Vol 82, No 5 (2011)
ARTICLES
Application of neuron networks in the diagnostics of endometrial pathologies
Maria Kotarska, Agata Smoleń, Norbert Stachowicz, Jan Kotarski
Vol 82, No 5 (2011)
ARTICLES
Abstract
Abstract Aim: The aim of the study was to construct neuron networks utilizing selected risk factors and ultrasonographic (USG) examination parameters in a two-dimensional (2D) and three-dimensional (3D) presentation in relation to endometrial pathologies. Materials and methods: The following risk factors were statistically analyzed: age and menopausal status, parity, using hormonal replacement therapy (HRT), BMI, 2D USG of the endometrium (thickness, uterine artery blood flow indices) and 3D USG (volume, vascularization indices) in relation to the result of histopathological examination of the endometrial tissue in 421 women, aged 22-87 years, with abnormal bleeding from the uterus. The changes of the sensitivity and specificity in the applied models corresponding to changes of the limit value, were presented in the form of receiver operating characteristic curves (ROC) and the comparison of the values of the area under the curve (AUC). The threshold value for the obtained models was established and models of artificial neuron networks (ANN) were constructed on the basis of the ROC. Conclusion: Application of artificial neural networks in medicine has been developing rapidly. They have been applied in pre-surgical differentiation of ovarian tumors and other neoplasms. In case of endometrial carcinoma the degree of clinical usage of artificial neural networks has been limited, despite the fact that, from the mathematical point of view, the differentiation using neural networks would be much more precise than the one that could be obtained by chance.
Abstract
Abstract Aim: The aim of the study was to construct neuron networks utilizing selected risk factors and ultrasonographic (USG) examination parameters in a two-dimensional (2D) and three-dimensional (3D) presentation in relation to endometrial pathologies. Materials and methods: The following risk factors were statistically analyzed: age and menopausal status, parity, using hormonal replacement therapy (HRT), BMI, 2D USG of the endometrium (thickness, uterine artery blood flow indices) and 3D USG (volume, vascularization indices) in relation to the result of histopathological examination of the endometrial tissue in 421 women, aged 22-87 years, with abnormal bleeding from the uterus. The changes of the sensitivity and specificity in the applied models corresponding to changes of the limit value, were presented in the form of receiver operating characteristic curves (ROC) and the comparison of the values of the area under the curve (AUC). The threshold value for the obtained models was established and models of artificial neuron networks (ANN) were constructed on the basis of the ROC. Conclusion: Application of artificial neural networks in medicine has been developing rapidly. They have been applied in pre-surgical differentiation of ovarian tumors and other neoplasms. In case of endometrial carcinoma the degree of clinical usage of artificial neural networks has been limited, despite the fact that, from the mathematical point of view, the differentiation using neural networks would be much more precise than the one that could be obtained by chance.
Keywords
neural networks, uterine bleeding, endometrial disease, sonography
Title
Application of neuron networks in the diagnostics of endometrial pathologies
Journal
Ginekologia Polska
Issue
Vol 82, No 5 (2011)
Page views
493
Article views/downloads
624
Bibliographic record
Ginekol Pol 2011;82(5).
Keywords
neural networks
uterine bleeding
endometrial disease
sonography
Authors
Maria Kotarska
Agata Smoleń
Norbert Stachowicz
Jan Kotarski