Predict the Therapeutic Effect of Bevacizumab Treatment Using a Channel Attention Convolution Neural Network
Hongtao Xu · Highlights in Science Engineering and Technology · 2022
One of the malignancies with the highest mortality rates among women worldwide is ovarian cancer. Epithelial ovarian cancer (EOC) is the most common kind of ovarian cancer which takes ~90% of ovarian cancer patients. peritoneal serous surface papillary carcinoma (PSPC) is rare cancer whose incident rate is 7% in women. Bevacizumab has been used as a monotherapy along with chemotherapy to treat advanced EOC and PSPC. Bevacizumab has a significant effect on chemotherapy, however, due to the high cost and side effects of the bevacizumab, how to predict the therapeutic effect of Bevacizumab treatment is very important. In this paper, the author uses the proposed attention module, ECA, embedding to the ResNet, compose as ECA-Net, to predict the treatment effect of the bevacizumab in current tissue is effective or invalid through the histopathological image. As a result, the ECA-Net gained novel performance, scoring highly on several evaluation metrics. Specifically, the classification accuracy of the ECA-Net is 94.54% and the f1 score is 95.00%. Bevacizumab is pricey and has side effects, the classification model will forecast its therapeutic impact. In this situation, the experiment will assist the gynecologist in selecting the best course of therapy while also saving money.