Enhancing Cervical Pre-Cancerous Lesion Detection by Adaptive Transformer-Efficient Unet++ with Multiencoder from Colposcopy Images

Precancerous pathology analysis involves a special medical screening, such as a Pap smear for cervical cancer, which can identify the abnormal cells from the cervix and when there is a change in cells that can transform into cancer if it is not treated well. The Human Papillomavirus (HPV) is capable of causing cervical cancer and it needs a periodic examination to identify it at the starting stage because it is optimized as a deadly disease for women. Moreover, the analysis method can provide an exception for negative findings because of human negligence. Hence, the intelligence-based model is proposed as a new technology that can identify cervical cancer pathology. Initially, the colposcopy input images are retrieved from public websites. The gathered colposcopy image is given for the detection process and it is executed using Adaptive Transformer-Efficient UNet++ with Multi-Encoder (AT-EUNet++ME). This method is developed with graph convolutional neural network (GCNN), Visual Geometric Group19 (VGG19) and residual network (ResNet). Moreover, the reformulated Botox optimization (RBO) is a new upgraded optimization process and the introduced RBO can have the proposed AT-EUNet++ with a fine-optimized framework. In overall operation, the detection of cervical cancer is well achieved. Hence, the developed AT-EUNet++ME is compared with the existing system, thus the proposed model can acquire a highly effective performance. Here, the designed approach has attained 97% accuracy compared to other conventional methods in the detection performance, which can optimally enhance early diagnosis.