PURPOSE: To identify a combination of sonographic features that best predicts ovarian malignancy. METHODS: Subjects included 249 women who had a transvaginal sonogram for a pelvic mass at Brigham and Women's Hospital between December 2005 and February 2010. Subjects underwent surgery for removal of the mass and pathologic diagnosis was available. Images were reviewed retrospectively by one sonologist blinded to diagnosis and clinical information. Twelve sonographic features were scored for each mass. The dataset was divided into training (n = 149) and testing (n = 100) sets. Within the training set, a stepwise logistic regression was used to weigh each variable and combination of features to identify those associated with malignancies. Using the results from the logistic regression analyses, we created a three-level risk stratification that was applied to the sonograms of subjects in the testing set to assess its ability to distinguish benign lesions from invasive and borderline cancers. RESULTS: High risk lesions included all masses with internal vascularity. In our testing set, this feature was present in 9 out of 12 (75%) invasive cancers, 1 out of 6 (16.7%) borderline lesions, and 9 out of 82 (11%) benign masses. The intermediate risk level included lesions with a thick wall or thick septa without internal blood flow. This combination of features identified one additional invasive cancer and 5 out of 6 (83.3%) borderline tumors. Masses with low risk features had a 2/49 (4.0%) incidence of malignancy. CONCLUSIONS: In the absence of high or intermediate risk sonographic features, the risk of malignancy is low.
PURPOSE: To identify a combination of sonographic features that best predicts ovarian malignancy. METHODS: Subjects included 249 women who had a transvaginal sonogram for a pelvic mass at Brigham and Women's Hospital between December 2005 and February 2010. Subjects underwent surgery for removal of the mass and pathologic diagnosis was available. Images were reviewed retrospectively by one sonologist blinded to diagnosis and clinical information. Twelve sonographic features were scored for each mass. The dataset was divided into training (n = 149) and testing (n = 100) sets. Within the training set, a stepwise logistic regression was used to weigh each variable and combination of features to identify those associated with malignancies. Using the results from the logistic regression analyses, we created a three-level risk stratification that was applied to the sonograms of subjects in the testing set to assess its ability to distinguish benign lesions from invasive and borderline cancers. RESULTS: High risk lesions included all masses with internal vascularity. In our testing set, this feature was present in 9 out of 12 (75%) invasive cancers, 1 out of 6 (16.7%) borderline lesions, and 9 out of 82 (11%) benign masses. The intermediate risk level included lesions with a thick wall or thick septa without internal blood flow. This combination of features identified one additional invasive cancer and 5 out of 6 (83.3%) borderline tumors. Masses with low risk features had a 2/49 (4.0%) incidence of malignancy. CONCLUSIONS: In the absence of high or intermediate risk sonographic features, the risk of malignancy is low.
Authors: Jesus Rolando Delgado-Balderas; Maria Lourdes Garza-Rodriguez; Gabriela Sofia Gomez-Macias; Alvaro Barboza-Quintana; Oralia Barboza-Quintana; Ricardo M Cerda-Flores; Ivett Miranda-Maldonado; Hugo Mauricio Vazquez-Garcia; Lezmes Dionicio Valdez-Chapa; Mauro Antonio-Macedo; Michael Dean; Hugo A Barrera-Saldaña Journal: Genes (Basel) Date: 2018-07-11 Impact factor: 4.096