The diagnostic efficacy of Gynecology Imaging Reporting and Data System (GI-RADS): single-center prospective cross-sectional study
Egyptian Journal of Radiology and Nuclear Medicine volume 50, Article number: 61 (2019)
To assess the validity and accuracy of GI-RADS classification in the prediction of malignancy and in triaging the management protocol in ovarian lesions.
One hundred fifty-six ovarian lesions were detected in the examined 116 women. The prevalence of malignant tumors was 44%. Overall GI-RADS classification rates were as follows: 41 cases (26.3%) were classified as GI-RADS 1, 26 cases (16 .7%) as GI-RADS 2, 34 cases (21.8%) as GI-RADS 3, 14 cases (8.9%) as GI-RADS 4, and 41 cases (26.3%) as GI-RADS 5. No follow-up was done in GI-RADS 1 patients. A final diagnosis of all GI-RADS 2 ovarian masses such as functional cyst (n = 10), hemorrhagic cysts (n = 8), corpus luteal cysts (n = 6), and some GI-RADS 3 as simple cysts (n = 10) was made by spontaneous resolution of these masses at follow-up after 6 weeks. Fifteen cases of GI-RADS 3 as mature teratoma, serous and mucinous cystadenoma, endometrioma, and ovarian torsion and all GI-RADS 4 and 5 underwent laparoscopic or surgical removal of the ovarian mass with histopathological examination. The diagnostic performance of the GI-RADS in predicting the risk of malignancy in ovarian masses was as follows: 98.11% sensitivity, 95.15% specificity, 91.2% positive predictive value (PPV), 99.2% negative predictive value (NPV), and 20.2 positive likelihood ratio, and the overall accuracy was 96.2% (area under receiver operating curve (AUC) = 0.96, P < 0.001).
GI-RADS classification performs well as a reporting system of the ovarian masses with high diagnostic performance in the prediction of malignancy, and it seems to be a helpful tool in triaging management in patients with ovarian masses.
The trial was registered in the US National Library of Medicine, under clinical trial number NCT03175991. Also, the ethical committee approval number of the Faculty of Medicine, Assiut University, was 17100016 on February 28, 2017.
Ovarian cancer is fatal cancer among gynecological malignancies . In Egypt, ovarian cancer represented 2.2% of all incident cancers and accounted for 4.4% of all newly diagnosed female cancers . The assessment of an adnexal mass is difficult and meticulous preoperatively, which leads to a disproportionate number of women with benign ovarian tumors being referred to specialized centers, and conversely women with ovarian malignancy being inappropriately operated in non-specialized centers .
Ultrasonography (US) is currently considered as the primary imaging modality for the detection and characterization of adnexal masses . Despite the progress in its diagnostic capability, there is a high false-positive rate (24%) reported by a large multicenter study that could be explained by dependence on operator experience and a transmission problem of sonographic information from the sonographer to the clinician who makes a final decision [5, 6].
Several studies have proposed for the characterization of the ovarian masses, including examiner’s subjective impression , mathematically developed scoring systems , simple descriptive scoring systems , logistic regression models , and neural networks . The subjective impression of an experienced radiologist is currently considered to be superior to other methods [4, 12], but its subjective nature affects the performance of the method and the examiner’s confidence in providing a diagnosis .
The International Ovarian Tumor Analysis (IOTA) consensus applies a standardized nomenclature and definition for all tumor features evaluated by ultrasound that improve the characterization of adnexal masses . However, there is still significant variation in the ultrasound reporting for adnexal masses that can be confusing for clinicians . In 2009, Amor et al. proposed a unified and structured language for an ultrasonographic report of adnexal masses similar to that used for a breast ultrasound (BI-RADS) called Gynecology Imaging Reporting and Data System (GI-RADS) . This system is based on pattern recognition analysis  and prior risk estimation of the probability of malignancy, based on previous studies . GI-RADS was developed to facilitate communication between radiologists and referring clinicians aiming to reduce the confusion and to help predict the probability of malignancy, thereby improving and individualizing treatment options. A prospective multicenter study of GI-RADS was published in 2011; that study reported that GI-RADS is effective in the prediction of malignancy in adnexal masses and in clinical decision making among patients from Spain and Chile; however, these results still required verification in other countries .
In this study, we apply the GI-RADS classification in ovarian lesions aiming to assess its validity and accuracy in the prediction of malignancy and triaging the management protocol in our locality.
This was a prospective cross-sectional study comprising 116 out of 300 women who are suspected of having ovarian lesions on the basis of previous examination, previous clinical or US examination by the obstetric and gynecological surgeon, accidentally discovered ovarian lesion on US examination by non-radiologists, ovarian lesion on computed tomography, high CA125, and clinical symptoms of ovarian lesions such as pelvic pain and back pain, between March 2017 to August 2018. The sample size was calculated using Open Epi software program, version 23.1. A total of 184 patients were excluded from our study when they underwent neoadjuvant chemotherapy before US examination (n = 50), had previous surgery on the ovary (n = 47), and had no pathological reports after surgery (n = 83). This study was approved by our institutional review board. Written informed consent for participation and publication was obtained from each patient after receiving information about the details of the study. Confidentiality of patient’s records was assured and maintained throughout the study. The trial was registered in the US National Library of Medicine, under clinical trial number NCT03175991.
All patients’ pelvises were examined by using the ProSound Alpha 7 ultrasound (Hitachi Aloka Medical America, Inc. Germany) by transvaginal ultrasound in lithotomy position using endovaginal transducer and/or transabdominal ultrasound in the supine position using a 3.75-MHz sector transducer, in transverse and longitudinal plane and evaluated by B-mode ultrasonography, color, and spectral Doppler. Two expert examiners (G.S.S and L.M.R.K), with more than 10 years’ experience in gynecological ultrasound, performed all examinations and stored between one and four representative images on the database.
Sonographic data analysis
A morphologic evaluation was performed according to the International Ovarian Tumor Analysis Group (IOTA) recommendations for the following parameters: wall thickness, septation, papillary projections, presence and echogenicity of solid areas, presence of mixed component, cystic component, and presence of ascites , and intra-abdominal metastases (peritoneal deposits, liver metastasis, and malignant abdominal lymphadenopathy) was also recorded. Pattern recognition analysis was also used for ovarian masses .
Then, the lesion volume was calculated according to the prolate ellipsoid formula (length × width × height × 0.523, expressed in cubic centimeters). Ten cubic centimeters for postmenopausal women and 20 cm3 for premenopausal women were considered as a cutoff point between the normal and suspicious ovarian lesion . This feature was not taken into consideration for assigning a GI-RADS classification.
After the morphologic evaluation was performed, the color Doppler was activated to identify vascular color signals within the tumor with no aliasing. A tumor was considered to have no flow when no signal could be detected. If blood flow was detected, it was stated as “peripheral” (color signals in the tumor wall or periphery of a solid tumor) or “central” (blood flow detected in septa, papillary projections, solid areas, or the central part of a solid tumor). The central blood flow was used for analysis when there was both peripheral and central blood flow. The evaluation of the amount of flow was subjective and stated as scanty, moderate, or abundant . Then, the pulsed Doppler was activated at the lowest pulse repetition frequency to calculate the resistive index (RI) and pulsatility index (PI). The lowest RI was used for analysis when there is more than one vessel. Morphological features considered suspicious of malignancy included thick wall ≥ 3 mm, thick septum ≥ 3 mm, solid papillary projection, solid and mixed component, presence of ascites, intra-abdominal metastasis, and central blood flow.
After the examinations, a combination of morphological features, color and spectral Doppler features, and then the lesion was evaluated according to GI-RADS classification, and the suggested management protocol was based on the risk of malignancy  as follows:
GI-RADS 1 patients did not undergo follow-up on the basis that these lesions are considered to be normal.
GI-RADS 2 patients were treated expectantly and underwent follow-up after 6 weeks on the basis that these lesions were functional.
GI-RADS 3 patients that did not resolve on follow-up by the radiologists on the basis that these lesions are most probably benign and underwent laparoscopic removal of the lesion.
GI-RADS 4 and 5 patients were referred to gynecological oncologists and surgeon for surgical removal on the basis that these lesions were very probably malignant, taking into consideration that the diagnosis of GI-RADS 4 depends on the presence of one or two sonographic findings suggestive of malignancy and three or more sonographic findings suggestive of malignancy in GI-RADS 5.
Finally, the referral to surgery and decision-making was consulted in accordance with a multidisciplinary team meeting (MDT). A definitive histopathological diagnosis was obtained as a gold standard test for all patients with GI-RADS 4 and 5 and 15 cases of GI-RADS 3 patients after laparoscopic or surgical removal of the masses. Resolution of the lesions on follow-up was considered as a gold standard test for all patients with GI-RADS 2 and 21 patients with GI-RADS 3.
A histopathological examination of all the surgical specimens was done. Tissue sections with formalin fixed and paraffin processed were stained with hematoxylin and eosin. Tumors were classified according to the WHO criteria . Borderline tumors were considered as malignant for analytic purposes.
Data was collected and analyzed using SPSS (Statistical Package for the Social Science, version 20, IBM, and Armonk, New York). Continuous data were expressed in the form of mean ± SD or median (range) while nominal data were expressed in the form of frequency (percentage). Categorical variables were compared using the chi-square test, and tumor volumes were compared using the Mann–Whitney U test. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (LR+), and negative likelihood ratio (LR−) of the GI-RADS system for identifying ovarian masses at high risk of malignancy were calculated using the receiver operating characteristic (ROC) curve. For interrater reliability testing between sonographic findings and histopathological data, Cohen’s kappa (κ) test was used and the result was interpreted as perfect agreement when its value lies between 0.81 and 1.00.
One hundred and fifty-six ovarian lesions were detected in the examined 116 women, 68 women (68.6%) were pre-menopausal, and 48 women (41.4%) were post-menopausal. Their mean age were 42 ± 16.16 years, range 10–82 years. Malignant tumors were more frequent among postmenopausal women than premenopausal women P < 0.001. Additionally, patients with malignant tumors had larger tumor volume, non-hyperechoic solid and mixed component, thick internal septation, ascites, and intra-abdominal metastases more than patients with benign tumors (P < 0.001).
The final diagnosis of the examined ovarian lesions with their corresponding GI-RADS scoring
All GI-RADS classification rates were demonstrated in Fig. 1. Most referring clinicians managed their patients according to their GI-RADS classification. No further follow-up was done in GI-RADS 1 patients, and a final diagnosis of all GI-RADS 2 ovarian masses such as functional cyst, hemorrhagic cysts, corpus luteal cysts, and 21 cases of GI-RADS 3 masses as simple cysts (Fig. 2) were made by spontaneous resolution of these masses at follow-up after 6 weeks as summarized in Table 1. Fifteen cases of GI-RADS 3 as mature teratoma, serous and mucinous cystadenoma, endometrioma, and ovarian torsion and all GI-RADS 4 and 5 (Fig. 3) underwent laparoscopic or surgical removal and a histopathological examination.
There was a strong agreement between the GI-RADS diagnosis and the final diagnosis as its kappa value was 0.91. The GI-RADS classifications in our studied lesions when compared with the gold standard test that is specific for each category demonstrated that among 103 benign ovarian lesions and normal ovaries, 100 lesions (97.1%) were diagnosed as GI-RADS 1, 2, and 3 by US, and this diagnosis was proved pathologically in 15 cases of GI-RADS 3. The missed three masses were classified as GI-RADS 4 (false positive), but the histopathological examination diagnosed them as serous cystadenoma = 1 case and mucinous cystadenoma = 2 cases (Table 1 and Fig. 4). This could be explained by the presence of echogenic locules that is misdiagnosed as a solid element, falsely indicated malignancy in this benign neoplasm in addition to the presence of ascites. Furthermore, 52 (98.1%) masses out of 53 malignant masses had GI-RADS 4 and 5. There was one false-negative mass was classified as GI-RADS 3 (Table 1); this case was a 63-years-old woman with the feature of mature teratoma, but histopathological examination proved it to be early-stage immature teratoma which was a rare entity in postmenopausal women (Fig. 5).
The diagnostic performance of the GI-RADS in predicting the risk of malignancy in ovarian masses
The AUC of the diagnostic performance of the GI-RADS in predicting the malignant ovarian masses was 0.96, and it was highly significant, with P value < 0.002 as summarized in Table 2.
Differentiation between benign and malignant ovarian masses is a common problem in clinical practice. Sonography is considered the first-line imaging modality used for this purpose, and it has been shown to be useful for determining optimal treatment . Many radiologists use pattern recognition approach , others use the scoring system , and considerable efforts have been made by some authors in the characterization of the ovarian masses . However, sometimes, the sonographic reports are misleading and confusing for the clinician . As a matter of fact, the decision of clinical management is based on the data provided in the sonographic report. Consequently, a strategy that provides a structured reporting system of the ovarian masses called GI-RADS, which is based on the concept developed for breast imaging (the BI-RADS classification), has been advised recently. As for BI-RADS, the lexicon of our new system is intended to provide a unified language for ultrasound reporting that improves the communication between the radiologists and clinicians, and recommendations for patient’s management . In this study, we assessed the validity and accuracy of our GI-RADS reporting system for ultrasound evaluation of ovarian masses in the prediction of malignancy and clinical decision-making preoperatively.
Description of the ovarian masses on this system depends on the basis of using the pattern recognition approach and a priori risk for malignancy in each group. On this basis, the GI-RADS classification helps the radiologists to give the clinician as much information as possible in a summarized way, as well as an estimated risk of malignancy, based only on the sonographic features of the lesions . To the best of our knowledge, this is the first standardized reporting system applicable to ovarian masses in our locality.
The GI-RADS classification in our study performed well as a diagnostic tool for prediction of malignancy in ovarian masses as it reported high sensitivity, specificity, and accuracy. This is not a surprising result as the sonographic evaluation of the ovarian masses in this study is based on the IOTA criteria, which have been tested in several multicenter studies and shown to be good criteria that can be used in the discrimination between benign and malignant adnexal masses .
Furthermore, PPV and NPV were high, and these values are not affected by disease prevalence in our study, as there is one selection bias in our study which is the relatively high prevalence of normal ovary and benign tumors. Amor et al.  reported similar high sensitivity, specificity, accuracy, LR−, PPV, and NPV, but their LR+ values were lower than in our study; the difference in this result between the two studies may be due to a large number of studied lesions in Amor et al.  as it was 432 because it is a multicenter study. Also, Amor et al.  used the bilaterality as a parameter in the evaluation of the ovarian masses, but we did not use it, and we use the presence of mixed component, intra-abdominal metastasis and a measurement of PI in our analysis algorithm in addition to the previously mentioned parameters in which both studies were similar in it. It is noteworthy that the results of the study done by Amor et al.  show nearly the same results as ours, as regards the high sensitivity, specificity, PPV, and NPV, but differ in the LR+ and LR− as they were higher in Amor et al. . This could be explained by the difference in the number of studied lesions. In contrary to our study, Migda et al.  reported low sensitivity and high specificity (66.0 and 93.8%, respectively) for GI-RADS when it added to the CA-125 marker, but it showed higher sensitivity and lowest specificity for GI-RADS 4 and 5 (94.3 and 72.2%, respectively).
There is a strong agreement found between GI-RADS classification and the final diagnosis in our study, its kappa value was 0.91. Therefore, the GI-RADS classification system was a useful tool used for identifying malignant ovarian masses, triaging the management, and making clinical decisions as it can detect the ovarian masses at high risk of malignancy. Consequently, ovarian malignancy is appropriately operated in a specialized center. This was similar to the results of previous studies [23, 24].
There were some limitations to our study. First, the performance of all ultrasound by expert radiologists affect diagnostic performance; therefore, further research into how this reporting system performs when used by non-expert radiologists is needed. Second, there is a relatively small sample size relative to the previous researches. Third, there is a high prevalence of normal ovary and benign ovarian masses because our selection criteria depend on all the referral patients to the radiodiagnosis department who are suspected of having an ovarian mass. Fourth, MRI was not applied in our study to compare their diagnostic accuracy with transvaginal US in the evaluation of GI-RADS; we recommended another study to compare between these two modalities. A further weakness is that our study was done in ovarian masses only not in adnexal masses.
The strength of this study is the meticulous prospective recording of the ultrasound data based on IOTA and recognition pattern on GI-RADS reports that leads to high sensitivity and specificity.
In conclusion, this prospective study demonstrated that GI-RADS classification performs well and valid as a reporting system of the ovarian masses with high diagnostic performance in prediction of malignancy, and it seems to be a helpful tool in triaging patient’s management and clinical decision making. The goal of the GI-RADS classification should be explained to the referring clinicians before the application of the treatment as it can improve patient care.
Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Area under receiver operating curve
Gynecology imaging reporting and data system
International Ovarian Tumor Analysis
Negative likelihood ratio
Positive likelihood ratio
Negative predictive value
Positive predictive value
Receiver operating characteristic
Jemal A, Tiwari RC, Murray T, Ghafoor A, Samuels A, Ward E et al (2004) Cancer statistics. CA: a Cancer Journal for Clincians. 54(1):8–29
A.S. Ibrahim, I.A. Seif-Eldin, K. Ismail, et al (2007) Cancer in Egypt, Gharbiah: Triennial Report of 2000-2002, Gharbiah. Population-based Cancer Registry, Middle East Cancer Consortium, Cairo, Egypt.
Pitta Dda R, Sarian Lo Fau-Barreta A, Fau-Campos EA BA, LldA CEF-A, AMD ALF-F, Fachini Am Fau-Campbell LM, Campbell LmFau-Derchain S, Derchain S (2013) Symptoms, CA125 and HE4 for the preoperative prediction of ovarian malignancy in Brazilian women with ovarian masses. BMC Cancer. 18(13):423. https://doi.org/10.1186/1471-2407-13-423
Valentin L, Hagen B, Tingulstad S, Erik-Ness S (2001) Comparison of “pattern recognition” and logistic regression models for discrimination between benign and malignant pelvic masses: a prospective cross-validation. Ultrasound Obstet Gynecol. 18(4):357–365
Timmerman D, Testa AC, Bourne T, Ferrazzi E, Ameye L, Konstantinovic ML et al (2005) Logistic regression model to distinguish between the benign and malignant adnexal mass before surgery: a multicenter study by the International Ovarian Tumor Analysis Group. J Clin Oncol. 23(34):8794–8801
Yazbek J, Raju SK, Ben-Nagi J, Holland TK, Hillaby K, Jurkovic D (2008) Effect of quality of gynecological ultrasonography on management of patients with suspected ovarian cancer: a randomized controlled trial. The Lancet Oncology. 9(2):124–131
Timor-Tritsch IE, Goldstein SR (2005) The complexity of a “complex mass” and the simplicity of a “simple cyst”. J Ultrasound Med. 24(3):255–258
Valentin L (1999) Pattern recognition of pelvic masses by gray-scale ultrasound imaging the contribution of Doppler ultrasound. Ultrasound Obstet Gynecol. 14(5):338–347
Alcázar JL, Mercé LT, Laparte C, Jurado M, López-García G (2003) A new scoring system to differentiate benign from malignant adnexal masses. Am J Obstet Gynecol. 188(3):685–692
Granberg S, Wikland M, Jansson I (1989) Macroscopic characterization of ovarian tumors and the relation to the histological diagnosis: criteria to be used for ultrasound evaluation. Gynecol Oncol. 35(2):139–144
Alcazar JL, Errasti T, Laparte C, Jurado M, Lopez-Garcia G (2001) Assessment of a new logistic model in the preoperative evaluation of adnexal masses. J Ultrasound Med. 20(8):841–848
Timmerman D, Verrelst H, Bourne TH, De Moor B, Collins WP, Vergote I et al (1999) Artificial neural network models for the preoperative discrimination between malignant and benign adnexal masses. Ultrasound Obstet Gynecol. 13(1):17–25
Timmerman D (2004) The use of mathematical models to evaluate pelvic masses; can they beat an expert operator? Best Pract Res Clin Obstet Gynaecol. 18(1):91–104
Yazbek J, Ameye L, Testa AC, Valentin L, Timmerman D, Holland TK et al (2010) Confidence of expert ultrasound operators in making a diagnosis of adnexal tumor: effect on diagnostic accuracy and interobserver agreement. Ultrasound in obstetrics & gynecology: the official journal of the International Society of Ultrasound in Obstetrics and Gynecology. 35(1):89–93
Timmerman D, Valentin L, Bourne TH, Collins WPVH (2000) Terms, definitions and measurements to describe the sonographic features of adnexal tumors: a consensus opinion from the international ovarian tumor analysis (IOTA)group. Ultrasound Obstet Gynecol. 16:500–505
Amor F, Vaccaro H, Alcazar JL, Leon M, Craig JM, Martinez J (2009) Gynecologic imaging reporting, and data system: a new proposal for classifying adnexal masses based on sonographic findings. J Ultrasound Med. 28(3):285–291
Le T, Fayadh RA, Menard C, Hicks-Boucher W, Faught W, Hopkins L et al (2008) Variations in ultrasound reporting on patients referred for investigation of ovarian masses. J Obstet Gynaecol Canada. 30(10):902–906
Amor F, Alcázar JL, Vaccaro H, León M, Iturra A (2011) GI-RADS reporting system for ultrasound evaluation of adnexal masses in clinical practice: a prospective multicenter study. Ultrasound Obstet Gynecol. 38(4):450–455
Prorok PC, Andriole GL, Bresalier RS et al (2000) Design of the prostate, lung, colorectal, and ovarian (PLCO) cancer screening trial. Control Clin Trials. 21(6 Suppl):273S–309S
Kurman RJ, Carcangiu ML, Herrington CS, Young RH (2014) World Health Organization classification of tumors of female reproductive organs. Lyon: IARC Press; str. 15-40
Alcázar JL, Royo P, Jurado M, Mínguez JÁ, García-Manero M, Laparte C et al (2008) Triage for surgical management of ovarian tumors in asymptomatic women: assessment of an ultrasound-based scoring system. Ultrasound Obstet Gynecol. 32(2):220–225
Brown DL, Doubilet PM, Miller FH, Frates MC, Laing FC, DiSalvo DN et al (1998) Benign and malignant ovarian masses: selection of the most discriminating gray-scale and Doppler sonographic features. Radiology. 208(1):103–110
Migda M, Bartosz M, Migda MS, Kierszk M, Katarzyna G, Maleńczyk M (2018) Diagnostic value of the gynecology imaging reporting and data system (GI-RADS) with the ovarian malignancy marker CA-125 in preoperative adnexal tumor assessment. J Ovarian Res. 11(1):92. https://doi.org/10.1186/s13048-018-0465-1
Zhang T, Li F, Liu J, Zhang S (2017) Diagnostic performance of the gynecology imaging reporting and data system for malignant adnexal masses. Int J Gynaecol Obstet. 137(3):325–331
The authors state that this work has not received any funding.
Ethics approval and consent to participate
This study was approved by the Reasearch Ethics Committee of the Faculty of Medicine at Assiut University in Egypt on February 28, 2017, and its number is 17100016. Written informed consent was obtained from all patients to participate in this study.
Consent for publication
All patients included in this research gave written informed consent to publish the data contained within this study.
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Khalaf, L.M.R., Desoky, H.H.M., Seifeldein, G.S. et al. The diagnostic efficacy of Gynecology Imaging Reporting and Data System (GI-RADS): single-center prospective cross-sectional study. Egypt J Radiol Nucl Med 50, 61 (2019). https://doi.org/10.1186/s43055-019-0071-2