Obstetrics and Gynaecology Section DOI : 10.7860/JCDR/2017/28533.10614
Year : 2017 | Month : Sep | Volume : 11 | Issue : 9 Page : QC12 - QC15

Role of Sonographic Gray-Scale Pattern Recognition in the Diagnosis of Adnexal Masses

Jyothi Shetty1, Grishma Reddy2, Deeksha Pandey3

1 Professor, Department of Obstetrics and Gynaecology, Kasturba Medical College, Manipal Academy of Higher Education, Manipal, Karnataka, India.
2 Postgraduate Student, Department of Obstetrics and Gynaecology, Kasturba Medical College, Manipal Academy of Higher Education, Manipal, Karnataka, India.
3 Associate Professor, Department of Obstetrics and Gynaecology, Kasturba Medical College, Manipal Academy of Higher Education, Manipal, Karnataka, India.


NAME, ADDRESS, E-MAIL ID OF THE CORRESPONDING AUTHOR: Dr. Deeksha Pandey, Associate Professor, Department of Obstetrics and Gynaecology, 4/1 Kasturba Medical College,Manipal Academy of Higher Education, Manipal-576104, Karnataka, India.
E-mail: deekshiiiobg@gmail.com
Abstract

Introduction

Characterization of adnexal masses as benign or malignant is of utmost importance for optimal management and prognostication. Ultrasound examination plays an important role in the differentiation of adnexal masses. Various sonographic characteristics have been recognised to differentiate benign and malignant adnexal masses. Subjective evaluation of gray-scale ultrasound images by an experienced ultrasound examiner to discriminate adnexal masses is known as “pattern recognition”.

Aim

To access the efficacy of pattern recognition at predicting an accurate histological diagnosis of adnexal masses.

Materials and Methods

All adnexal masses diagnosed clinically or during screening sonography were included in the study (n=136). Sonographic pattern recognition was performed and documented with specific diagnosis whenever feasible. Risk of Malignancy Index 3 (RMI3) score was also calculated. Results were compared with the gold standard histology. Chi-square test was used to assess the significance of the results and a p-value <0.05 was considered statistically significant.

Results

In the final cohort of 136 women, on pattern recognition, 91 were suspected to have benign adnexal masses and 45 were reported as malignant adnexal masses. However, on final histo-pathology, 94 patients had benign tumours and 42 patients had malignant disease. The benign group pattern recognition could render a specific diagnosis in 85.7% as compared to RMI3 pattern recognition conferred a sensitivity of 95.2% (RMI3 78.6%), with a slight compromise in the specificity (94.7% versus 96.8%).

Conclusion

Pattern recognition is a sensitive and specific sonographic tool in discriminating benign and malignant adnexal masses. Moreover, it is also useful in differentiating various benign adnexal masses.

Introduction

Adnexal masses are one of the common challenges faced by a gynaecologist in day to day clinical practice. Characterization of these as benign or malignant is of utmost importance for optimal management and prognostication. Malignancy of the ovary is notoriously known for getting missed or misdiagnosed in early stages. With more than 225,000 new cases annually leading to 140,000 deaths, timely diagnosis of ovarian cancer remains a major health concern worldwide as well as a challenge [1,2].

The subjective evaluation of gray-scale ultrasound images by an experienced ultrasound examiner to discriminate adnexal masses is known as “pattern recognition”. Pattern recognition can help to discriminate an adnexal mass as benign or malignant and at times to recognize a specific diagnosis as endometrioma, dermoid, mucinous cystadenoma, fibroma and likewise. This pattern recognition has been shown to accurately make a specific diagnosis in almost half of adnexal masses preoperatively [3]. The role of CA125 alone or in various combinations with sonographic tools has also been studied. Risk of Malignancy Index (RMI) which is scoring system, was developed to differentiate benign and malignant ovarian tumours, so as to facilitate timely management/referral. One of the most popular versions of this index is known as RMI3 [4]. Three-dimensional power Doppler is another modality which has been investigated in some studies [5].

As compared to the above mentioned modalities and scores, pattern recognition is a simple, cost effective gray-scale ultrasound based tool. It has recently attracted many investigators for triaging adnexal masses. It has been studied with varied combinations of other screening tools like RMI score or CA125 [6]. Therefore, this study was carried out to access the efficacy of pattern recognition alone at predicting an accurate histological diagnosis of adnexal masses.

MATERIAL AND METHODS

This prospective cohort study was carried out in a tertiary care center over a period of two years between January 2011-January 2013 on approval by Institutional Ethical Committee (IEC192/2011). All adnexal masses which were either diagnosed by history and examination or were incidentally diagnosed during preliminary sonography and now were planned for histological confirmation of diagnosis either following surgical removal or by biopsy were included in the study. All enrolled women provided written informed consent for participation. The investigator involved in pattern recognition during sonographic examination was not blinded to the history, clinical finding or any sonographic/laboratory work up done previously, to minimize the bias. Non gynaecological tumours and borderline ovarian tumour were excluded from the final analysis.

Sample Size Calculation

Anticipating 86% sensitivity for RMI with the precision of 10% at 95% confidence level a minimum of 46 malignant and 92 benign cases are required for this study [7].

A) Ultrasonography (pattern recognition): All sonographic examinations were performed with a transabdominal (5 MHz) and/or transvaginal (7.5 MHz) probes using Toshiba Nemio machine. Benign and malignant tumours manifest a characteristic appearance at gray-scale imaging known as “pattern recognition”. Based on the subjective evaluation of the gray-scale ultrasound images by a single observer, the sonographer classified each tumour as probably benign or malignant. Whereever feasible, a specific diagnosis such as endometrioma, dermoid, paraovarian cyst, simple cyst, peritoneal inclusion cyst or hydrosalpinx was made [8].

B) Risk of malignancy index 3 (RMI3): This was calculated by multiplying the ultrasound score (U), menopausal status (M) and serum CA125 level {RMI 3 = U × M × CA125} [4]. If the RMI3≥200, the diagnosis is suggestive of malignant ovarian tumour and if RMI3 is <200, it is diagnosed as benign tumour.

Statistical Analysis

The statistical analysis was done using SPSS version 22.0.0.0. The Pearson Chi-square test was applied to test the significance of differences of age, menopausal status and ultrasound score. The Mann-whitney U-test was applied when testing differences in distribution of CA125 among women with benign and malignant adnexal masses. A probability (p) value of <0.05 was considered significant.

Results

A total of 140 women with an adnexal mass were enrolled. Four patients were excluded from the final analysis, as three of those had borderline pathology and one was diagnosed to have Gastro-Intestinal Stromal Tumour (GIST). Thus, the final study population included 136 women with adnexal mass. On pattern recognition, 91 patients were suspected to have benign adnexal masses and 45 patients were reported as malignant adnexal masses. However, on final histopathology, 94 patients had benign tumours and 42 patients had malignant disease. As histopatholgy is the ‘goldstandard’ to differentiate between benign and malignant pathology, analysis was performed and comparisons were done based on the final report.

Mean age of the study population (n=136) was 40.5 years. Mean age for malignant group is 46.5 years and benign group was 39.3 years. Nine patients in malignant group and 24 patients in benign group were nulliparous. Around 60% women in malignant group as compared to 25% in the benign group had attained menopause. The differences in age and menopausal status were statistically significant, as expected [Table/Fig-1].

Demographic characteristics of the study population.

CharacteristicsTotal no. of patients n=136(%)Benign group n=94(%)Malignant group n=42(%)p-value
Mean age (range) in years41.5 (14-75)39.3 (16-63)46.5 (14-75)0.04
Median age423648
ParityNulliparous33 (24.3)24 (25.5)9 (21.4)0.09
Parous103 (75.7)70 (74.5)33 (78.6)
Menopausal statusPre menopausal88 (64.7)71 (75.5)17 (40.5)<0.001
Post menopausal48 (35.3)23 (24.5)25 (59.5)

Pattern Recognition

Among the 91 tumours recognised as benign by pattern recognition, a specific diagnosis was suggested in 85.7% (78/91). The proposed diagnosis was mucinous cystadenoma (19), serous cystadenoma (12), endometriotic cysts (17), dermoid cysts (12), fibroma (4), paraovarian cyst (4), peritoneal inclusion cyst (5), haemorrhagic cyst (2) and hydrosalpinx (3). Sixty eight (74.7%) of the proposed 78 diagnoses based on gray-scale imaging were correct [Table/Fig-2]. The incorrect diagnosis of three mucinous cystadenoma included two serous cystadenoma and another simple cyst. The incorrect diagnosis of four serous cystadenoma was of four mucinous cystadenomas. The incorrect diagnosis of one endometriotic cyst was of a granulosa cell tumour, the incorrect diagnosis of two haemorrhagic cysts comprised one each of peritoneal inclusion cyst and granulosa cell tumour. In 13 tumours, which were considered as benign no specific diagnosis could be ascertained.

Diagnosis of benign adnexal masses (n=91) based on pattern recognition compared with histology.

Ultrasound characteristicsNumber n = 91 (%)USG specific diagnosisHistological diagnosis
Multiloculated ovarian cyst26 (28.5)Mucinous cystadenoma (19)Mucinous cystadenoma (16)Serous cystadenoma (2)Simple cyst (1)
No specific diagnosis (7)Infarcted ovarian cyst (2)Chronic oophoritis (2)Pyosalpinx (1)Torsion ovary (1)Mucinous cystadenoma (1)
Ground glass appearance17 (18.7)Endometriotic cysts (17)Endometriotic cysts (16)Granulosa cell tumour(1)
Clear cyst13 (14.3)Serous cystadenoma (12)Serous cystadenoma (8) Mucinous cystadenoma (4)
No specific diagnosis (1)Torsion ovary (1)
Solid tumours9 (9.4)Fibroma (4)Fibroma (4)
No specific diagnosis (5)Fibroma (1)Haemorrhagic cyst (2)Ruptured corpus luteal cyst (2)
Cysts with bright echoes with posterior acoustic shadowing7 (7.7)Dermoid cysts (7)Dermoid cyst (7)
Cysts with hyperechoic lines5 (5.5)Dermoid cysts (5)Dermoid cyst (5)
Spider in a web appearance5(5.5)Peritoneal inclusion cyst (5)Peritoneal inclusion cyst (5)
Unilocular cyst with normal ovary adjacent to the cyst4(4.4)Paraovarian cyst (4)Paraovarian cyst (4)
Reticular lacy network2 (2.2)Haemorrhagic cyst (2)Peritoneal inclusion cyst (1)Granulosa cell tumour (1)
Unilocular with incomplete septation3 (3.3)Hydrosalpinx (3)Hydrosalpinx (3)

Malignancy was suspected in 45 tumours with pattern recognition. Though five of these were benign by histopathology. In 45 malignant masses, a specific diagnosis was proposed in 35% (16/45) tumours [Table/Fig-3]. Out of those three proved to be wrong. The included two mucinous cystadenocarcinoma which turned out to be serous cystadenocarcinoma and fallopian tube carcinoma. One which was thought as borderline tumour was found to be endometrioma.

Diagnosis of malignant adnexal masses (n=45) based on pattern recognition compared with histology.

Ultrasound characteristicsNumber n = 45 (%)USG specific diagnosisistological diagnosis
Multilocular with solid area11 (24.4)Mucinous cystadeno- carcinoma (4)Mucinous cystadenocarcinoma (2)Serous cystadenocarcinoma (1)Fallopian tube carcinoma (1)
No specific diagnosis (6)Serous cystadenocarcinoma (6)
Borderline tumour (1)Endometriotic cyst (1)
Unilocular cyst with solid area which has irregular borders10 (22.2)No specific Diagnosis (10)Clear cell tumours (3)Mucinous cystadenocarcinoma (1)Serous cystadenoma* (2)Serous cystadenocarcinoma (1)3 - Endometrioid tumour (3)
Unilocular, solid with papillae4 (8.8)Serous cystadeno-carcinoma (4)Endometrioid tumour (1)Serous cystadenocarcinoma (3)
Multilocular, solid area with papillae5 (11.1)No specific diagnosis (5)Serous cystadenocarcinoma (5)
Solid tumours15 (33.3)Germ cell tumour (6)Dysgerminoma (2)Yolk sac tumour (2)Mixed germ cell tumour (1)Granulosa cell tumour (1)
Sexcord stromal tumour (1)Sertoli leydig cell tumour (1)
No specific diagnosis (8)Endometrioid tumour (4)Serous cystadeno carcinoma (1)Steroid cell tumour (1)Fibroma (2)

RMI – Risk of Malignancy Index.


Pattern recognition suspected five benign cases falsely as malignant. It misdiagnosed two fibromas as malignant due to solid appearance. Two serous adenomas were misdiagnosed as malignant due to their solid papillary projections. Pattern recognition has identified two malignancies falsely as benign. Two granulosa cell tumours were misdiagnosed as benign due to their appearance as cyst with haemorrhage and ground glass appearance.

RMI3

The median RMI was 31 (Range: 6-1500) for benign tumours as compared to 943 (Range: 9-35,127) in malignant tumours. False positive RMI (RMI ≥200) was seen in three benign cases (3.2%). False negative RMI (RMI<200) was found in two cases of malignant ovarian tumour.

[Table/Fig-4] summaries the comparison of pattern recognition and RMI3 in the preoperative diagnosis of benign and malignant ovarian tumours.

Comparison of RMI 3 versus pattern recognition.

Final Diagnosed as per histopathologyMalignant (42)Benign (94)Test efficacy
RMISensitivity: 78.6%Specificity: 96.8%PPV: 91.7%NPV: 91%
RMI ≥200True malignant-33False malignant-3
RMI<200False benign-9True benign-91
Pattern recognitionSensitivity: 95.2%Specificity: 94.7%PPV: 88.9%NPV: 97.8%
Benign patternFalse benign-2True benign-89
Malignant patternTrue malignant-40False malignant-5

Discussion

Present study proves that the overall efficacy of pattern recognition is superior to RMI3 in differentiating benign and malignant adnexal masses. Pattern recognition not only differentiates benign and malignant tumours, it also confers a specific pathologic diagnosis. Specific diagnosis might not be much useful for malignant tumours as most of the times treatment is radical. However, it has big significance in differentiating various benign tumours and optimizing the management [3].

A recent study even tried to find out the sensitivity of pattern recognition in the specific diagnosis of benign adnexal masses [9,10]. They found that the highest sensitivities were achieved with simple cysts (100%) and hydrosalpinx (100%). This was followed by mature teratomas (88%), endometriomas (75%), ovarian fibromas (88%), tubo-ovarian abscesses (88%) and serous cystadenocarcinomas (82%). Serous cystadenomas were misdiagnosed most commonly (40.5%). In our study, however five cases of serous cystadeno-carcinoma having sonographic pattern of ‘multilocular, solid area with papillae’ could not be diagnosed with confidence. The sensitivity of subjective assessment in diagnosing adnexal torsion in the aforementioned study was 54%. Similarly in our study too, both the cases of adnexal torsion were not given any specific diagnosis.

On excluding the ten cases of non epithelial origin from our study sample, the sensitivity of RMI3 increased from 78.6% to 94% with a specificity of 96.8%. Thus, it can be inferred that the efficacy of RMI3 is equal to pattern recognition in identifying epithelial ovarian cancers.

More or less similar efficacy has been reported for RMI3 in other recent studies too (sensitivity of 85%, a specificity of 91%, a PPV of 60%, and a NPV of 97.8%) [11-13]. Pattern recognition confers a sensitivity of 95.2% as compared to 78.6% by RMI3, with a slight compromise in the specificity (94.7% versus 96.8%). The reason for a lower sensitivity of RMI3 may be due to the fact that CA125 is usually not much elevated in stage I disease, which is a determinant in RMI3 index calculation [14]. Another interesting observation was made by Lennox GK et al., that the scoring systems for triaging ovarian malignancies do not perform well in patients with stage I disease where endometriod and clear cell histologies predominate [12]. Similar trend was seen in our study too where RMI score was not found to be similarly efficient for early stage or non epithelial ovarian cancer. We realized that six of the nine malignant cases which were misdiagnosed as benign by RMI3, were non epithelial in origin. So in case if we have suspicion by sonography and RMI is not matching, it is worthy to test for other tumour markers and then replacing CA125 in RMI with the specific tumour marker [Table/Fig-5].

RMI score incorporating specific tumour marker instead of CA125. HPE- Histopathological evaluation, LDH- Lactate dehydrogenase, AFP- Alpha fetoprotein, U- Ultrasound Score, M- Menopausal Status, RMI- Risk Of Malignancy Index.

Sl. No.HPEMeno-pausal statusSpecific tumour markerU × M × specific tumour markerOriginal RMI
1Stage 3 dysgerminomaPre-menopausalLDH = 4387 IU/L3 (solid tumour with ascites)x1x4387 =13,1613 × 1 × 13.2=39.6
2Stage 1 dysgerminomaPre-menopausalLDH = 2064 IU/L1x1x2064 =20641 × 1 × 11.8=11.8
3Stage 1 yolk sac tumourPre-menopausalAFP=26,957 ng/ml1 × 1 × 26,957 =26,9571 × 1 × 133 =133
4Stage 3 yolk sac tumourPre-menopausalAFP=9423 ng/ml3(solid tumour with ascites)x1x9423 =28,2693 × 1 × 390 =1170
5Stage 3 mixed germ cell tumourPre-menopausalAFP=984 ng/ml3 (solid tumour with ascites) × 1 × 984 = 2,9523 × 1 × 108 =324

Limitation

Though the sample size here is inadequate to propose for the substitution of a specific tumour marker instead of CA125 in non epithelial tumours. Further prospective studies with large number of non epithelial tumours are required to validate this proposal.

Conclusion

Pattern recognition was superior to RMI3 for discriminating benign and malignant adnexal masses. RMI3 was not able to identify epithelial ovarian cancer stage I, germ cell tumours, and sex cord stromal tumours. RMI3 may be useful when experienced sonographer is unavailable.

References

[1]Siegel R, Naishadham D, Jemal A, Cancer statistics 2013 CA Cancer J Clin 2013 63(1):11-30.  [Google Scholar]

[2]Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D, Global cancer statistics CA Cancer J Clin 2011 61(2):69-90.  [Google Scholar]

[3]Valentin L, Pattern recognition of pelvic masses by gray-scale ultrasound imaging:the contribution of Doppler ultrasound Ultrasound in obstetrics & gynecology:the official journal of the International Society of Ultrasound in Obstetrics and Gynecology 1999 14(5):338-47.  [Google Scholar]

[4]Tingulstad S, Hagen B, Skjeldestad FE, Halvorsen T, Nustad K, Onsrud M, The risk-of-malignancy index to evaluate potential ovarian cancers in local hospitals Obstet Gynecol 1999 93(3):448-52.  [Google Scholar]

[5]Utrilla-Layna J, Alcazar JL, Auba M, Laparte C, Olartecoechea B, Errasti T, Performance of three-dimensional power Doppler angiography as third-step assessment in differential diagnosis of adnexal masses Ultrasound Obstet Gynecol 2015 45(5):613-17.  [Google Scholar]

[6]Manegold-Brauer G, Buechel J, Knipprath-Meszaros A, Schoetzau A, Hacker NF, Tercanli S, Improved detection rate of ovarian cancer using a 2-step triage model of the risk of malignancy index and expert sonography in an outpatient screening setting International journal of gynecological cancer:official journal of the International Gynecological Cancer Society 2016 26(6):1062-69.  [Google Scholar]

[7]Akturk E, Karaca RE, Alanbay I, Dede M, Karasahin E, Yenen MC, Comparison of four malignancy risk indices in the detection of malignant ovarian masses Journal of Gynecologic Oncology 2011 22(3):177-82.  [Google Scholar]

[8]Jung SI, Ultrasonography of ovarian masses using a pattern recognition approach Ultrasonography 2015 34(3):173-82.  [Google Scholar]

[9]Sayasneh A, Kaijser J, Preisler J, Smith AA, Raslan F, Johnson S, Accuracy of ultrasonography performed by examiners with varied training and experience in predicting specific pathology of adnexal masses Ultrasound in Obstet Gynecol 2015 45(5):605-12.  [Google Scholar]

[10]Bayoglu Tekin Y, Dede FS, What is the success of ultrasonography of benign adnexal masses? J Obstet Gynaecol Res 2014 40(2):473-78.  [Google Scholar]

[11]Rossi A, Forzano L, Romanello I, Ambrosini G, Iuri V, Marchesoni D, Comparison of Pelvic Masses Score (PMS) and Risk of Malignancy Index (RMI 3) in the evaluation of pelvic masses European Journal of Gynaecological Oncology 2014 35(4):421-24.  [Google Scholar]

[12]Lennox GK, Eiriksson LR, Reade CJ, Leung F, Mojtahedi G, Atenafu EG, Effectiveness of the risk of malignancy index and the risk of ovarian malignancy algorithm in a cohort of women with ovarian cancer:Does histotype and stage matter? International Journal Of Gynecological Cancer: Official Journal Of The International Gynecological Cancer Society 2015 35(4):421-24.  [Google Scholar]

[13]Menon U, Talaat A, Rosenthal AN, Macdonald ND, Jeyerajah AR, Skates SJ, Performance of ultrasound as a second line test to serum CA125 in ovarian cancer screening BJOG:An International Journal of Obstetrics and Gynaecology 2014 7:35-39.  [Google Scholar]

[14]Gupta D, Lis CG, Role of CA125 in predicting ovarian cancer survival- a review of the epidemiological literature Journal of Ovarian Research 2009 9(2):13  [Google Scholar]