Pap Smear Image Segmentation and Classification Methods For Cervical Cancer Detection Using Machine Learning

Cervical cancer ranks as a major cause of death among women globally, making essential for effective treatment. The automated examination of Pap smear images through sophisticated machine learning and deep learning methods has notably enhanced diagnostic precision. This study conducts a comprehensive review of segmentation and classification methods utilized for Pap smear images, with an emphasis on UNet and Region-based Convolutional Neural Networks (R-CNN) for segmentation, Support Vector Machine (SVM) and Convolutional Neural Networks (CNN) for classification, along with Stochastic Gradient Descent (SGD) for optimization. U-Net has shown remarkable effectiveness in segmenting both cytoplasm and nuclei from Pap smear images, thus enhancing feature extraction for classification purposes [3]. Likewise, RCNN has proven to be adept at detecting and pinpointing abnormal cervical cells, yielding accurate segmentation outcomes [6]. In terms of classification, SVM has been frequently employed because of its ability to differentiate between normal and abnormal cells based on the features extracted [10]. CNN, utilizing deep learning, has exhibited exceptional performance in autonomously learning structured features for the precise classification of cervical cells [12]. Furthermore, SGD has been vital in optimizing deep learning architectures, enhancing both training efficiency and convergence rate [15]. The combination of these methods has resulted in improved accuracy and reliability in the detection of cervical cancer, promoting early diagnosis and minimizing the manual workload in pathology laboratories. This review sheds light on the latest developments in segmentation and classification approaches, underscoring their influence on enhancing cervical cancer screening.