Chapter 32: COLOR DOPPLER SONOGRAPHY OF PELVIC MASSES

Transvaginal sonography (TVS) affords detailed assessment of the morphology of pelvic masses; however, the morphology of adnexal masses depicted by TVS is nonspecific at times. Color Doppler sonography (CDS) with spectral analysis has added to our understanding and characterization of these lesions based on its depiction of vascularity of the mass. Using two-dimensional (2D) and/or three-dimensional (3D) CDS, it is possible to demonstrate the origin of the mass and to detect areas of abnormal neovascularity.

Color Doppler sonography depicts the pelvic vessels supplying the ovary and uterus. Spectral analysis correlates a variety of anatomic and physiologic parameters such as:

1. Vessel characteristics such as the presence of a muscular media

2. Certain physiologic processes, such as vasodilation or vasoconstriction

3. Vascular configuration of certain parenchymal beds

4. Inherent interstitial pressure within the tumor

In general, CDS serves as an adjunct to the morphologic assessment of pelvic masses demonstrated on sonography (Figure 32-1). A double-blinded study from our institution showed that CDS added clinically important information in approximately 40% of the patients undergoing TVS for a pelvic mass.¹ A combination of TVS and CDS provides assessment of adnexal masses on a physiologic, rather than on a merely morphologic, basis.^2,3

This chapter discusses and illustrates the technique for performing CDS on patients with pelvic masses. It also outlines the role of CDS in the diagnosis of the adnexal masses, including its differentiation of benign from malignant masses. Limitations in assessment of benign versus malignant histopathology and pitfalls of the examination are presented.^4,5 Special indications including the use of this modality in screening programs and in postmenopausal women who have adnexal masses also are addressed.⁶

The ovary is an ovoid structure that is nestled between the roughly middle positioned uterus and the internal iliac vessels, which are located along the lateral pelvic sidewall. The volume of the ovary changes with the patient's hormonal status and has an average volume of 4.2 cm³ in the adolescent period, 9.8 cm³ in the premenopausal adult, and 5.8 cm³ in the postmenopausal woman.² The formula for calculating the size of the ovary is that of a prolate ellipse that is equal to 0.523 (length × width × height).⁷ The ovary is readily identified in women of child-bearing age morphologically by the presence of small, rounded, anechoic spaces that represent developing follicles or corpora lutea. In the postmenopausal woman, however, only approximately 60% of ovaries can be definitely identified due to their small size and location near the bowel.⁸ Some contain cystic structures up to around 3 cm, which may represent inclusion cysts.

The ovary receives a dual blood supply. One arises from the abdominal aorta and the ovarian artery, which courses toward the ovary in the infundibulopelvic ligament. The other arterial blood supply is derived as an adnexal branch of the uterine artery. Once the "feeding" vessels penetrate the ovarian capsule, intraovarian branches become coiled and tortuous, especially in areas devoid of follicles. When a corpus luteum forms, small blood vessels form in its wall, resulting in a "vascular arcade." The blood flow within this vascular arcade has relatively high diastolic flow and low pulsatility and can appear as a ring of vascularity ("ring of fire") (Figure 32-2).

Understanding this anatomic arrangement is important in optimizing TV-CDS technique for depiction of adnexal masses for several reasons. First, if scanned in the luteal phase of the cycle, many premenopausal women have areas of low-impedance flow within the ovary. Therefore, one should avoid scanning women in this phase, if an evaluation of an adnexal mass is ordered. Second, the relative impedance seen in masses will be related to which vessels are sampled and their proximity to tumor growth.

In premenopausal women, vascular flow within the ovary depends on the presence of a functioning corpus lu-teum and the neovascularity relates to hormonal changes that change during the menstrual cycle. In the men-strual and follicular phases, there is high resistance to diastolic flow. After the formation of the corpus luteum, low-resistance patterns may be identified as a result of the neovascularity of the corpus luteum.^9,10 The ovaries of postmenopausal women are usually quite hypovascular, with only the main arterial branches showing color. Continued research is necessary on the effects of estrogen replacement on ovarian flow as depicted by CDS. As the sensitivity of CDS scanners improves, flow in smaller vessels with lower velocities will be depicted.

Although this chapter focuses on the role of CDS in the assessment of adnexal masses, it is important to appreciate the morphologic characteristics depicted by conventional gray-scale analysis when interpreting CDS. Of all ovarian masses, 80% are benign, 10% to 15% are malignant, and approximately 5% represent metastatic lesions. When evaluated in relation to age, benign lesions predominate in women younger than 40 years, whereas malignant lesions are predominantly seen in women older than 40 years. The ovary is composed histologically of 3 cell types. These are the coelomic epithelial layer, the germ cell layer, and the stromal layer.

Epithelial lesions account for 75% to 80% of all ovarian neoplasms and represent more than 90% of all ovarian malignancies. Serous lesions account for 30% of the total, whereas 20% are mucinous lesions. Other epithelial lesions, including endometrioid carcinoma and Brenner cell tumors, are seen in 14% of ovarian masses. Of malignant epithelial lesions, serous cystadenocarcinoma is the dominant cell type. Overall, germ cell tumors account for approximately 20% of ovarian masses identified. The majority of these are benign teratomas, although a malignant form is seen, especially in younger populations. Stromal or sex cord tumors are often hormonally active and are seen in 10% of ovarian masses. Typically, these tumors demonstrate low-impedance flow probably secondary to their increased perfusion and vasodilation of "feeding" vessels.

Predominantly cystic adnexal masses include benign processes, such as physiologic cysts and endometrio-mas, as well as paratubal masses. Also included in this category are the epithelial ovarian lesions, which includes both benign and malignant cell types. With the resolution of the currently available sonography systems, many benign ovarian masses will show scattered low-level echoes within the mass. These echoes may arise from proteinaceous fluid within the cyst. Complex masses are identified in a variety of conditions. These include benign ovarian processes, such as ovarian teratomas and tubo-ovarian complexes, and endometriomas and hemorrhagic cysts with various stages of clot retraction. In general, most solid adnexal masses will be malignant. They are usually of germ cell or stromal origin. An exception is the Brenner tumor, which is a benign epithelial lesion even though it is solid.

Morphologic features that suggest malignancy include broad-based papillary excrescences or irregular solid areas. Other sonographic findings in metastatic ovarian disease include ascites and omental thickening. These features should be assessed before CDS evaluation because abnormal flow typically is seen in areas of morphol-ogic abnormality.

Diagnosis of ovarian tumor with CDS is based on the detection of the low-impedance, high-velocity flow associated with tumor neovascularity as well as dilation of clustered vessels in a morphologically abnormal area of interest. Tumor microvascularity also tends to have a chaotic arrangement of vessels. Detection of tumor neovascularity by CDS is based on recognition of the tumor vascularity that is required for tumor growth and nutrition. Tumor neovas-cularity involves the generation of new vessels from either normal or abnormal existing ones. Normal arterioles have an inner muscular lining that regulates organ perfusion. In tumor angiogenesis a paucity of this lining is pre-sent. Typically there is a substantial amount of arteriovenous shunting within tumor microvasculature. Increased diastolic flow relative to systolic flow occurs due to a decrease in resistance to forward flow. Normally, this process of neoangiogenesis is self-limited, such as the cyclical changes related to the formation of the corpus luteum or the neovascularity and vasodilatation found in acute inflammation. In tumoral vessels, however, this process is not self-limited and is a requisite for continued tumor growth over 3 to 5 mm.¹¹

Any sonographic evaluation of a pelvic mass begins with its morphologic analysis. The dominant sonographic character of the mass is noted. The mass is characterized as predominantly cystic, complex, or solid. Next, its internal morphology is assessed for the presence of papillary projections and septations, and low-level echoes. Finally, secondary signs of malignancy, such as ascites or pelvic peritoneal implants, are identified.

Similar to characterizing masses by their morphology, CDS requires that multiple parameters need to be evaluated. These include vessel location, relative impedance, velocities, and the presence or absence of a notch in the waveform during diastole.¹² The presence of a diastolic notch is typically seen in larger arterioles that contain a muscular media or at their branch points.

Color Doppler sonography aids in vessel localization and characterization of the vascular pattern. The vessel is localized as peripheral, central, or miscellaneous (ie, within projections, septations). It is clear that the detection of flow is most likely where there is active tumor growth. Spectral analysis is used to characterize the waveform. Both are angle independent. The pulsatility index (PI) takes into account more of the velocities composing the waveform. Pulsatility or resistive indices (RI) are calculated from the waveforms, and, finally, the presence of a diastolic notch is noted.

It is clear that certain non-neoplastic lesions, such as corpora lutea, may demonstrate the neovascularity patterns associated with malignant lesions. Ancillary features may help to differentiate malignant from benign proc-esses. Malignant masses tend to have an increase in vascularity either centrally or within papillary projections. As mentioned previously, the presence of a diastolic notch in the waveform obtained from a vessel within a mass is a nonspecific finding and can be absent in benign conditions that are associated with low-impedance flow, vasodila-tion, or both.¹²

Color Doppler sonography can be performed using either abdominal or vaginal probes. This is determined by the size of the lesion and its location. The vessels are located using CDS. Flow detection is predicated on the presence of a minimal threshold velocity (typically approximately 1 to 2 cm/s) and an optimal angle of interroga-tion. The color image is superimposed on the gray-scale image. After the vessels are localized, spectral analysis is performed. When interrogating vessels using spectral analysis, the entire volume of large vessels should be sampled. Spectral analysis of intraparenchymal vessels requires that the smallest sample volume that will en-compass the vessels be used. The waveform pattern will permit assessment of the type of vascular flow within the lesion. Analysis of the waveflow is performed using 2 standard indices. The first is the RI, which is defined as the maximum systolic velocity minus the end-diastolic velocity divided by maximum systolic velocity. The second index is the PI, which is defined as the maximum systolic velocity minus the end-diastolic velocity divided by the mean velocity. There is debate as to which index is preferred, and several investigators have reported that both are equally accurate.^12,13 At our institution, the PI is preferred because it takes into account the entire shape of the waveform produced during 1 complete cardiac cycle.

At present, it is difficult to establish an exact PI or RI value that is indicative of malignancy. Some investigators have assigned an arbitrary cutoff of PI less than 1.0 or RI less than 0.4.^5,6 Not all investiga-tors, however, have established strict criteria. It may be that an absolute value cannot be established, but rather that these indices must be interpreted in the context of the morphologic and clinical findings. Certain investigators have reported the lowest obtainable PI as representative of the neovascularity of the mass. Most studies have shown a significant difference in the impedance patterns of benign versus malignant masses irrespective of the index used.

Some of the controversy regarding discriminatory values can be attributed to different sensitivities of each type of CDS scanner. With some scanners a waveform can be obtained even in areas that do not demonstrate color; with others a waveform cannot be generated even though there is color. In addition, as scanners have become more sensitive to slow flow, the impedance values in certain tumors may decrease because intraparenchymal vessels that may not have been previously detected with other equipment will be displayed.

The big question regarding CDS of ovarian masses involves whether macroscopic differences in tumor neovas-cularity correspond and/or reflect that of tumor microvascularity, which is at the capillary level. To better delineate the histopathology of these conditions, multiple parameters, including vessel location and arrangement, and impedance indices, have been added to the sonographic evaluation.¹² This use of multiple parameters has led to the development of scoring systems. Timor-Tritsch et al examined the role of morphology and CDS in 115 patients.² They found that abnormal morphology and low PIs accurately predicted malignancy. Con-versely, normal morphology and high PIs accurately predicted benignity. When there was discordance in the morphology and the PI, however, there was overlap in benign and malignant conditions.

Kurjak et al used a scoring system that incorporates the variables of morphology, type of vascularity, and RI.¹⁴ This produced high specificity and sensitivity. They defined lesions as benign, questionable, suspicious, and malignant. The questionable and suspicious lesions occasionally demonstrated discordant findings in relation to morphology and CDS.⁵ This experience has been reported by several investiga-tors.^15,16 It seems that each discordant case must be individually assessed according to clinical and laboratory findings.¹⁷

When assessing adnexal lesions, close adherence to optimal Doppler technique is necessary. For example, sampling error can be noted if the wall filters or pulse repetition frequency (PRF) is inappropriately set. This is due to the slow flow of the vessels and the small size of the lesions. Certain limitations arise when using spectral analysis for assessing small intraparenchymal vessels. Color assignment will change due to the tortuosity of the vessels. Due to this tortuosity, it is often difficult to establish the optimal angles for interrogation, making measurements of velocity inaccurate. Many of the intraparenchymal vessels demonstrate extremely slow flow. Examination of these vessels is optimized by adjusting the color gain, PRF settings, and wall filters.³

False-negative examinations will occur if the vessel interrogated is not the tumoral vessel. For this reason, in suspicious lesions, multiple areas must be investigated. The size of the lesion may affect the presence of abnormal flow. In very small lesions (<2 mm) it may be that neovascularity has not been established; therefore, abnormal waveform patterns will not be demonstrated.¹² Large lesions may have absent or diminished flow due to the necrosis from a torsion of the pedicle of the mass.¹² False-positive values can be noted within the corpus luteum of hormonally active women.¹² This is especially problematic with the cyclical variation in premenopausal women. This can be minimized by examining the patient noted in the early follicular phase. At this time, intraparenchymal normal vessels typically show high pulsatility. Other conditions that will cause false-positive values include inflammation and benign but hormonally active neoplasms of the ovary.

General Principles

Adnexal masses may be difficult to characterize on a morphologic basis alone. Since most benign lesions, as well as most neoplasms, are predominantly cystic, a method to differentiate these lesions would be clinically valuable. Un-fortunately, morphologic evaluation by gray-scale sonography alone may at times be inadequate. Sonographic characterization of the internal contents, such as low-level echoes, septation, and projections, may suggest the most likely type of tumor as well as the malignant potential of these lesions. Color Doppler sonography affords additional information concerning the growth and malignant potential of these masses.¹²

Recently, a study of the various gray-scale and CDS parameters indicated that the presence of a solid component is the most statistically significant predictor of a malignant ovarian mass.¹⁸ For masses that did not contain a solid component, the location of flow (central), amount of intraperitoneal fluid, and presence and thickness of septations seemed to have statistically significant discrimination power. Others have confirmed that the utility of CDS is related to the proportion of multiloculated masses with solid components that are stud-ied.¹⁹

The pelvis is the source of benign and malignant ovarian lesions and of adnexal masses, including paraovarian cysts, tubo-ovarian complexes, and uterine leiomyomata. Doppler characteristics of these lesions reflect their vascu-larity. In general, benign pelvic masses exhibit high-resistance waveform patterns (ie, high PI). Functional ovarian masses often present with patterns that demonstrate low resistance to diastolic flow (ie, low PI). With hormonal ma-nipulation, cyclical variation, or both, these lesions typically regress or revert to a high-resistance waveform pat-tern.²⁰ Of the benign ovarian lesions, 2 in particular are great mimickers. These are endometriomas and teratomas. Morphologically, they may appear suspicious for malignancy due to their solid components. Doppler interrogation of these lesions may differ with areas of high or low diastolic flow.^5,7 The pattern of low impedance (low PI) may be problematic when assessing for the possibility of malignancy because they may contain vessels with low-impedance and high-diastolic flow. The flow within these masses had been related to their biologic activity, the presence of acute hemorrhage, or both.²⁰ The clinical presentation of patients with these masses may be useful in further distinguishing them from women at risk for ovarian malignancies.

Malignant ovarian masses often show characteristic findings on gray-scale sonography. In those cases in which the morphology is nonspecific, Doppler interrogation may add additional information. The CDS analysis obtained from malignant vessels will usually show an increase in the number of vessels, as well as central or scattered locations of these vessels. The diastolic notch is characteristically absent.⁶ This high-diastolic flow (low PI) reflects the neovascularity of the malignant process. It seems to be most evident in rapidly growing tumors as compared with those late-stage tumors that have become necrotic. In our experience, 3 stage I ovarian cancers were found out of the 9 diagnosed based on their CDS findings. All of the late-stage tumors, however, were diagnosable based on their morphologic features alone.^7,21,22 Some have reported differences in velocity in malignancies, although our series showed a trend but not a statistically significant differ-ence.^21,22

Other lesions that may contain high-diastolic flow (low PI) include inflammatory tubal or bowel masses and uterine leiomyomata. Morphologic changes that indicate that these masses arise outside of the ovary, as well as the patient's clinical history, are useful in assessing these patients. Color Doppler sonographic evaluation of tubo-ovarian complexes will differ depending on the presence of concurrent inflammation. Characteristically, in those patients actively infected, there will be low resistance to diastolic flow (low PI). This reflects the active inflammatory process with vasodilation occurring in these lesions. However, in patients with adequate medical treatment, the tubo-ovarian complex is a residuum of the previous disease. These noninfected masses often demonstrate increased diastolic resistance and high PI values.²³ With uterine leiomyomata, a rim of stroma may be seen along the uterine wall, separating it from the ovary. This rim of stroma represents the peritoneal surface of the leiomyomata and is quite useful in determining the epicenter of the mass. The resistive indices in leio-myomata will differ depending on the vascularity of the mass and whether or not normal vessels are parasitized.

Figures 32-1, 32-2, 32-3, 32-4, 32-5, 32-6, 32-7, 32-8, 32-9 and 32-10 illustrate the morphologic and spectral analysis of selected pelvic masses. Table 32-1 lists the multiple parameters used and typical findings in benign and malignant masses. Table 32-2 illus-trates the expected sonographic characteristics compared to impedance patterns of selected adnexal masses.

Using a combination of morphology and spectral analysis, investigators have attempted to determine the histopathology of pelvic masses. Fleischer et al²¹ examined 62 patients in whom 25 malignancies were pathologically proven. Twenty of the malignancies showed low PIs and an absent diastolic notch; however, 3 benign lesions also shared similar findings. This resulted in a 98% negative predictive value and a positive pre-dictive value of 83%.²¹ A high negative predictive value would be expected of any test in a disease that has low prevalence such as ovarian carcinoma. Kurjak et al identified 624 benign masses and 56 malignant masses, including 16 stage I lesions, in more than 14,000 women.¹⁴ Neovascularity was seen in 15 of 16 of the stage I carcinomas and in 39 of the 40 stage III or IV ovarian carcinomas. At our institution, retrospective analysis of 96 patients with pathologic correlation after TV-CDS indicated that in approximately 40% of cases, CDS provided enhanced specificity concerning the organ of origin and the cell type.¹ This specificity was most evident in the detection of ovarian malignancy, ovarian torsion, and ectopic pregnan-cies. In 40% of cases, the specificities of TVS and CDS were equal; in 6% of the patients, TVS was more specific than CDS. In 2 lesions, neither exam was histologically definitive. These included a 2-mm metastatic ovarian can-cer, which was not diagnosed, and an intraligamentous leiomyomata, which was misdiagnosed as an ovarian neoplasm.¹

When interpreting the sonographic findings, it is important to be aware of the patient's clinical history, including age, prior surgery, risk factors for infection, and hormonal status. Laboratory values, such as white cell count and CA-125 values, are also important in evaluating this information. Because low-impedance flow occurs in a number of clinical situations, it is only by integrating these factors that the significance of the lesion be appreci-ated. It would appear that CDS can accurately exclude the possibility of malignancy in most situations. This enables the clinician to identify those patients in whom aggressive management is indicated.

The true sensitivity of CDS awaits further evaluation and study. At present, many investigators have reported on small groups of patients in whom it was felt that the added specificity afforded by CDS substantiated its use, especially in those cases in which gray-scale sonography was equivocal or nondiagnostic (Table 32-3). In these studies, true-positive rates for the diagnosis of ovarian cancer (percentage of lesions with low impedance that were ovarian cancer) ranges from 80% to 100%, whereas false-positive (percentage of benign masses with low imped-ance) ranges from 0 to 20%. These results need to be analyzed relative to the population studied and techniques used. Although only approximately 50 stage I ovarian cancers have been reported worldwide, it would appear that CDS, coupled with the morphologic appearance of the mass, adds important clinical information.

Adnexal Torsion

Color Doppler sonography provides important data concerning the evaluation of ovarian (adnexal) torsion. Adnexal torsion is usually associated with abnormalities in the blood supply and venous return from the ovary. This, in turn, may be associated with anatomic derangement and associated intraovarian hemorrhage as well.

In order to better understand the sonographic findings in torsion, it is important to consider the arterial and venous flow to and from and within the ovary. The ovary receives arterial blood from 2 sources. Variation in the completeness of the obstruction of this vascular supply may explain the variety of the spectrum of CDS findings that have been identified in ovarian torsion.^24,25 Another parameter that may factor into what is seen with CDS is the pressures inside the ovary. This can be altered if there is increased venous pressure from interstitial (intraovarian) hemorrhage. Clearly, the CDS findings in torsion depend on the severity and chronicity of the torsion. In incomplete or intermittent torsion, the CDS findings may be minimal, whereas in patients with several twists or complete torsion, no flow will be detected. Typically, the first finding is that of lack of venous flow, in which case arterial flow may still be demonstrated, although there may be high-resistance flow in these cases.^1,3,4 Subsequently, absent arterial flow is demonstrated. It is important to examine the inner portions of the ovary, because the capsular vessels may not be compro-mised. Demonstration of an enlarged ovary with small central cystic spaces resulting from macroscopic hemorrhage and compromised vascularity on CDS suggests a diagnosis of ovarian torsion. We have also reported on the CDS findings in isolated tubal torsion, which usually occurs in women who have undergone minor tubal ligation.²⁵ The fusiform mass representing the distended torsed tube usually lacks any demonstrable flow on CDS. The interested reader is referred to Chapter 36 on pelvic pain for more extensive discussion on illustration of this topic.

The Postmenopausal Ovary

Eighty percent of ovarian carcinomas arise in women older than 50, and the majority of these are of the cystic epithelial type. It is evident that demonstration of a cystic mass in a postmenopausal woman is of clinical signifi-cance. A cyst that fulfills all criteria for a simple cyst in the premenopausal woman has been shown to be benign in nearly 100% of cases in premenopausal women and in 95% of cases in postmenopausal women.¹⁴ Wolf et al examined 149 postmenopausal patients, 15.8% of whom demonstrated a benign cystic adnexal mass.²⁶ The prevalence of ovarian carcinoma at age 60 is 70 per 100,000 women, whereas the prevalence of uniloculated, nonseptated ovarian cysts in this study was 14,800 per 100,000 women. This study would indicate that postmenopausal ovarian cysts are 200 times more frequent than ovarian carcinoma. In a follow-up study, Levine et al followed the natural history of simple adnexal cysts in postmenopausal women. In the 72 masses that were followed, 53% disappeared and 39% increased in size or remained constant.²⁷ In a similar study performed in Scandinavia with a follow-up of 3 years, 78% of masses disappeared or did not change. Most of the cysts that regressed were in women less than 60 years of age. In 13%, new cysts were observed during the study interval.²⁸ The postmenopausal ovary has long been viewed as a static organ. These studies, however, would tend to lend support to the fact that the ovary is a dynamic structure even into menopause.

Little is known concerning the natural history of ovarian carcinoma. It is not clear what role, if any, the formation of benign cysts play in tumorigenesis. In view of the prevalence of cysts in the ovaries of peri- and postmenopausal patients, the area lends itself to future research.²⁹ These studies underscore the need to determine the malignant potential of these lesions to decrease unnecessary surgeries. Current management would be to consider sonographic follow-up for those women with small simple cysts demonstrating reassuring Doppler interrogation and normal CA-125 values. All other cases should be referred for further evaluation and possible surgery.

Ovarian carcinoma is the most common lethal gynecologic malignancy, despite the fact that endometrial and cervical malignancies have a greater incidence. This is due in part to the paucity of symptoms and the resultant delay in diagnosis. It was envisioned that CDS could diagnose ovarian carcinoma in its early stages; however, to date, only approximately 50 stage I carcinomas have been reported as detected by CDS (see Table 31-3). Most of these lesions have been recognized by their low-impedance patterns. It seems reasonable to hypothesize that this tumor neovascularity is most evident during the rapid growth phase of an ovarian malignancy, thereby making it most readily detectable.

Recent data suggest significant potential for contrast-enhanced transvaginal sonography as a means to detect early ovarian cancers. This technique has a similar basis, which is the detection of tumor neovascularity. It is important to realize that contrast-enhanced transvaginal sonography (CE-TVS) does not utilize CDS due to its limited resolution of tumor neovascularity, which consists of capillaries. Alternatively, CE-CDS utilizes harmonic pulse in-version imaging of microbubbles as displayed in gray scale.

Three-dimensional CDS can depict the abnormal branching pattern and irregular caliber of tumor vessels. Quantitative techniques are now available to objectively quantify this assessment using fractal mathemat-ics.^{30, 31, 32, 33 and 34}

Sonographic recognition of abnormal morphologic features of an ovarian mass are usually sufficient in the diagnosis of late-stage disease. Typical sonographic findings include large pelvoabdominal masses with solid areas, ascites, and focally irregular walls with papillary excrescences. Color Doppler sonography is of little incremental benefit in these patients.

The accuracy of CDS appears to be superior to CA-125 for diagnosing stage I lesions. When used in a high-risk population as an adjunct to TVS, CDS seems to be most clinically useful.⁶ CA-125 values are elevated in 30% to 50% of stage I lesions and in more than 80% of stage III or IV lesions. The lack of sensitivity of CA-125 in stage I lesions precludes its use as a screening tool.

CDS is probably most accurate in detecting rapidly growing tumors that have achieved a critical tumoral mass. False-negative results will be obtained when the lesion is too small to characterize, lacks neovascular-ity, or, in the case of necrotic lesions, has outgrown its blood supply. In large lesions it may be necessary to interrogate multiple sites to identify areas of low-impedance flow. It is important that multiple sites, especially central vessels or projections, be examined to assess for neovascularity in these lesions. Of special interest are those epithelial lesions of low malignant potential. Doppler interrogation may be able to demonstrate areas of low-impedance flow in morphologically benign-appearing areas, suggesting the possibility of malignancy (Table 32-4 and Figure 32-11).

In the largest series of stage I ovarian cancer detected by TV-CDS, Kurjak et al³⁵ reported detection of 18 cases (see Figure 32-11). Stage I ovarian cancer was found in 4 asymptomatic women in whom the ovary was morphologically normal or had a unilocular cyst.³⁵ Of the group of 18 women who were retrospectively analyzed, 1 patient with stage Ia and another with stage Ib were missed. The world experience in detection of stage I ovarian cancer totals approximately 50 patients to date. More experience is needed with con-ventional TV-CDS, power CDS, and CE TV-CDS to determine the relative efficacy of these tech-niques.³⁶

There is a lack of universal criteria for "cutoff" values for impedance values to distinguish between benign and malignant lesions. Part of the discrepancy between studies can be attributed to extensive differences, such as system sensitivity, or intrinsic differences, such as the population of tumors studied. Studies need to specify which vessels were interrogated and the stage of the tumors detected. Most studies have indicated that morphologic evaluation of ovarian tumors may be equal or more accurate than CDS.^37,38 Further investigation concerning conventional CDS and newer types of CDS such as "power" CDS and CE TV-CDS is required before these techniques can be universally applied.

The potential for extensive clinical application of CDS in the evaluation of adnexal masses seems great because it affords an additional parameter to improve the characterization of adnexal masses. Because it depicts relative vascularity, it may also be useful as a means to monitor response to treatment. Because vascular patterns can be affected by a number of benign conditions, including cyclical changes, hormonally active lesions, and inflammatory processes, it is important that these conditions be considered. In patients in whom concordant findings are identified (ie, both morphology and spectral analysis are indicative of either a benign or malignant process), it is safe to recommend therapy as indicated by these findings. In cases of discordant findings in which the morphology is not in agreement with the spectral analysis, it is important to search for other etiologies, such as hormonal changes in the premenopausal woman or an infectious process. If these cannot be ascertained, then it may be necessary to rule out the possibility of an underlying malignancy.

The role of the sonologist is to determine those cases in which active intervention is indicated. These represent situations in which expectant management will be inappropriate. The end point should be whether or not the patient requires intervention, not whether or not the patient has an underlying malignancy. It is not a failure of the technique that in certain clinical situations high-diastolic flow is noted but represents a benign process (ie, inflammation or metabolically active tumors). In these circumstances, intervention is required in either the form of antibiotics or surgery. In premenopausal women in whom a hormonal etiology is felt to be the likely cause of these discordant findings, it is important to recommend serial assessment to demonstrate interval resolution. After a period of observation, most of these lesions will regress, obviating more aggressive interventional management. In a study of 65 women with adnexal masses, lesions that had high-impedance flow and were smaller than 5 cm with a cystic or complex texture tended to regress after 6 or 12 weeks (Figure 32-12).³⁹

There have been several new developments in TV-CDS of ovarian masses in the past few years. Scanners have become more sensitive to intraovarian flow, thus resulting in the depiction of more vessels in and around the ovary. Improved resolution has also allowed for better depiction of arterial and venous ovarian flow (Figures 32-13, Figures 32-14 and 32-15).

Another major improvement is the ability to depict tumor neovascularity in 3D. This affords display of vessel branching and caliber, both of which are abnormal in tumors. Recent software improvements allow for these parameters to be quantitated, thus objectifying assessment of vascularity with 3D-CDS.

Which parameters to assess and the selection of "cutoff" values remain unsettled. It is clear that malignancies have lower impedance and higher velocity flow than do benign lesions. However, many overlaps have been observed between these 2 groups, making it difficult to use a single sonographic parameter (Table 32-1 and Figure 32-16). Multiparameter analysis is encouraged but makes determination of what is potentially malignant from probably benign more complicated. It is also clear that TV-CDS provides a means to determine the location and arrangement of vessels. This should decrease sampling error because one can assess only those vessels that have significant Doppler shifts versus those that are associated with tumor growth. Technical errors such as improper scanning angle, PRF, and gray-scale versus color priority settings will continue to complicate diagno-ses.⁴⁰ However, improvements in 3D sonography with contrast may improve determination of vessel density and arrangement, important parameters in distinguishing benign from malignant ovarian masses.⁴¹

Detection and quantification of tumor angiogenesis will become of primary importance in assessing the possibility of malignancy. It has been shown that tumor angiogenesis is a prognostic factor in patients with advanced ovarian carcinoma.⁴² Other studies have suggested that intratumoral blood flow correlates with the endothelial cell activity of the tumor vessels and not with the mean vascularity den-sity.⁴³ Combining the TV-CDS findings with assay of thymidine phosphorylase, an enzyme involved in tumor angiogenesis, may allow assessment of malignant possibility and potential response to treat-ment.⁴⁴

Future developments may include improved evaluation of "tumor enhancement kinetics" and enhanced evaluation of overall blood flow in normal and tumorous areas. This technique has significant potential as a diagnostic and therapeutic means, particularly in light of the increased use of antiangiogenesis agents and the importance of monitoring tumor response. The interested reader is referred to Chapter 35 devoted to early detection of ovarian cancer for further information and discussion.

Transvaginal sonography, coupled with CDS and spectral analysis, has added to our understanding of adnexal masses.^{45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59 60, 61, 62, 63, 64, 65, 66 and 67} CDS should not be seen as an absolute test. No single pattern can be taken as a definitive indication of malignancy or benignity (Figures 32-17 and 32-18). The strength and weakness of the modality must be recognized in order to optimize its clinical use. It is important that the findings demonstrated at the time of the examination be correlated with the patient's clinical and hormonal status, as well as with laboratory assessment, including CA-125, when indicated. To facilitate evaluation, close communication with both the patient and her clinician is paramount to an understanding of the significance of the sonographic findings.

Continued research, such as in the 3D quantification of vessel density and objective assessment of flow patterns, will better define the role of color Doppler sonography in the assessment of adnexal masses. Recent refinements in CDS including 3D and quantification may improve the accuracy of transvaginal sonography in determining which lesions are malignant (Figures 32-19, Figures 32-20, 32-21, 32-22, 32-23 and 32-24). The interested reader is referred to Chapter 35 devoted to early detection of ovarian cancer with sonography for a more detailed discussion of new techniques such as constrast-enhanced transvaginal sonography for the early detection of ovarian cancer (Chapter 35).