The fifth leading cause of cancer death among women is ovarian cancer. It is a disease of post-menopausal women and sometimes prepubescent girls. Risk factors include age more than 50, positive family history, infertility, and previous cancer [9]. This study included 56 ovarian lesions in 40 patients (16 cases showed bilateral masses). The patients’ age ranged from 14 to 75 years old (mean age 45 +/− 16.073 SD).
Ovarian masses are commonly seen in clinical practice and may be incidentally detected in symptomatic patients. Characterization of an ovarian lesion represents a diagnostic challenge; it is of great importance in the preoperative assessment in order to plan adequate therapeutic procedures and may influence the patient’s management [1]. US is the first-line imaging modality for adnexal lesions and is a useful preoperative test for the characterization of noncomplex masses. MRI may be of great help in identifying malignant lesions before surgery, particularly when US findings are suboptimal or indeterminate [6]. In this study, we analyzed the diagnostic performance for the pre-contrast MR sequences, DWI, DCE-MR imaging, and MRS in the evaluation of adnexal masses. We found 15 lesions (26.8%) were classified as GI-RADS 5, 20 lesions (35.7%) were classified as GI-RADS 4, all of GI-RADS 4–5 lesions were found malignant by conventional and functional MRI, and 21 lesions (37.5%) were classified as GI-RADS 3 which in MRI showed an overlap between borderline/ benign lesions.
DW-MRI is an important technique that enables the radiologist to move from morphological to functional assessment of diseases of the female pelvis [10]. In this study, 70% of the lesions showed restricted diffusion (hyper-intense signal in DWI and hypo-intense in ADC map), whereas 30% displayed facilitated diffusion. All the malignant (n = 27, 100%) and two of the borderline lesions (n = 2, 16.6%), as well as 10 (58.8%) benign lesions that showed restricted diffusion (7 TOAs, 2 broad ligament fibroid, 1 chronic ectopic pregnancy).
Regarding the mean ADC values, we found the solid components for the benign lesions differed significantly from that of the borderline and invasive malignant lesions (P ≤ 0.001). We conclude that ADC measurement in the solid components was more specific for differentiating benign from malignant lesions; this agrees with a study carried out by Zhang and colleagues [6], they concluded that the presence of a solid component with high or low signal intensity on T2-weighted images and restricted on DWI with low ADC values (less than 1.20 × 10−3 mm2/s) at b = 1000 s/mm2 are predictive of malignancy. However, the presence of a solid component with high or low signal intensity on T2-weighted images and facilitated on DWI being high signal intensity on DWI with high ADC values (greater than 1.20 × 10−3 mm2/s), or low signal intensity on T2-weighted images and DWI with lower ADC values at b = 1000 s/mm2 are predictive of benignity.
In contrast, Fujii and colleagues [11] conducted a study on 123 ovarian lesions, recorded that the most malignant ovarian tumors, as well as some of the mature cystic teratomas, and showed high signal intensity on DWI. Also, they concluded that the mean ADC value of the solid portion in malignant tumors did not significantly differ from that in the benign lesions (mean ADC for the benign lesions was 1.47 ± 0.42 and mean ADC for the malignant lesions was 1.41 ± 0.34 (× 10−3 mm2/s)). They attributed this finding to the inclusion of sex cord-stromal tumors, Brenner’s tumor, and cyst-adenofibroma, all of which have a dense network of collagen fibers and thus resulting in low ADC values similar to the malignant lesions.
In the current study, also, there was a statistical significance (P = 0.002) between the mean ADC values of the cystic components of the benign and borderline/invasive malignant lesions. It may be attributed to the inclusion of the TOA cases which had low ADC values of their cyst contents. All the seven lesions of the TOAs showed diffusion restriction of their cyst contents. Their mean ADC value ranged from 0.6 to 0.8 × 10−3 mm2/s (± 0.19 SD), with a mean value of 0.7 × 103 (+/− 0.16SD) mm2/s. However, the rest of the benign lesions (which was 1.97 ± 0.428 after the exclusion of the TOAs) showed no significant difference from that of the borderline (1.83 +/− 0.27) or the invasive malignant tumors (2.1 +/− 0.44). This agrees with a study conducted by Wang et al [12] on 69 patients (34 patient with TOA for characterizing the TOA mimicking ovarian malignancy); it showed that the mean ADC value of TOA cystic component was lower than that in malignant tumors (1.04 ± 0 0.41 × 10−3 mm2/s vs. 2.42 ± 0.38 × 10−3 mm2/s; P < 0.001).
In our study, the addition of the DWI improved the sensitivity, PPV, NPV, and accuracy of the conventional MRI from 74%, 76%, 44%, and 66% to 89%, 78%, 64%, and 75%, respectively, yet the specificity decreased from 47% to 41%. Such low specificity elicited in our research is explained by the presence of benign cases that have mimicked malignancy on DWI, starting from their misleading signal intensities of restricted diffusion, down to the low ADC values measured; such cases include TOA (n = 7), broad ligament fibroid (n = 2), and chronic ectopic pregnancy (n = 1) showed restricted diffusion due to mixed cellularity of such tumors.
A study done by Mansour et al [13] found that the solo performance of DWI is not an applicable way to discriminate benign from malignant ovarian masses. DWI has high sensitivity and specificity in diagnosing a malignancy in suspicious ovarian masses, provided (1) inclusion of the conventional MRI data, (2) combined analysis of DWI quantitative and qualitative criteria and (3) awareness of the sequence pitfalls. The sensitivity, specificity, PPV, NPV, and accuracy of adding DWI to conventional MR imaging was 93%, 85%, 88%, 94%, and 82%, respectively, compared to 93%, 100%, 100%, 92%, and 95% after adding DCE-MRI to the conventional MR.
In contrast to a prospective analysis done by Thomassin-Naggara et al [14], they characterized 77 adnexal masses using DWI. They considered the SI at the DWI to be the accurate tool for predicting benign/malignant criteria, not the ADC values; however, in our study, the ADC value was the most accurate tool.
Challenging cases in our study included (1) TOAs, in which their cystic components showed high signal on the DWI with low ADC values that ranged from 0.6 to 0.8 × 10−3 mm2/s in the ADC maps, giving a picture similar to a malignant tumor. However, some of those can be diagnosed by detailed clinical history and tumor markers. (2) Broad ligament fibroids, as they demonstrated intermediate to high signals in DWI with low ADC values, fortunately, their iso-intense T1 signals suggested their benign pathology. (3) Ectopic pregnancy, it was predominantly cystic lesion with internal fine mural nodules < 3 mm and showed enhancement of the wall and the fine mural nodules in the post-contrast series and demonstrated high signal in DWI with low ADC values, it was only proven to be ectopic pregnancy only after surgical removal.
We had 12 BOTs that showed variable appearance on DWI. About 10 cases showed facilitated diffusion of its solid component with a high ADC value (1.4 × 10−3 mm2/s) suggesting benign pathology (false negative). Two cases showed restricted diffusion with high signal on DWI and mean ADC value of 0.6 × 10−3 mm2/s, suggesting a malignant invasive pathology. Also, borderline pathology is suggested based on its morphological features “presence of vegetation or thick septations.”
Quantitative DCE-MRI provides an accurate method for the prediction of malignancy, particularly in preoperative indeterminate cases [15]. Sohaib and colleagues [16] described that malignant lesions show greater enhancement than benign lesions during the early phase of enhancement rather than the late phase of enhancement, while benign ovarian tumors showed a gradual increase in enhancement without a well-defined peak, while, borderline ovarian tumors showed moderate initial enhancement followed by a plateau.
Another study carried out by Thomassin-Naggara et al [17] showed that curve type 3 appeared specific for invasive tumors. Curve type 1 was more frequent in benign than in malignant tumors. No difference was found among the three groups (benign, borderline, and malignant) regarding the frequency of curve type 2. However, there was an overlap among the curve types of the benign and borderline lesions; therefore, their capacity to differentiate benign from borderline tumors was low.
In our work, 7 lesions out of 17 benign lesions (41.2%) showed a steady rise with no definite peak (type 1 curve), while the other ten lesions (58.8%) showed a rapid rise with a plateau. Some of the invasive malignant lesions 18 (66.6%) showed an initial rapid steep early enhancement (type 3 curve). However, there was a great overlap between benign and borderline lesions, as we had 6 borderline tumors (borderline serous cystadenomas) which had type 1 curve. We also had 9 cases of malignant tumors which demonstrated type 2 curve (2 bilateral dysgerminomas, 2 mucinous cystadenocarcinomas grade I, 2 granulosa cell tumor, 3 moderately differentiated serous cystadenocarcinoma). Because of the overlap of curve patterns with respect to the histological type, we combined the semi-quantitative parameters.
A study was carried by Kazerooni et al [18] which showed that early enhancement features, i.e. TTP as the most sensitive and WIR as the most specific single classifier, can best describe the properties of ovarian masses, so both have been reported as helpful indicators for discriminating benign and malignant ovarian masses. Therefore, by inspecting DCE-MRI curves in malignant lesions, a steep slope of enhancement, higher WIR, and smaller initial time to enhancement peak can be observed. A study conducted by Li and colleagues [19], differentiating malignant and benign ovarian lesions on 48 ovarian tumors (13 benign and 35 malignant) investigated the TTP, produced a sensitivity of 100% and specificity of 92.31%.
In this work, we had found a Tmax cutoff of ≤ 145 s as predictive of borderline/malignant has produced a maximum statistical significance (P < 0.001) with a sensitivity of 78.7% and specificity of 84.6%.
We agreed with a study conducted by Mansour and colleagues [15] on 150 complex or purely solid ovarian masses evaluating the ability of dynamic post-contrast sequence to specify indeterminate ovarian masses, which showed that MRE% was higher for malignant than for benign and borderline masses (P < 0.001), but no significant difference was noted between benign and borderline ones (P > 0.05).
In this study, we found that MRE% was significantly higher in malignant (mean value of 125.3 ± 54.8 SD) than in benign lesions (mean value 78.25 ± 51.3 SD), and even higher than in borderline lesions (mean value 129 ± 5.2 SD). However, there was some overlap regarding the MRE values between the benign and borderline lesions. However, this can be used to exclude invasive malignancy.
In the study of Bernardin et al [20], calculating mean SImax of the malignant (invasive/borderline lesions) calculated on solid enhancing target components was 712 (278.6 SD), while for benign lesions, it was 491.2 (467.2 SD) with statistical difference (P = 0.018).
In our study, using a SImax threshold of ≥ 990 as predictive of malignant invasive pathology gave a sensitivity of 90.8% and a specificity of 62.4%. However, using WIR, we found that the mean value of the WIR in the malignant invasive lesions (19.2 ± 7.4 SD) was much higher than that of the benign (8.17 ± 4.6SD) and borderline (10.3 ± 8.4SD) lesions. Using a WIR cutoff of ≥ 12.8 as predictive of invasive malignant lesions produced a sensitivity of 68.5 % and specificity of 75.9%.
This was in agreement with Bernardin et al [20] that found a statistically significant difference in SImax, SIrel, and WIR between borderline and invasive malignant tumors. Applying a cutoff WIR ≥ 9.5 as predictive of borderline/invasive malignancy produced optimal diagnostic performance providing a sensitivity of 67% and specificity of 88%.
The TOA were such challenging cases with the use of DCE parameters as they had high MRE% and WIR data (mean values 97.3 ± 52.5 SD and 16.6 ± 4.4 SD, respectively), they had a mean Tmax 210.8 ± 68.3. Applying a Tmax threshold of ≤ 145, we had three cases of tubo-ovarian complexes that remained indeterminate. However, fortunately, these types of lesions can be easily diagnosed based on conventional MRI and diffusion findings.
So based on the DCE findings in our work, Tmax and WIR had the best results showing the highest sensitivity and specificity among the other DCE criteria, along with the type curve. Although DCE-MRI has high specificity for depicting invasive lesions, there was no statistical significance between the benign and borderline lesions regarding the DCE parameters (P > 0.05).
DCE-MRI techniques are also prone to pitfalls. False-negative results may occur with poorly vascularized malignant tumors, and false-positive enhancement characteristics may be seen in benign lesions with a high blood supply, such as tubo-ovarian abscess, which may appear complex and indeterminate with all imaging modalities [21].
Molecular imaging through MRS can detect metabolic features characteristic of malignancy. As molecular changes often precede morphologic alterations, sensitivity is expected to improve by MRS [22].
A study conducted by Malek et al [23] on 23 ovarian masses showed the presence of choline peak in 17 of 19 malignant masses (sensitivity 89%) and in 3 of 4 benign masses. So they considered that it could not be used in differentiating between benign and malignant tumors.
In our study, we assessed the metabolites’ peaks in 40 lesions. Choline peak was detected in 32 lesions (80%), including 10 out of 16 (62.5%) benign lesions (4 out of 6 abscesses, mucinous, and serous cystadenomas). It was above the noise level but lower than twofold. Two borderline (100%) lesions and 22 malignant (100%) with sharp peaks were found in serous and mucinous cystadenocarcinomas.
This agreed with a study conducted by El Sorogy et al [22] showed Cho peak was detected in all cases of solid tumor or solid parts in cystic tumors. However, the Cho peak was found even in benign tumors and could not be used for the differential diagnosis between benign and malignant tumors.
A study conducted by Stanwell et al [24] showed a higher Cho/Cr ratio in malignant ovarian cancer than in benign cystic ovarian tumors. Also, a study conducted by El Sorogy et al [22] showed the sensitivity of Cho/Cr ratio in differentiating benign from malignant ovarian lesions was 0.83 and specificity 0.82.
In this study, mean Cho/Cr ratio was 1.29 ± 0.98 SD for malignant lesions, while the mean value in borderline lesions was 0.63 ± 0.15 SD and the mean value for the benign lesions was 0.65 ± 0.34. There was a statistical significance regarding the Cho/Cr ratio between the benign and invasive malignant lesions (P ≤ 0.001) as well as between the borderline and invasive lesions (P = 0.05), but not between the benign and borderline lesions (Table 8).
A study conducted by Stanwell et al [24] reported the presence of NAA in all of the teratomas and serous cyst-adenomas, as well as a portion of the serous carcinomas. Also, its amplitude was greater in malignant than in benign tumors. Kolwijck et al [25] concluded that both NAA and N-acetyl groups from glycoproteins and/or glycolipids may contribute to the Delta 2.0–2.1 ppm resonance complex in the ovarian cyst fluid.
In our study, we found NAA peaks in serous and mucinous cystadenomas and tubo-ovarian abscesses. NAA peak was also detected in one borderline tumor and most of the malignant tumors (19 lesions).
A study conducted by El Sorogy and colleagues [22] found a very striking finding is that lactate signal was found in almost all benign lesions, and not obtained in any of malignant lesions except dysgerminoma. Another study conducted by MA and colleagues [26] on 69 patients, choline peak was detected in all cases (100%), NAA peak in 67 cases (97%, 25 benign and 42 malignant), lipid peak in 47 cases (17 benign and 30 malignant), and lactate peak in only eight cases (four benign and four malignant).
In our study, sharp lactate peaks were found in serous benign lesions, TOA, broad ligament fibroids, solid portions of the borderline lesions as well as the solid portions of the malignant serous and mucinous cystadenocarcinoma lesions (Table 9).
BOTs are known also as tumors of low malignant potential. They are characterized by atypical epithelial proliferation and moderate nuclear atypia but without stromal invasion. Patients with BOTs generally have an excellent prognosis after surgical excision because they are more likely to present in an early stage; however, they show a high recurrence rate. It is important to preoperatively discriminate BOTs from invasive malignant tumors because a conservative fertility-sparing laparoscopic surgery can be performed in the former [27]. The preoperative diagnosis of BOTs remains challenging regarding the clinical, laboratory, and imaging findings [28]. Morphological imaging features suggesting borderline tumors have been investigated; however, there are overlapping with those of invasive epithelial tumors. So, better imaging tools are needed to improve the distinction of invasive, borderline, and benign adnexal lesions [20].
In our study, BOTs presented a great challenge for us. Their morphological features can mimic benign or malignant lesions, yet they lack the invasive behavior of the malignant lesions. However, in functional imaging, most of them mimic benign lesions, which is consistent with their non-invasive behavior.