LEAF-Mamba: Feature Adaptation and Fusion for Whole Slide Image Analysis

LEAF-Mamba is a new framework for whole slide image analysis that jointly adapts and fuses frozen pretrained features. It uses a reconstruction‑regularized adapter to learn compact, task‑specific representations and combines parallel Mamba and convolutional streams to capture both long‑range dependencies and local interactions, with a gate that modulates their joint output before slide‑level pooling. Experiments show that this approach improves diagnostic classification and survival prediction over strong multiple‑instance learning baselines, and component analyses confirm the benefits of the adaptation and fusion modules.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Computer Vision.