Abstract
BACKGROUND: Cervical cancer is a prevalent malignancy among women, often asymptomatic in early stages, complicating detection. OBJECTIVE: This study aims to investigate innovative techniques for early cervical cancer detection using a novel U-RCNNS model. METHODS: Cervical epithelial cell images stained with hematoxylin and eosin (HE) were analyzed using the U-RCNNS model, which integrates U-Net for segmentation and R-CNN for object detection, incorporating dilated convolution techniques. RESULTS: The U-RCNNS model significantly improved the accuracy of detecting and segmenting cervical cancer cells, with the enhanced Mask R-CNN showing notable advancements over the baseline model. CONCLUSION: The U-RCNNS model presents a promising solution for early cervical cancer detection, offering improved accuracy compared to traditional methods and highlighting its potential for clinical application in early diagnosis.