Comparison of diagnostic usefulness of predictive models in preliminary differentiation of adnexal masses

The purpose of this study was to compare prognostic models evaluating the probability of an ovarian cancer occurrence based on a number of clinical and ultrasonographic data in women with adnexal masses. A total of 686 women with adnexal masses underwent the examinations between 1994 and 2002. The recorded parameters included: age, menopausal status, body mass index, the grayscale and Doppler ultrasonographic examination, and selected markers concentration levels. In order to find the best combination of features, which significantly influences the probability of malignancy, stepwise logistic regression analysis, as well as artificial neural network, was used. The diagnostic efficiency of received models was estimated and compared using receiver-operating characteristics (ROC) curve. The results indicate that 431 and 255 patients had a benign and malignant ovarian tumor, respectively. Application of stepwise logistic regression analysis revealed statistically significant importance of eight features. The sensitivity and specificity for the received model were 65.71% and 77.59%, respectively. Three-layer perceptron network shows 13 features as significant predictors of malignancy. The network gave a sensitivity of 85.7% and specificity of 93.1%. Comparison of area under ROC curve for received models was 0.9679 vs 0.9716. Prognostic values of the analyzed neural model are not optimal but seem to surpass logistic regression model in terms of the predictive possibilities.

[1]  M. Zweig,et al.  Receiver-operating characteristic (ROC) plots: a fundamental evaluation tool in clinical medicine. , 1993, Clinical chemistry.

[2]  T. Bourne,et al.  Distinguishing the benign and malignant adnexal mass: an external validation of prognostic models. Letters to the Editor , 2001 .

[3]  David W. Hosmer,et al.  Applied Logistic Regression , 1991 .

[4]  J Vandewalle,et al.  Artificial neural network models for the preoperative discrimination between malignant and benign adnexal masses , 1999, Ultrasound in obstetrics & gynecology : the official journal of the International Society of Ultrasound in Obstetrics and Gynecology.

[5]  J. Shepherd.  Revised FIGO staging for gynaecological cancer , 1989, British journal of obstetrics and gynaecology.

[6]  D. Oram,et al.  A risk of malignancy index incorporating CA 125, ultrasound and menopausal status for the accurate preoperative diagnosis of ovarian cancer , 1991, British journal of obstetrics and gynaecology.

[7]  L. Valentin,et al.  Gray scale sonography, subjective evaluation of the color Doppler image and measurement of blood flow velocity for distinguishing benign and malignant tumors of suspected adnexal origin. , 1997, European journal of obstetrics, gynecology, and reproductive biology.

[8]  G Gacci,et al.  Predicting ovarian malignancy: application of artificial neural networks to transvaginal and color Doppler flow US. , 1999, Radiology.

[9]  B. Mol,et al.  Distinguishing the benign and malignant adnexal mass: an external validation of prognostic models. , 2001, Gynecologic oncology.

[10]  J. Hanley,et al.  A method of comparing the areas under receiver operating characteristic curves derived from the same cases. , 1983, Radiology.

[11]  P. Lavin,et al.  Comparison of Serum CA 125, Clinical Impression, and Ultrasound in the Preoperative Evaluation of Ovarian Masses , 1988, Obstetrics and gynecology.

[12]  A. Tailor,et al.  Sonographic prediction of malignancy in adnexal masses using multivariate logistic regression analysis , 1997, Ultrasound in obstetrics & gynecology : the official journal of the International Society of Ultrasound in Obstetrics and Gynecology.