Deep Learning assisted Cervical Cancer Classification with Residual Skip Convolution Neural Network (Res _ Skip _ CNN)- based Nuclei segmentation on Histopathological Images

Abstract

Cervical cancer (CC) remains the second most typical cancer in women internationally having a death too of sixty percent. CC starts with nil obvious symptoms and possesses a lengthy latency turning initial diagnosis via periodical health checks extremely significant. The problem of automatic technique to identify CC is proffered for enhancing the detection's precision. The CC histology source images will be required for employing image preprocessing to lessen the effect resulting from the images noise alongside the effect upon succeeding accurate feature extraction resulting in impertinent background. The images will be segmented employing the Residual Skip Convolution Neural Network (Res_Skip_CNN) centered upon 4 renowned discriminative attributes: i) nuclei's proportion to the cytoplasm, ii) nuclei's diameter, iii) shape factor, iv) nuclei's roundness. The random forest classification paradigm would be considered as an input to the cervix's segmented image having the epithelium layer for aiding the diagnostician in CC detection. We assess the execution of the proffered Res_Skip_CNN with RF (Res_Skip_CNN-RF) classifier opposing the Herley datasets employing a variable classes quantity while doing image classification. Consequently, it is observed that the proffered Res_Skip_CNN-RF attains 95% of Accuracy, 93% of Precision, 89% of Recall, and 85% of Fl-Score.