A deep learning-based, threshold-agnostic, subpixel-accurate 2D and 3D spot detection method for fluorescence microscopy and spatial transcriptomics.
Spotiflow is a deep learning-based software tool for detecting spot-like structures in fluorescence microscopy images and 3D volumes with subpixel accuracy. It solves the problem of accurate, automated transcript detection in spatial transcriptomics and general spot detection in bioimaging without requiring manual threshold setting. The method uses stereographic flow regression to precisely localize spots.
Bioimage analysts, computational biologists, and researchers working with fluorescence microscopy data, particularly in spatial transcriptomics (e.g., multiplexed FISH) or any field requiring precise detection of spot-like signals in 2D or 3D.
Developers choose Spotiflow for its combination of subpixel accuracy, threshold-agnostic design, and ease of integration into existing workflows via its CLI, API, and plugins for popular platforms like Napari and QuPath. Its availability of pre-trained models allows for immediate use without extensive training.
Accurate and efficient spot detection for microscopy data
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Achieves precision finer than a pixel, critical for spatial transcriptomics where exact transcript localization is needed, as highlighted in the publication.
Eliminates manual intensity threshold selection, reducing human error and variability in spot detection workflows.
Offers CLI, Python API, Docker, and plugins for Napari, QuPath, and TrackMate, making it easy to embed into existing bioimage pipelines.
Includes ready-to-use models like 'general' and 'hybiss' for common microscopy modalities, allowing immediate use without training.
Requires PyTorch installation with specific CUDA versions for GPU support and on Windows, additional Visual Studio Build Tools, which can be daunting for beginners.
Deep learning models demand significant GPU memory and computational power, potentially limiting use on standard or older hardware.
The conda-forge version may be outdated compared to pip, as noted in the installation section, leading to potential compatibility issues.