The field of medical imaging is witnessing significant advancements in semi-supervised learning, driven by the need to overcome the limitations of scarce labeled training data. Researchers are exploring innovative approaches to leverage unlabeled data, including class-conditioned image translation, frequency prior guided matching, and synthesizing images to match pseudo labels. These methods have shown promising results in improving model performance and generalizability across different datasets and domains. Notably, the use of ensemble-based pseudo-labeling, uncertainty injection, and feature distillation has enabled robust classification and segmentation performance even with minimal labeled data. Overall, the field is moving towards developing more practical and effective solutions for medical imaging applications where annotation costs are prohibitive. Noteworthy papers include: SPARSE Data, Rich Results: Few-Shot Semi-Supervised Learning via Class-Conditioned Image Translation, which introduces a novel GAN-based semi-supervised learning framework. Frequency Prior Guided Matching: A Data Augmentation Approach for Generalizable Semi-Supervised Polyp Segmentation, which presents a novel augmentation framework that leverages polyp-specific structural properties. SynMatch: Rethinking Consistency in Medical Image Segmentation with Sparse Annotations, which proposes a framework that synthesizes images to match pseudo labels. PSScreen: Partially Supervised Multiple Retinal Disease Screening, which develops a model that leverages multiple partially labeled datasets to train a model for multiple retinal disease screening.