Advanced Deep Learning for Histopathological Image Segmentation

Created using ChatSlide

This presentation delves into the problem of effective breast cancer segmentation in histopathology slides, addressing the complexity and manual challenges through advanced deep learning models. Leveraging a diverse dataset from multiple scanners with both dense and sparse annotations, it explores segmentation methodologies using adapted U-Net and Swin Unet architectures, highlighting comparative performances with nnU-Net achieving 79.59% accuracy. Special focus is placed on the high...