Xu Weize, Cai Huaiyuan, Zhang Qian, Wang Zhengze, Yang Jiajun, Wu Xiaofeng, Li Chengwen, Cui Chenghua, Liu Changzhi, He Jin, Mueller Florian, Dai Jinxia, Hao Chen, Ouyang Wei, Cao Gang
Faculty of Life and Health Sciences, Shenzhen University of Advanced Technology, 518107, Shenzhen, China.
College of Veterinary Medicine, Huazhong Agricultural University, Wuhan, China.
Genome Biol. 2025 Sep 1;26(1):261. doi: 10.1186/s13059-025-03736-x.
Imaging-based spatial-omics advances biomedical discoveries with subcellular resolution and high sensitivity, but accurately identifying signal spots from diverse images remains challenging. We develop U-FISH, a deep learning method that enhances images for consistent spot detection across various spatial-omics data. We establish a comprehensive FISH image dataset from seven spatial-omics methods. Benchmark analysis shows U-FISH has superior accuracy and generalizability compared to existing methods and can effectively decode 3D FISH data. U-FISH is the first spot detection software integrated with large language models, as demonstrated in AI-assisted diagnostics. Our study provides a valuable tool for spatial-omics analysis and diagnostics.
基于成像的空间组学以亚细胞分辨率和高灵敏度推动了生物医学发现，但从各种不同图像中准确识别信号点仍然具有挑战性。我们开发了U-FISH，这是一种深度学习方法，可增强图像以便在各种空间组学数据中进行一致的斑点检测。我们从七种空间组学方法建立了一个全面的荧光原位杂交（FISH）图像数据集。基准分析表明，与现有方法相比，U-FISH具有更高的准确性和通用性，并且可以有效地解码三维FISH数据。如在人工智能辅助诊断中所示，U-FISH是首个与大语言模型集成的斑点检测软件。我们的研究为空间组学分析和诊断提供了一个有价值的工具。
