U-FISH: a fluorescent spot detector for imaging-based spatial-omics analysis and AI-assisted FISH diagnosis

Area of Science:

• Biomedical research
• Computational biology
• Medical imaging

Background:

• Imaging-based spatial-omics offers high-resolution biological insights but faces challenges in accurate signal spot identification.
• Existing methods struggle with consistency across diverse spatial-omics datasets.

Purpose of the Study:

• To develop a robust deep learning method for accurate and consistent signal spot detection in spatial-omics imaging.
• To improve the analysis of complex biological data from various spatial-omics techniques.

Main Methods:

• Developed U-FISH, a deep learning image enhancement technique for consistent spot detection.
• Created a comprehensive dataset of Fluorescence In Situ Hybridization (FISH) images from seven spatial-omics methods.
• Benchmarked U-FISH against existing methods for accuracy and generalizability.

Main Results:

• U-FISH demonstrated superior accuracy and generalizability in spot detection across diverse spatial-omics data.
• The method effectively processed and decoded complex 3D FISH data.
• U-FISH is the first spot detection software integrated with large language models, showing promise in AI-assisted diagnostics.

Conclusions:

• U-FISH provides a significant advancement for spatial-omics data analysis, enabling more reliable signal spot identification.
• The integration with large language models opens new avenues for AI-assisted diagnostics in spatial biology.
• This tool offers enhanced capabilities for biomedical discoveries using spatial-omics data.