Data Analysis for Risk Prediction of Cervical Cancer Metastasis and Recurrence Based on DCNN-RF. Issue 1 (February 2021)

• Record Type: Journal Article
• Authors: Zhou, Xiaohong; Li, Weihong; Wen, Zhicheng
• Is Part Of: Journal of physics. Volume 1813:Issue 1(2021)
• Publication Date: 2021-02
• Subjects: Recurrence and metastasis of cervical cancer -- H&E pathological images -- Convolution neural network -- Random forest -- Transfer learning.; Physics -- Congresses; 530.5
• DOI: 10.1088/1742-6596/1813/1/012033
• Languages: English
• ISSNs: 1742-6588

Abstract

Abstract: In allusion to the problem of the low survival rate of cervical cancer metastasis and recurrence, combining the advantages of deep learning, a hybrid DCNN-RF method based on patched pathological image was proposed to predict the risk of metastasis and recurrence in cervical cancer patients. In order to improve the generalization ability of the model, according to the features, predicting result could be obtained from random forest, and the integration result was invoked as the result of haematoxylin and eosin pathological whole-slide images (WSI). The experimental results show that the model yielded an accuracy of 90.32% for prediction based on sliding window in cross-validation, and 0.83 AUC in the WSI.