AI-generated from publicly available materials.EPFL has unveiled spora, a groundbreaking open tissue atlas aimed at enhancing AI research in cancer diagnostics and treatment.
Launched in collaboration with institutions in Zurich and Geneva, spora is touted as the largest freely accessible spatial proteomics atlas in the world. This innovative resource contains over 13,500 multiplex images derived from nearly 5,500 patients, encompassing data from 14 different tissue types and utilizing five imaging technologies. The comprehensive dataset not only includes spatial proteomics images but also pairs with histopathology images when available, facilitating a better understanding of cellular arrangements and interactions within tissues.
The atlas is structured around three main components: spora[data], which provides standardized images and annotations; spora[io], a Python library for data preparation and cohort combination; and spora[bench], a benchmarking suite designed for tasks like cell identification and clinical variable prediction. This framework allows researchers to validate findings across different studies and assess the applicability of AI models trained on diverse datasets.
Led by Charlotte Bunne of EPFL's AIMM group, spora is part of a broader initiative under the National AI Initiative for Precision Oncology. The project's ambition is to translate insights from extensive tissue sample analysis into personalized treatment strategies for patients. Future plans include expanding the atlas to cover additional tissues and diseases, further solidifying its role in advancing cancer research through AI.