Does quantitative analysis of three-dimensional power Doppler angiography have a role in the diagnosis of malignant pelvic solid tumors? A preliminary study

Authors


Abstract

Objective

To investigate the role of a simplified method for the three-dimensional (3D) quantification of tumor vascularity in the differential diagnosis of solid pelvic masses.

Methods

Twenty-four patients with a solid pelvic mass on B-mode ultrasound evaluation underwent 3D power Doppler evaluation before surgery. The volume of interest was obtained by drawing the largest section of the mass in three orthogonal planes. The following 3D vascular parameters were calculated for all patients: relative color, average color and flow measure. Receiver–operating characteristics curve analysis was used to choose the best cut-off value. The overall agreement between the test result and the actual outcome was calculated using kappa statistics.

Results

Fifteen of 24 subjects were found to have pelvic malignancy. The relative color and the flow measure were significantly higher in malignant (9.7 (8.98) and 8.95 (8.78) (median (interquartile range (IQR)), respectively) than in benign masses (2 (4.42) and 1.79 (4.71), respectively; P < 0.05) but there was no difference in the average color. The best cut-off values of relative color and flow measure were 3.2 and 2.8, respectively. The use of relative color showed a sensitivity of 80% with a specificity of 67% with an overall agreement that was higher, though only marginally so, than that of qualitative 3D power Doppler (kappa = 0.47 and 0.44, respectively).

Conclusions

In a small group of pelvic masses that appear malignant on B-mode ultrasonography, the use of 3D quantification of tumor vascularity yields a diagnostic accuracy that is similar to that of subjective evaluation of vascularity. We suspect, however, that the quantitative method will produce more consistent results between operators. Copyright © 2005 ISUOG. Published by John Wiley & Sons, Ltd.

Introduction

Solid pelvic masses, defined as a tumor where the solid component comprises 80% or more of the tumor when assessed in a two-dimensional (2D) section1, are at high risk of being malignant2, 3. Using B-mode ultrasonography, virtually all solid masses should be suspected to be malignant although a high false-positive rate should be considered in the differential diagnosis of adnexal malignancies. Conventional color and pulsed Doppler have been introduced to improve the diagnostic accuracy of gray-scale morphological ultrasonography but the results show it is of limited value3. Three-dimensional (3D) power Doppler sonography has been used to characterize microvasculature of normal and pathological conditions in the uterine cervix4, benign ovarian masses5 and normal ovary6–8. In an attempt to improve the reproducibility of subjective evaluation of vessel architecture, Pairleitner et al.5 described a method of selecting a region of interest within a 3D volume; specifically designed software was used to generate indices that represented the whole vascularization in the defined region of interest. Other authors tested the reproducibility of 3D technology by describing the volume of interest of a normal ovary on the basis of manual definition of contours, which were drawn when six image planes had been generated around a maximum longitudinal section6, 7. Raine-Fenning et al.8 examined the interobserver reliability of 3D power Doppler ovarian data acquisition from 20 patients at various stages of in-vitro fertilization using a computerized tool. These authors showed that 3D ultrasound can be reliably used between observers to acquire power Doppler information from the ovary, thus supporting the current use and further development of this technique in clinical practice8. We recently described a simplified method of quantifying power Doppler signals in a tumor ‘volume’ obtained on 3D ultrasound examination of pelvic masses and achieved quite acceptable rates of intra- and interobserver variability9.

Several studies have investigated the clinical role of 3D ultrasound examination with or without 3D power Doppler in the diagnosis of pelvic malignancies with controversial results10–16. To our knowledge, 3D quantification of tumor vascularity has not been prospectively evaluated in the differential diagnosis of solid pelvic tumors.

The purpose of this prospective study was to evaluate a simplified method for the 3D quantification of tumor vascularity in the differential diagnosis of malignancies from benign solid pelvic masses. In addition, we compared this approach with the color score semiquantitative evaluation1 and with qualitative 3D power Doppler evaluation of tumor vessels12.

Methods

Twenty-four patients with solid pelvic masses on B-mode ultrasound evaluation underwent 3D power Doppler evaluation. In four patients the masses were bilateral and with the same B-mode appearance, but only one mass for each patient was included in the study. All participating patients gave their informed consent. The study was approved by our Institutional Review Board. All the examinations were performed by an experienced sonologist (A.C.T.).

In the 2 days prior to surgery, all patients underwent ultrasound examination using gray-scale, color and spectral Doppler modes available on a commercial Technos MP ultrasound machine (ESAOTE S.p.A., Genova, Italy). Scans were performed using a multifrequency transvaginal probe (EC 123 range from 9.0 MHz to 5.0 MHz). In the color, spectral and power modes, the Doppler ultrasound had a frequency of 5 MHz. The wall filter was set at 100 Hz. Color signals along the wall and within the septa were identified and the pulsed Doppler gate was superimposed; the lowest resistance index (RI) and the corresponding value of pulsatility index (PI) were calculated according to the methods previously published17.

All color Doppler examinations began with the same settings of the ultrasound system, which were maximized to achieve the highest sensitivity for detection of color Doppler signals allowing detection of blood flow velocities ≥ 3 cm/s, and the color gain was set just below the background noise level to increase, as far as possible, the Doppler sensitivity for low-velocity flow detection.

A subjective semiquantitative assessment of the amount of blood flow within the examined lesion was made (color score) according to the IOTA consensus statement1. The amount of blood flow within the mass was scored as follows: a score of 1 was given when no blood flow could be found in the lesion; a score of 2 was given when only minimal flow could be detected; a score of 3 was given when moderate flow was present, and a score of 4 was given when the pelvic mass appeared highly vascularized with marked blood flow1. A color score ≥ 3 was considered suggestive of malignancy.

Acquisition data from volumetric reconstruction were stored inside the ultrasound unit for a postprocess exam analysis. The ultrasound volume acquisition was performed using the Flock of Birds technique (Ascension Technology Corporation, Burlington, VT, USA), which consists of an ultrasound scanner plus a tracking system that enables the correlation between the spatial coordinate of the probe position and the images generated during the scan. The tracking system consists of an electromagnetic transmitter positioned close to the patient, plus a receiver that is fixed (by a specific holder) to the probe. The electromagnetic field generated from the transmitter is recognized by the receiver. The spatial coordinates of the probe are calculated by analyzing the signal obtained from the receiver.

During the ultrasound scan the tracking system provides a set of spatial coordinates plus three angles to identify the probe position in the space. At the end of the scan the system collects a series of spatial probe coordinates and a set of images. These data are processed in order to generate a volume (‘resampling’). In the 3D reconstruction of the ‘resampled’ volume, the smallest measurable volume is called a voxel. Voxels with a power Doppler signal are defined as ‘color voxels’. In the obtained 3D volume it is possible to define the volume of interest.

For these ovoid masses, we used a previously described and validated method9 to select the region of interest, different from the method provided by the manufacturer: in the three orthogonal planes the margins of the largest section of the mass are drawn, cutting away the surrounding tissue. In the selected volume, the following 3D parameters were calculated: ‘total volume’, which corresponds to the sum of all the voxels; ‘color volume’, which corresponds to the sum of all the color voxels; and ‘relative color’, which is equal to the color volume measurement, divided by the total volume measurement. Each color voxel has a different intensity derived from the intensity of the power Doppler signal and the ‘average color’ corresponds to the sum of the intensity of each color voxel divided by the color volume. The ‘flow measure’ (known also as ‘flow intensity’) corresponds to the sum of the intensity of each color voxel, divided by the total volume.

The intraobserver agreement was high for each 3D-derived parameter calculated from two observations made 24 h apart, calculated in a previous study9, with intraclass correlation coefficient (intra-CC) values of 0.920 (CI 95%, 0.914–0.926), 0.978 (CI 95%, 0.970–0.986) and 0.978 (CI 95%, 0.977–0.993) for the ‘relative color’, the ‘average color’ and ‘flow measure’, respectively. Furthermore, the reproducibility of 3D indices obtained by two different operators was high with interclass correlation coefficient (inter-CC) values of 0.978 (CI 95%, 0.970–0.986), 0.966 (CI 95%, 0.957–0.975) and 0.997 (CI 95%, 0.986–1.000) for the ‘relative color’, the ‘average color’ and ‘flow measure’, respectively9. Receiver–operating characteristics (ROC) curves were used to select the best cut-off value of significantly different parameters. All masses were also evaluated by a qualitative 3D power Doppler analysis. The presence of irregular and randomly dispersed vessels with complex branching were suggestive of pelvic malignancy12.

The 3D parameters and the ultrasonographic impressions were then compared with the final histopathological diagnosis obtained after surgery. Malignant tumors were staged surgically according to the criteria established by the International Federation of Gynecology and Obstetrics (FIGO).

Statistical analysis was performed using non-parametric Mann–Whitney U-test (α = 0.05) for two group comparisons and P < 0.05 was considered statistically significant. Non-parametric testing was chosen because all variables manifested skewed distribution. Fisher's exact test was used for categorical variables (α = 0.05). The sensitivity, specificity, positive (PPV) and negative predictive values (NPV), of each test were calculated for each pelvic mass18. To evaluate the overall agreement between a test result and the actual outcome, the kappa index was calculated according to Fleiss19. Kappa values ranging between 0.40 and 0.75 were assumed to indicate a fair-to-good agreement. The z statistic for the comparison of two proportions20 was used to evaluate the specificities and sensitivities obtained using the different approaches. The diagnostic value of each test was also assessed using likelihood ratios (LR) that are not affected by the prevalence of disease in the studied population. Fair-to-good test performance is associated with a positive LR (LR+) between 2.0 and 5.0 and a negative LR (LR−) between 0.5 and 0.221, 22.

Results

Nine of the 24 pelvic masses were benign, and 15 were malignant (prevalence of malignancy per mass, 62.5%). Clinical characteristics of the patients and tumor histopathological results are reported in Table 1. Five ovarian carcinomas were FIGO Stage III.

Table 1. Characteristics of patients and tumors
Characteristicn
Age (years, mean ± SD)52 ± 10
Postmenopausal status (%)16 (67)
Histopathology 
 Benign9
  Ovarian fibroma7
  Brenner tumor1
  Schwannoma-like leiomyoma1
 Malignant15
  Serous ovarian carcinoma5
  Clear cell ovarian carcinoma1
  Krukenberg3
  Granulosa cells tumor1
  Lymphoma1
  Carcinosarcoma1
  Liposarcoma1
  Sarcoma peritonei1
  Malignant tumor of smooth muscle1

Ultrasonographic and color Doppler characteristics of the tumors examined are reported in Table 2. All masses were found to be vascularized on color Doppler evaluation. Color score was ≥ 3 in 11 (73%) malignant masses and in 2 (22%) benign masses (P = 0.033). No differences were observed for PI and RI, according to the histopathological results (Table 2).

Table 2. Ultrasonographic and color Doppler characteristics of the tumors examined
CharacteristicBenign (n = 9)Malignant (n = 15)P
  1. IQR, interquartile range; PI, pulsatility index; RI, resistance index.

Median (IQR) diameter of the mass (mm)61 (19)87 (49)0.06
Color score ≥ 3 (%)2 (22)11 (73)0.033
Lowest PI (median (IQR))0.78 (0.53)0.52 (0.4)0.36
Lowest RI (median (IQR))0.54 (0.19)0.41 (0.21)0.39

Qualitative 3D power Doppler analysis suggested pelvic malignancy in 13 (87%) malignant masses and in 4 (44%) benign masses (P = 0.061). The relative color and the flow intensity were significantly higher in malignant than in benign pelvic masses (P = 0.016 and 0.034, respectively) but the associated interquartile ranges (IQR) were wide (Table 3). No difference in the average color of benign vs. malignant pelvic masses was found (Table 2). ROC curve analysis showed that the best cut-off values were 3.2 for relative color (Figure 1) and 2.8 for flow intensity. The areas under the curve were similar but greater using relative color (0.800 and 0.763, respectively). A relative color > 3.2 was observed in 12 (80%) malignant masses and in 3 (33%) benign masses (P = 0.036) (Table 4).

Figure 1.

Receiver–operating characteristics curve analysis of relative color with the best cut-off value at 3.2 (area under the curve: 0.800, SE 0.09).

Table 3. 3D power Doppler sonography derived indices in the study population
3D parametersBenign (n = 9)Malignant (n = 15)P
  1. IQR, interquartile range.

Relative color (median (IQR))2 (4.42)9.7 (8.98)0.016
Average color (median (IQR))91 (17)89 (14)0.83
Flow intensity (median (IQR))1.79 (4.71)8.95 (8.78)0.034
Table 4. Comparison of diagnostic capabilities of different Doppler methodologies for the detection of malignancy among solid pelvic masses
Doppler methodologyMalignant (n)Benign (n)Total (n)
Color score   
 ≥ 311213
 < 34711
Qualitative power 3D   
 Positive13417
 Negative257
Quantitative power 3D (relative color)   
 ≥ 3.212315
 < 3.2369

The sensitivity, specificity, PPV, NPV, LR+, LR− and kappa values of relative color are reported in Table 5. The values of sensitivity, specificity, PPV and NPV of flow intensity in the diagnosis of pelvic malignancy were the same as those of relative color. The use of relative color showed a sensitivity of 80% with a specificity of 67% and a LR− of 0.3 (Table 5). Of the 15 masses suspected of being malignant because they produced a relative color > 3.2, 12 were confirmed by pathology (Figure 2). The three false-positive cases (a fibroma, a benign Brenner tumor (Figure 3) and a schwannoma-like leiomyoma) showed values of relative color of 7.6, 6.2, and 5.1, respectively. Of nine masses that showed values of relative color lower than 3.2, six were confirmed by pathology (Figure 4). The three false-negative cases were a carcinosarcoma (Figure 5), a liposarcoma, a sarcoma peritonei with values of relative color of 1.2, 1, and 1.2, respectively.

Figure 2.

A true positive case of three-dimensional color indices. The histopathological evaluation confirmed the presence of an ovarian carcinoma.

Figure 3.

A false-positive case of three-dimensional color indices. The color indices were elevated but histopathological evaluation showed the presence of a benign Brenner tumor.

Figure 4.

A true negative case of three-dimensional color indices. The histopathological evaluation confirmed the presence of an ovarian fibroma.

Figure 5.

A false-negative case of three-dimensional color indices. The color indices were low but histopathological evaluation showed the presence of ovarian carcinosarcoma.

Table 5. Comparison of diagnostic parameters of different Doppler methodologies for the detection of malignancy among solid pelvic masses
ParameterDoppler methodology
Color scoreQualitative power 3DQuantitative power 3D (relative color)
  1. LR−, negative likelihood ratio; LR+, positive likelihood ratio; NPV, negative predictive value; PPV, positive predictive value.

Sensitivity (%)738780
Specificity (%)785667
PPV (%)857680
NPV (%)367167
Accuracy (%)757575
LR+ (95% CI)3.3 (0.94,11.63)1.95 (0.91, 4.16)2.24 (0.92, 11.63)
LR− (95% CI)0.34 (0.14, 0.85)0.24 (0.06, 0.99)0.3 (0.1, 0.91)
Kappa value0.490.440.47

The use of relative color was marginally less accurate than 2D color score in the diagnosis of adnexal malignancies (kappa values were 0.47 and 0.49, respectively), but there was no significant difference in terms of sensitivity and specificity (P = 0.983 and P = 0.996, respectively) (Table 5). Furthermore, the use of qualitative 3D power Doppler was slightly less accurate than relative color in the diagnosis of adnexal malignancies (kappa value was 0.44), in the absence of any significant difference in terms of sensitivity and specificity (P = 0.990 and P = 0.996, respectively).

Discussion

This preliminary study on the 3D quantitative analysis of the vascularization in ovarian masses focused on solid adnexal masses since neither B-mode nor color/power Doppler parameters are accurate in discriminating between benign and malignant neoplasms3. Although we evaluated a relatively small sample, the 3D quantification of tumor vascularity seems to be associated with an acceptable specificity although it is characterized by a 20% false-negative rate. It is worth noting that the three false-negative cases corresponded to ovarian masses with sarcoma components at histology. The hypothesis that necrosis could influence the percentage of vascularization in these masses was considered but not confirmed. Since this is the first study to investigate 3D quantitative analysis for the detection of pelvic malignancy, no comparison is possible with other published articles. For this reason, in the present study, we performed a comparison with other previously used methods. 3D quantitative analysis did not significantly improve the accuracy provided by semiquantitative 2D color Doppler analysis or qualitative 3D analysis. However, it had the advantage of providing an objective evaluation of vascular flow.

In pelvic solid tumors, where the risk of malignancy is very high, the morphological analysis of ovarian tumors with 3D ultrasound is impractical because all masses appear malignant10–12, 23. A more promising 3D technique, which may be used to evaluate adnexal masses, is 3D power Doppler. The efficiency of this method was first described by Kurjak and collaborators12–15 but further studies by others have mostly not been able to reproduce the results obtained16, 24, 25. With such controversial results, the use of more objective parameters seems to be mandatory. Our objective 3D approach was validated in terms of reproducibility before assessing a potential clinical application.

Subjective evaluation of masses with central flow continues to be associated with variable reproducibility due to the absence of a standardized definition of abnormal vascularization. Because the three approaches used in this study had a similar accuracy, the more objective evaluation obtained by using 3D indices should be preferred. Theoretically, this method should allow a better comparison of results between different clinics and operators. Whether the use of different machines will influence these results remains to be seen.

Ancillary