SimNorth: A novel contrastive learning approach for clustering prenatal ultrasound images
samweiss
This paper presents SimNorth, an unsupervised deep learning method for organizing non-standard fetal ultrasound images. By learning feature embeddings using a novel contrastive loss and clustering similar anatomical structures, SimNorth outperforms existing methods like Autoencoders, MoCo, and SimCLR in identifying meaningful image groups. The post SimNorth: A novel contrastive learning approach for clustering prenatal ultrasound images appeared first on Department of Psychiatry .