iHAST: Integrating Hybrid Attention for Super-Resolution in Spatial Transcriptomics


Xi Li (University of California, Irvine), Jing Zhang (Donald Bren School of Information and Computer Sciences, University of California, Irvine), Ziheng Duan (University of California, Irvine), Yi Dai (University of California, Irvine), Siwei Xu (Donald Bren School of Information and Computer Sciences, University of California, Irvine)
The 35th British Machine Vision Conference
PDFPosterVideo (Right click to download)

Abstract

Spatial transcriptomics (ST) technologies have revolutionized genomic research by enabling spatially-resolved gene expression profiling within their native contexts. These technologies provide a unique opportunity to explore the spatial heterogeneity of gene expression and create a detailed map of cellular function and organization. However, most current ST assays are limited to resolving cell clusters or multicellular structures instead of individual cells, leading to the merging of signals from multiple cell types. This averaging effect may obscure the spatial dynamics of diverse cell populations within complex tissues. To overcome this limitation, we conceptualize ST data as a specific type of image, where cells represent pixels and genes are analogous to channels. In response, we propose a novel arbitrary scale image super-resolution framework, named iHAST, aimed at enhancing the spatial resolution of ST data. To support this framework, we have compiled an extensive spatial transcriptomics dataset comprising over 800 examples for training and testing. This dataset encompasses a variety of tissue types and includes samples from different ST technology and platform. Our approach outperforms all state-of-the-art methods in this task.

Citation

@inproceedings{Li_2024_BMVC,
author    = {Xi Li and Jing Zhang and Ziheng Duan and Yi Dai and Siwei Xu},
title     = {iHAST: Integrating Hybrid Attention for Super-Resolution in Spatial Transcriptomics},
booktitle = {35th British Machine Vision Conference 2024, {BMVC} 2024, Glasgow, UK, November 25-28, 2024},
publisher = {BMVA},
year      = {2024},
url       = {https://papers.bmvc2024.org/0991.pdf}
}


Copyright © 2024 The British Machine Vision Association and Society for Pattern Recognition
The British Machine Vision Conference is organised by The British Machine Vision Association and Society for Pattern Recognition. The Association is a Company limited by guarantee, No.2543446, and a non-profit-making body, registered in England and Wales as Charity No.1002307 (Registered Office: Dept. of Computer Science, Durham University, South Road, Durham, DH1 3LE, UK).

Imprint | Data Protection