Low-Magnification MUSE Imaging Matches High-Res for Breast Cancer Margin Detection

A recent study featured on arXiv (2608.11317) investigated the practicality of low-magnification fluorescence imaging for detecting breast cancer margins through texture analysis and deep learning techniques. The research contrasted microscopy with ultraviolet surface excitation (MUSE) images at magnifications of 4x and 10x for patch-level classification. Both the texture analysis (TA) using local binary patterns (LBP) and the deep learning (DL) approach utilizing a Vision Transformer (ViT) model demonstrated comparable performance across the two magnifications. The DL method achieved 96.30% sensitivity, 100% specificity, and 98.18% accuracy at both levels. While the TA method showed 100% specificity at 4x and 100% sensitivity at 10x, both methods maintained identical accuracy (96.67%). The findings indicate that 4x imaging can provide similar diagnostic accuracy to 10x, potentially facilitating quicker and more economical intraoperative margin evaluations. This study underscores MUSE’s potential as an effective technique for assessing surgical margins in breast cancer operations, with low-magnification imaging presenting a practical benefit without sacrificing accuracy.