The lazy way of pathological images analysis in the era of large models

Histopathological slides remain the gold standard for disease diagnosis and the study of tissue architecture and pathology, offering an exceptionally rich yet often underutilized source of biomedical information. Unlike single-cell datasets, pathology slides are readily available in clinical workflows and comparatively inexpensive to generate.

Recent advances in foundation models have transformed our ability to extract meaningful insights from image data, leading to a surge of high-impact studies leveraging pathology images for survival prediction, subtyping, spatial integration, and more. However, for many bioinformaticians and systems biologists, the field of image analysis still feels inaccessible because of steep technical barriers: diverse file formats, fragmented tools, and a lack of intuitive frameworks for exploration and analysis.

Is there a way to analyze pathology images as intuitively as we analyze single-cell data? The answer is yes!

In this talk, I will introduce LazySlide, an open-source platform for pathology image analysis specifically designed for users scientists already comfortable with tools like Scanpy and AnnData. LazySlide supports multiple image formats and enables interaction with large models. This talk will demonstrate how LazySlide empowers scientists to “analyze images with text” and incorporate rich histopathological information into multi-omics workflows, without needing to become image processing experts. Whether you’re studying cancer progression, tissue aging, or immune landscapes, LazySlide unlocks a new dimension of spatial and morphological context at scale.

Get started with LazySlide in no time: https://github.com/rendeirolab/LazySlide