A dual branch based stitching method for whole slide hyperspectral pathological imaging

Abstract

Hyperspectral imaging technology integrates spatial information with broadband spectral data that extends beyond the visible spectrum, enabling in-depth analysis of spectral signatures unique to various tissue components. Whole slide image is an important media for pathologists to make diagnosis and image stitching is a basis technology for whole slide image. However, very few researches focus on microscopic hyperspectral image stitching due to its characteristics of memory-intensive and high-quality image feature sparsity. To address these limitations, we propose a High-quality Hyperspectral whole slide Image Stitching Method (HHISM) for pathological sample. Considering that color image can offer high spatial resolution, we introduce a color image guided dual branch method for hyperspectral image stitching. The proposed strategy with color-hyperspectral pair as model input works with common spectral scanning hyperspectral microscope hardware to improve data acquisition efficiency. To further enhance performance, we incorporate inter-branch data sharing module, enabling information exchange between color image stitching and hyperspectral image stitching branches. This interaction can enhance the quality of hyperspectral whole slide image stitching. We have conducted comprehensive experiments on three kinds of samples to evaluate the effectiveness of our proposed method. Experimental results on H&E samples demonstrate that our method outperforms the state-of-the-art medical image stitching methods in terms of quality and hardware-software interaction.