystadenoma mixed with a benign Brenner tumour in the postoperative pathological analysis.

Erroneous classifications by the JR and the SR. A A 61-year-old woman presented an incidental right ovarian mass (green arrow) with a high-risk TIC classified as score 5 by both readers but was finally a fibrothecoma. As specified by O-RADS MRI guidelines, unenhanced sequences should be acquired before the contrast bolus injection. However, in this case, there was an error in the acquisition as there were no unenhanced sequences before the injection of the contrast bolus (in both TIC the contrast uptake started at the second 0), and this can distort the TIC results and lead to misinterpretation. In clinical practice, cases like this should be considered O-RADs 0 (incomplete or erroneous MRI technique). B A 56-year-old woman presented an incidental left ovarian mass (blue arrow) with a high-risk TIC that was classified as score 5 by both readers but was finally a luteinised fibrothecoma

Errors by the JR due to misinterpretation of the classification. A A 31-year-old woman presented a right ovarian mass with macroscopic fat content (blue arrow) and a high amount of solid-enhancing tissue (T1W FS + C series). The mass does not have a Rokitansky nodule. It was classified as score 2 by the JR but the histological result showed an immature teratoma. Teratomas do not fit the classification well as they can have low, intermediate or high-risk TIC. In this case, the mass had an intermediate TIC. It is difficult to distinguish mature from immature teratoma and thus, it is stipulated that if they present a high amount of solid tissue, they should be classified with a score of 4. B A 79-year-old woman with a left ovarian mass (yellow arrow) with solid hyperenhancing tissue (T1W FS + C series). The TIC was interpreted as low risk by the JR and the mass was misclassified as score 3. Pathological analysis after surgery confirmed that it was a clear cell carcinoma arising from a cystadenoma. In this case, the TIC was an intermediate-risk curve as it had an initial increase lower than the myometrium, followed by a plateau

Errors by the JR, paradigmatic examples. A A 33-year-old woman with bilateral adnexal masses with a tree-like morphology (green arrow) and low contrast uptake shown by the TIC. This case was classified as score 3 by the JR as it was considered that it had a low-risk TIC. Pathological analysis after surgery confirmed that it was a borderline serous tumour. B A 67-year-old woman with a left ovarian mass that shows hyperintense T2w content and multiple thin septa (yellow arrow) typical of mucinous tumours. This mass was misclassified as score 3 by the JR as it was interpreted as having a low-risk TIC. The histological results showed metastasis of a mucinous appendix tumour

A 74-year-old woman with a right ovarian solid-cystic mass. This case was misclassified as score 4 by the JR considering it as having an intermediate-risk TIC. The postoperative pathological analysis revealed that it was a mucinous cystadenoma mixed with a Brenner tumour. In this case, the JR calculated the TIC out of the ovarian parenchyma surrounding the lesion (green arrow), but the true solid component was the thin septa (blue arrow) that corresponded to a score 3 as the SR perceived. As in this case, false positives may be due to errors performing the TIC

The diagnostic performance is summarised in Table 4.

The classification carried out by the SR obtained a sensitivity of 97.4% (95% CI: 90.9; 99.3), a specificity of 100% (95% CI: 95.2; 100.0), a PPV of 96.2% (95% CI: 87.0; 98.9), and a NPV of 100% (95% CI: 92.9; 100.0) for the prediction of malignancy.

The classification by the JR obtained a sensitivity of 96.1% (95% CI: 89.0; 98.9), a specificity of 92.0% (95% CI: 81.2; 96.8), a PPV of 93.9% (95% CI: 83.5; 97.9), and a NPV of 94.8% (95% CI: 87.4; 98.0) for the prediction of malignancy.

Considering the score range 2–3 as benign and the score range 4–5 as malignant, there was discordance in only 5 of the 126 studies (6%). The interobserver agreement between the JR and the SR was excellent, with a Kappa index of 0.92 (95% CI 0.86; 0.98).