Tissue Block and Digital Slide Registration Method for Microscopic Samples

With the digitalization of routine histology and pathology, there has emerged a need for automated histological instruments such as automated microdissectors, staining-coverslipper machines, and high-performance slide microscope scanners. Essential to the seamless integration of these tools are registration solutions that restore the tissue orientation lost between different workflow stages. In this way, samples can be evaluated together, accelerating and facilitating the diagnosis process. Image registration is one of the most researched topics today, with state-of-the-art machine learning solutions like SuperPoint and SuperSet effectively able to match similar images. However, they struggle to find correspondences in a multimodal environment between tissue block and section images. Our proposed solution addresses this issue by decomposing the registration process into tissue recognition and the matching of tissue regions.