Whole Slide Images Based Cervical Cancer Classification Using Self-Supervised Learning and Multiple Instance Learning

Most whole-slide picture classification systems now rely on manual pixel-level annotations, which are delicate and time-consuming, and necessitate the annotation of specialized topic expertise. We propose employing self-supervised learning and multiple instances learning to handle large WSI datasets with only the reported diagnoses as labels to address this issue. Here we use a machine learning technique i.e.  K-Nearest Neighbors (KNN) and the deep neural network i.e., convolutional neural network that showed better performance when compared to KNN and the features learned by CNN are better for classification applications.

Keywords: Whole Slide Images, KNN algorithm, Convolutional neural network, Features, Cervical Cancer.