Predicting Pre- and Post-Diagnostic Depression in Women with Abnormal Pap Screening Tests: A Neural Network Approach

利用神经网络方法预测宫颈涂片检查异常女性的诊断前和诊断后抑郁症

阅读:1

Abstract

(1) Background: After receiving an abnormal Papanicolaou smear result, very often women fail to adhere to further procedures due to depression. Using a neural network approach, this research aimed to predict pre- and post-diagnostic depressive symptoms in women with abnormal Pap screening tests. (2) Methods: The study was conducted at the Clinical Center of Kragujevac, Serbia, among 172 women with a positive Pap screening result before and after diagnostic procedures (colposcopy/biopsy/endocervical curettage). Just before and 2 to 4 weeks after the diagnostic procedures, women filled out a socio-demographic questionnaire and the Hospital Anxiety and Depression Scale (HADS). Multilayer perceptron neural networks were modeled. (3) Results: Depression was present in 37.2% of women before diagnostic procedures and in 48.3% after. Feature selection showed four variables that correlated with depression before diagnostic procedures-anxiety (according to the HADS), depression according to the CESD scale, worry score on the POSM scale and use of sedatives. Model for predicting pre-diagnostic depression yielded an accuracy of 79.41%, with a value of 0.842 for area under the receiver operating characteristic curve (AUROC). The HADS anxiety score, place of residence and CESD score were the most important attributes for predicting post-diagnostic depression, with an ANN model accuracy of 88.24% and AUROC 0.939. (4) Conclusions: This research revealed a possible way of predicting depression occurrence in those women who received a positive Pap screening test and who are undergoing follow-up diagnostics, aiding medical doctors in the provision of successful and on-time psychological assistance.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。