Three-Dimensional Doppler Ultrasonography in Assessing Nodal Metastases and Staging Head and Neck Cancer

Transcription

1 The Laryngoscope VC 2013 The American Laryngological, Rhinological and Otological Society, Inc. Three-Dimensional Doppler Ultrasonography in Assessing Nodal Metastases and Staging Head and Neck Cancer Yu-Shih Lai, MD, MS; Chun-Ying Kuo, MD; Mu-Kuan Chen, MD, PhD; Hui-Chuan Chen, PhD Objectives/Hypothesis: This study built a simple prediction system by three-dimensional (3D) Doppler ultrasonography to evaluate the metastases of cervical lymph nodes and the preoperative initial stage of head and neck cancer. Study Design: Retrospective review of cervical lymph node ultrasound features and prospective nodal staging of head and neck cancer. Methods: One hundred thirty-nine suspicious cervical lymph nodes, receiving 3D Doppler ultrasonography, were used to establish a predictive model. Then nodal metastasis was initially staged from 27 patients with head and neck cancer by this model. Results: The prediction system was constructed by major (internal matting, vascularity pattern) and minor (age 40 years, short/long ratio 0.5) sign categories. Cervical lymph node was regarded as metastasis with the presence of one major and any of the other factors. The predictive model resulted in sensitivity of 91.9%, specificity of 88.2%, and accuracy of 89.2%. Then we evaluated the initial staging of patients with head and neck cancer by this model, and the rate of correct N staging was 92.6%. Conclusions: According to this prediction system, 3D Doppler ultrasonography definitely provides a rapid and reliable method for initial staging of head and neck cancer. Key Words: Three-dimensional Doppler ultrasonography, metastatic cervical lymph nodes, prediction model, initial staging. Level of Evidence: 2b. Laryngoscope, 123: , 2013 INTRODUCTION Cervical lymph node metastases are common 1,2 and represent one of the most important prognostic factors in head and neck squamous carcinoma. 3 5 Therefore, evaluation of cervical lymph nodes and preoperative staging are crucial procedures for the planning of appropriate treatment for head and neck cancer patients. The most accurate method for examination of the neck remains controversial. Palpation is the most convenient method for the detection of cervical lymph nodes but requires skill, and its accuracy ranges between 59% and 84% 6,7 because of difficulty in identifying lymph nodes <1 cm in diameter. 8 Contrast-enhanced computed tomography (CT) and magnetic resonance imaging (MRI) remain the typical choices for staging of head and neck cancer because of high anatomic resolution for assessment of a primary tumor. However, the accuracy of From the Department of Otorhinolaryngology, Head and Neck Surgery, Changhua Christian Hospital, Changhua (Y.-S.L., C.-Y.K., M.-K.C.); Mingdao University, Changhua (M.-K.C.); and Providence University, Taichung (H.-C.C.), Taiwan Editor s Note: This Manuscript was accepted for publication May 1, The authors have no funding, financial relationships, or conflicts of interest to disclose. Send correspondence to Mu-Kuan Chen, Chief of Department of Otorhinolaryngology, Head and Neck Surgery, Changhua Christian Hospital, 135 Nanxiao St., Changhua City, Changhua County 500, Taiwan @cch.org.tw DOI: /lary differentiation between benign and malignant lymph nodes varies according to nodal size. 9 Fluorodeoxyglucose-positron emission tomography (FDG-PET)/CT has the potential to provide increased sensitivity and specificity for lymphatic and distant metastases but is too expensive for routine use. 10 Ultrasonography provides a less expensive and effective means of predicting metastasis of cervical lymph nodes and for the initial staging of head and neck cancer, compared to CT, MRI, and FDG-PET/CT. 11 Previous studies have described factors for evaluation in color Doppler ultrasonography, such as size, shape, long diameter, short diameter, short- to long-axis ratio, hilar echogenicity, internal echogenicity in B-mode ultrasonography, and vascular pattern. 12,13 However, a precise and easily used predictive model remains lacking. Therefore, the aim of this study was to develop a rapid and promising predictive model for the evaluation of metastatic cervical lymph nodes using ultrasonography reports. We used the model for the initial staging of neck metastases in a prospective cohort of consecutive patients with untreated head and neck cancer, comparing results with pathological neck dissection specimen findings to evaluate its validity. MATERIALS AND METHODS The institutional review board of Changhua Christian Hospital approved this study. Patient approval and informed consent for review of their images and charts were not required. Among 135 patients at the Department of Otorhinolaryngology, Head 3037

2 and Neck Surgery, Changhua Christian Hospital who underwent ultrasonographic examination, 139 consecutive untreated suspicious neck lymphadenopathy nodes were identified. From April 2011 to March 2012, 139 consecutive ultrasound (US)-guided fine needle aspiration (US-FNA; n 5 105) or excision biopsy (n 5 34) specimens of untreated suspicious neck lymphadenopathy were collected from 135 patients at the Department of Otorhinolaryngology, Head and Neck Surgery, Changhua Christian Hospital. US images were obtained using a Toshiba (Tokyo, Japan) SSA-680A machine with a high-resolution linear array transducer (5 12-MHz frequency range) prior to biopsy. Morphologic and internal echoic features were evaluated using gray-scale ultrasonography. Three-dimensional power Doppler mode ultrasonography was used on a high-sensitivity setting for the detection of patterns in complete lymph node vascularity. Imaging was performed by two doctors (Y.-S.L., C.-Y.K.) from the Otorhinolaryngology, Head and Neck Surgery Department. They entered the US features together, blinded to the pathological results. All lymphoma cases were excluded (n 5 6). Table I lists the final diagnoses for the 139 suspicious cervical lymph nodes. Parameters for the ultrasonographic features evaluated in this study were: 1) Side: left or right side; 2) long-axis diameter (L): the maximum diameter of the longitudinal plane; 3) shortaxis diameter (S): the largest diameter perpendicular to the L dimension; 4) short-axis/long-axis (S/L) ratio: the ratio of S and L; 5) margin: well-defined or ill-defined margin; 6) echogenic level: hypoechoic or isoechoic compared to the surrounding soft tissue; 7) internal echo: heterogeneous or homogeneous; 8) internal matting: present (Fig. 1A) or absent (Fig. 1C) (less than half the area of intranodal coarse content) (Fig. 1B); 9) central necrosis: presence or absence of intranodal cystic necrosis; 10) echogenic hilus: present or absent; 11) 3D Doppler vascular pattern: A (absent/hilar/central/treelike pattern) or B (aberrant/peripheral pattern). In our prediction model for cervical metastatic lymph nodes, the significance of continuous variables, including patient age, L, S, and S/L ratio, was assessed using Student t Diagnosis TABLE I. Final Diagnosis of 139 Cervical Lymph Nodes. Metastatic nodal disease 37 Oral cancer 6 Lung cancer 5 Oropharyngeal cancer 3 Nasopharyngeal carcinoma 3 Esophageal cancer 3 Laryngeal cancer 2 Hypopharyngeal cancer 2 Unknown primary carcinoma 5 Miscellaneous* 8 Benign nodal disease 102 FNAB 79 Reactive lymphadenopathy 15 Tuberculosis 5 Schwannoma 2 Kikuchi disease 1 *Including prostate cancer, gall bladder cancer, and metastatic lymph nodes without primary cancer survey. FNAB 5 fine needle aspiration biopsy. No. test. The association between all statistically significant (P <.05) continuous variables and metastatic nodal disease was evaluated using the receiver operating characteristic (ROC) curve, and an optimal cutoff point was determined. These variables were converted into categorical variables according to the cutoff point to establish accuracy. The associations among categorical variables, including the converted continuous variables mentioned (patient sex, side, margin, echogenic level, internal echo, central necrosis, echogenic hilus, and vascular pattern), were then evaluated using the v 2 test. Variables were eligible for entry into the multiple logistic regression model if they were associated significantly with nodal metastasis following univariate analysis. Coefficient parameters were calculated to construct the multivariate formula that estimated the relationship between the independent variables and outcome prediction, when all statistically nonsignificant (P >.05) variables had been eliminated from the multivariate model. Calibration of the prediction model was assessed using the Hosmer Lemeshow goodness-of-fit test. The association between the model and metastatic nodal disease was reevaluated using the ROC curve, and an optimal cutoff point was determined at the point of greatest sensitivity and the corresponding smallest 1 2 specificity value. All statistical analyses were performed using SPSS software, version 15.0 (SPSS, Chicago, IL). Another independent cohort of 27 consecutive patients of untreated head and neck cancer was initially evaluated for nodes using the study prediction model during May 2012 to October These patients underwent a neck dissection operation performed by a single doctor (M.-K.C.) following MRI or CT. Images were reviewed by three neuroradiologists with 10 years of experience in head and neck radiology, blinded to the clinical information. The accuracy of level by level and stage by stage initial staging was evaluated according to postoperative pathological results. RESULTS In this study, we evaluated 135 patients with 139 suspicious cervical lymph nodes. Our cohort contained 73 men and 62 women, with a mean age of 39.6 years (range years), who received ultrasonographic examination prior to US-FNA or excision biopsy. Table I lists the final diagnoses for the 139 suspicious cervical lymph nodes (37 metastatic carcinomas and 102 benign). Carcinoma nodal metastasis of any origin was included because of the similar ultrasonographic features to the head and neck cancer. Two schwannomas were enrolled in our study due to difficulty in differentiating them from lymphadenopathy by initial physical examination and ultrasonography. We converted all continuous variables that were significantly associated with nodal metastasis into categorical variables for division into two groups, including age (<40 years, 40 years), L (<25 mm, 25 mm), S (<10 mm, 10 mm), and S/L ratio (<0.5, 0.5), according to their ROC curve and optimal cutoff point (data not shown). The results from v 2 test indicated that sex, age, axis, L, S, S/L ratio, margin, echogenic level, internal echo, internal matting, central necrosis, echogenic hilus, and vascular pattern differed significantly between malignancy and benignity (P <.05; Table II). Following univariate analysis, we included all significant parameters in multivariate analysis, which identified the independent ultrasonographic features age, S/L ratio, internal 3038

3 Fig. 1. (A) A 44-year-old man with tongue cancer and neck metastasis. Gray-scale ultrasonogram showing hypoechoic homogenous areas with internal matting. (B) A 51-year-old woman with benign neck lymphadenopathy. Gray-scale ultrasonogram showing hypoechoic homogenous areas with partial internal matting (less than half the area). (C) A 28-year-old man with benign neck lymphadenopathy. Gray-scale ultrasonogram showing hypoechoic homogenous areas without internal matting. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.] matting, and vascular pattern as the strongest predictors of nodal metastasis. The Hosmer Lemeshow goodness-offit test provided no evidence of lack of fit in the selected model (P 5.895). A scoring scale was then constructed using these significant variables and their positive integers proportionate to the estimates (Table III): Score age ð40 years 3 1;<40 years 30Þ S=L ratio ð0:5 3 1; 0:530Þ13 3internal matting ðpresent 31; absent vascular pattern aberrant =peripheral 31; others 30 We regarded a nodal lymph node as malignant when the predictive score was 5 according to the Youden index (sensitivity 1 specificity 2 1) and the best cutoff point (ROC curve area , P <.0001). We then simplified the model to a prediction model with major (presence of internal matting and vascularity pattern) and minor (age 40 years and S/L ratio 0.5) sign categories. Cervical lymph node was regarded as metastasis with the presence of one major factor and any of the other factors, according to the results from the described scoring scale calculations. We identified the predictive model as having 91.9% sensitivity, 88.2% specificity, 73.9% positive predictive value (PPV), 96.8% negative predictive value (NPV), and 89.2% accuracy. We then evaluated another independent cohort of 27 consecutive patients with untreated head and neck cancer (15 patients with and 12 patients without cervical lymph node metastasis) using the ultrasonographic prediction model. When we compared the final pathologic results, the accuracy of level by level initial staging was 92.5% (123 of 133), with an understaged rate of 6.0% (eight of 133) and an overstaged rate of 1.5% (two of 133) in 133 levels. Table IV compares the results stage by stage with the different imaging modalities. The rate of correct N staging by ultrasonographic prediction model for patients was 92.6% (25 of 27) compared with 55.0% for MRI (11 of 20) and 42.9% for CT (three of seven). DISCUSSION In this study, we constructed a rapid and practical prediction model for the evaluation of metastatic cervical lymphadenopathies using ultrasonographic examination. Our results indicate that the model is a useful tool for the initial staging of patients with head and neck cancer and could potentially facilitate treatment planning. Physicians generally use high-resolution US for assessment of cervical lymph nodes, and its images have become more precise with rapid technological advances. Previous studies have identified ultrasonographic features including L, S, S/L ratio, hilar echogenicity, internal echogenicity, margin, intranodal necrosis, and vascular pattern as reliable factors for differentiation between benign and metastatic lymph nodes. 7,12 21 In our study, we identified S/L > 0.5, presence of internal matting, peripheral or aberrant vascular patterning, and patient age >40 years as independent factors that influence the prediction of cervical metastatic disease using logistic regression multiple analysis. Internal matting displayed the largest odds ratio (2.94). We determined that several metastatic lymph nodes were hyperechogenic to normal with coarse contents, regardless of the echogenic level and the internal echo. Previous studies have described similar features in tuberculous cervical adenitis, including intranodal cystic necrosis, pseudocysts, or caseous necrosis. 14,17 In other studies, coagulation necrosis was a characteristic of tuberculous and metastatic cervical lymph nodes. 16,22 In our opinion, after comparison with pathologic findings, this feature might be related to extensive tumor lymphangiogenesis 23 and microcystic necrosis in early metastatic lymph nodes, and follow central or cystic necrosis in late metastatic lymph nodes. In this study, we observed both intranodal matting and central necrosis in large or late metastatic lymph nodes. When evaluating metastatic lymphadenopathy, it is therefore important to consider this factor. However, tuberculous lymph nodes can sometimes mimic metastatic ones. Intranodal vascular patterning is another relevant factor during the evaluation of metastatic lymphadenopathy. Color Doppler ultrasonographic imaging has the advantage over CT, MRI, or PET/CT of ability to observe abnormal nodal vessels. However, the vascular patterning revealed by the 2D color Doppler mode can be misjudged, and interpretation can be problematic because of lack of complete gross lymph node images (Fig. 2A, 3A). To avoid this problem, we used a 3D color Doppler in 3039

4 TABLE II. Clinical and Ultrasonographic Features Associated With Metastatic Nodal Disease. Metastasis No, n Yes, n 5 37 Total, n Characteristic No. % No. % No. % P* Gender Female <.001 Male Age <40.5 years < years Long <24.6 mm < mm Short <10.05 mm < mm Long/short ratio < < Side Left Right Heterogeneous No <.001 Yes Signal Hypoechoic <.001 Isoechoic Matting Negative <.001 Positive Central necrosis Negative <.001 Positive Margin Well <.001 Ill Hilar Negative <.001 Positive D Doppler vascular pattern Aberrant/peripheral <.001 Others *Probability value by v 2 test. 3D 5 three-dimensional; mm 5 millimeter. power Doppler mode. Previous studies have used the 3D color Doppler technique for evaluation of breast mass or cardiac flow, as well as a pilot study on cervical lymphadenopathy. 23,24 Using this technique, the observer can easily perform and correctly determine vascular patterning using the image planes of the 3D dynamic phantom from any angle (Fig. 2B, 3B). In our study, we separated the lymph nodes into two groups according to their intranodal vascular pattern to facilitate evaluation. Our predictive tool displayed 87.1%/80.6% accuracy, 59.5%/47.8% sensitivity, and 97.1%/89.4% specificity. The reason for its poor sensitivity might be our inclusion of 3040

5 TABLE III. Results of Multiple Analysis and the Scoring Scale of the Prediction Model. Characteristic Estimate * OR 95% CI P Score Age L/S Matting < D Doppler vascular pattern TABLE IV. Numbers and Proportions of N Staging With the Different Imaging Modalities. Staging Ultrasonography, No./% MRI, No./% CT, No./% Total Correct 25/ /55.0 3/42.9 Understaged 1/3.7 2/10.0 1/14.2 Overstaged 1/3.7 7/35.0 3/42.9 CT 5computed tomography; MRI 5 magnetic resonance imaging. *Regression coefficient. 3D 5 three-dimensional; CI 5 confidence interval; L/S 5 long/short ratio; OR 5 odds ratio. some large severe central necrotic metastatic lymph nodes with avascular vascular patterning (seven of 37) into group A (benign). This drawback could be adjusted by other coefficient factors in the prediction model. In 2010, Liao et al. emphasized that important factors for evaluation in a malignant lymphadenopathy prediction model are patient age, S/L ratio, internal echo, and vascular pattern. 13 They created a predictive scoring scale and identified cervical lymphadenopathy as malignant (score 7). The accuracy of their scale (88.9%) was similar to that of our predictive tool; however, the coefficient parameters remain too complex to calculate. Wu et al. developed another scoring scale using the ultrasonographic features of patient age, vascular index, S, vascular pattern, and internal echo with positive integral estimated parameters. 12 For differentiation of lymphadenopathy, the scale provided 86.8% accuracy using a score 10 to indicate malignancy. However, quantification of the vascular index remained unstable and time-consuming because of thin-vessel caliber and aberrant vascular patterning, especially for the measurement of metastatic intranodal vascularity. Using our prediction model with major and minor sign categories, we were able to evaluate and obtain information on cervical lymph nodes within seconds. Fig. 2. A 35-year-old man with neck reactive lymphadenitis. (A) Twodimensional (2D) power Doppler mode ultrasonogram showing intranodal aberrant vascular patterning. (B) 3D power Doppler mode ultrasonogram showing extensive central treelike vascularity patterning within the same lymph node. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.] Fig. 3. A 49-year-old man with tongue cancer and neck metastasis. (A) Two-dimensional (2D) power Doppler mode ultrasonogram showing intranodal peripheral vascular patterning. (B) 3D power Doppler mode ultrasonogram showing aberrant vascularity patterning within the same lymph node. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.] 3041

6 In patients with head and neck cancer scheduled for neck dissection surgery or postoperative chemoradiotherapy, neck lymph node staging remains important and is typically based on imaging. Sumi et al. described how the performance of ultrasonography was significantly superior to that of CT for the depiction of cervical metastatic lymph nodes; however, its performance for the observation of deep cervical lymph nodes, such as those in the retropharyngeal space, was poor. 11 Stoeckli et al. reported that the sensitivity, specificity, PPV, NPV, and accuracy among CT, PET/CT, ultrasonography, and US-FNA, with regard to the endpoint N0 versus N 1, displayed nonsignificant differences. 9 However, US-FNA (69%) showed a higher correct level by level staging rate than PET/CT (63%), ultrasonography (62%), and CT (62%). US-FNA evaluation of every suspicious malignant lymph node is not feasible in initial staging, particularly in patients with advanced metastatic nodal disease. Our prediction model using a 3D color Doppler technique provided considerably higher accuracy in level by level initial staging (92.5%) and rate of correct staging of patients (92.6%) than US-FNA. Therefore, ultrasonography should represent the primary choice for the initial staging of metastatic cervical lymphadenopathy, combined with CT or MRI imaging for primary tumor analysis, in head and neck cancer. This study has several limitations. First, nodal lymphadenopathy from different levels have different normal characteristics (e.g., threshold size for pathology in level II is higher; shape of nodes in level I is rounder.) However, additional analysis of every nodal level was not available because of the small case number. Second, it is difficult to differentiate a large, irregularly shaped lymphadenopathy from several inseparable, extracapsularly spread lymph nodes by ultrasonography. This may lead to mistaken staging. Third, interobserver variation may have existed in ultrasonographic performance and interpretation. Last, only 27 patients with head and neck cancer underwent neck dissection and pathology correlation with preoperative imaging. Furthermore, it will be necessary to gather a large number of cases to make a future follow-up study more precise. CONCLUSION Ultrasonography provides an inexpensive and noninvasive examination technique for the diagnosis of cervical lymph nodes. In this study, we created a rapid prediction model with major (presence of internal matting and vascularity patterning) and minor (age 40 years and S/L ratio 0.5) sign categories. We regarded cervical lymph nodes as metastatic according to the presence of one major factor and at least one of any other factors. This prediction model using a 3D color Doppler technique can differentiate metastatic from benign lymph nodes, and provides a promising and reliable method for initial nodal staging in head and neck cancer. BIBLIOGRAPHY 1. Shah JP. Patterns of cervical lymph node metastasis from squamous carcinomas of the upper aerodigestive tract. Am J Surg 1990;160: Galer CE, Kies MS. Evaluation and management of the unknown primary carcinoma of the head and neck. J Natl Compr Canc Netw 2008;6: Cerezo L, Millan I, Torre A, Aragon G, Otero J. Prognostic factors for survival and tumor control in cervical lymph node metastases from head and neck cancer. A multivariate study of 492 cases. Cancer 1992;69: Kane SV, Gupta M, Kakade AC, D Cruz A. Depth of invasion is the most significant histological predictor of subclinical cervical lymph node metastasis in early squamous carcinomas of the oral cavity. Eur J Surg Oncol 2006;32: Mamelle G, Pampurik J, Luboinski B, Lancar R, Lusinchi A, Bosq J. Lymph node prognostic factors in head and neck squamous cell carcinomas. Am J Surg 1994;168: Dayanand SM, Desai R, Reddy PB. Efficiency of ultrasonography in assessing cervical lymph node metastasis in oral carcinoma. Natl J Maxillofac Surg 2010;1: Toriyabe Y, Nishimura T, Kita S, Saito Y, Miyokawa N. Differentiation between benign and metastatic cervical lymph nodes with ultrasound. Clin Radiol 1997;52: Ali S, Tiwari RM, Snow GB. False-positive and false-negative neck nodes. Head Neck Surg 1985;8: Stoeckli SJ, Haerle SK, Strobel K, Haile SR, Hany TF, Schuknecht B. Initial staging of the neck in head and neck squamous cell carcinoma: a comparison of CT, PET/CT, and ultrasound-guided fine-needle aspiration cytology. Head Neck 2012;34: Akkas BE, Demirel BB, Vural GU. Clinical impact of 18F-FDG PET/CT in the pretreatment evaluation of patients with locally advanced cervical carcinoma. Nucl Med Commun 2012;33: Sumi M, Ohki M, Nakamura T. Comparison of sonography and CT for differentiating benign from malignant cervical lymph nodes in patients with squamous cell carcinoma of the head and neck. AJR Am J Roentgenol 2001;176: Wu CH, Lee MM, Huang KC, Ko JY, Sheen TS, Hsieh FJ. A probability prediction rule for malignant cervical lymphadenopathy using sonography. Head Neck 2000;22: Liao LJ, Wang CT, Young YH, Cheng PW. Real-time and computerized sonographic scoring system for predicting malignant cervical lymphadenopathy. Head Neck 2010;32: Ahuja A, Ying M. Sonography of neck lymph nodes. Part II: Abnormal lymph nodes. Clin Radiol 2003;58: Khanna R, Sharma AD, Khanna S, Kumar M, Shukla RC. Usefulness of ultrasonography for the evaluation of cervical lymphadenopathy. World J Surg Oncol 2011;9: Ahuja A, Ying M, Yang WT, Evans R, King W, Metreweli C. The use of sonography in differentiating cervical lymphomatous lymph nodes from cervical metastatic lymph nodes. Clin Radiol 1996;51: Ahuja A, Ying M, Evans R, King W, Metreweli C. The application of ultrasound criteria for malignancy in differentiating tuberculous cervical adenitis from metastatic nasopharyngeal carcinoma. Clin Radiol 1995;50: Gupta A, Rahman K, Shahid M, et al. Sonographic assessment of cervical lymphadenopathy: role of high-resolution and color Doppler imaging. Head Neck 2011;33: Dragoni F, Cartoni C, Pescarmona E, et al. The role of high resolution pulsed and color Doppler ultrasound in the differential diagnosis of benign and malignant lymphadenopathy: results of multivariate analysis. Cancer 1999;85: Wu CH, Hsu MM, Chang YL, Hsieh FJ. Vascular pathology of malignant cervical lymphadenopathy: qualitative and quantitative assessment with power Doppler ultrasound. Cancer 1998;83: Steinkamp HJ, Teichgraber UK, Mueffelmann M, Hosten N, Kenzel P, Felix R. Differential diagnosis of lymph node lesions. A semiquantitative approach with power Doppler sonography. Invest Radiol 1999;34: Ahuja AT, Ying M. Sonographic evaluation of cervical lymph nodes. AJR Am J Roentgenol 2005;184: Tsujino H, Jones M, Shiota T, et al. Real-time three-dimensional color Doppler echocardiography for characterizing the spatial velocity distribution and quantifying the peak flow rate in the left ventricular outflow tract. Ultrasound Med Biol 2001;27: Carson PL, Fowlkes JB, Roubidoux MA, et al. 3-D color Doppler image quantification of breast masses. Ultrasound Med Biol 1998;24: