A hybrid deep learning approach for cervical cancer segmentation and classification

One of the most common cancers worldwide is cervical cancer and is ranked as fourth among all gynaecological malignancies. In this research, a model named fractional gorilla chimp optimisation-based Shepard convolutional neural network (FGChO-based ShCNN) is developed for the classification of cervical cancer. Here, the image denoising is done by the adaptive bilateral filter. Furthermore, the U-Net++ model with gorilla chimp optimisation (GChO) is used for segmentation purposes. After that, the necessary features like grey level co-occurrence matrix (GLCM) features and texture features are extracted. Then, the automatic cervical cancer classification is executed with FGChO-based ShCNN wherein the ShCNN undergoes training using FGChO. The proposed FGChO-based ShCNN has the accuracy of 93.1%, false positive rate (FPR) of 0.055, positive predictive value (PPV) of 90.5%, negative predictive value (NPV) of 89.9%, true negative rate (TNR) of 94.5%, and true positive rate (TPR) of 92.4%.