RESPAN: A Deep Learning Pipeline for Accurate and Automated Restoration, Segmentation, and Quantification of Dendritic Spines.
Download the latest version of the RESPAN Windows executable here. This file is zipped using 7zip.
Please allow a minimum of 20GB of disk space for the software and ensure there is sufficient space for processing your data. RESPAN includes lossless compression of image files to ensure a minimal footprint for generated results and validation images.For more information and troubleshooting, please refer to our user guide. Pretrained models for a variety of image modalities are available for download here with addition information on each model noted below.| Segmentation Model | Modality | Resolution | Annotations | Details |
|---|---|---|---|---|
| Model 1 | Spinning disk and Airyscan/laser scanning confocal microscopy | 65 x 65 x 150nm | spines, dendrites, and soma | 112 datasets, including restored and raw data and additional augmentation |
| Model 2 | Spinning disk confocal microscopy | 65 x 65 x 65nm | spines, necks, dendrites, and soma | isotropic model, 7 datasets, no augmentation |
| Model 3 | Two-photon in vivo confocal microscopy | 102 x 102 x 1000nm | spines and dendrites | 908 datasets, additional augmentation |
For detailed protocols using RESPAN, please refer to our manuscript.
- Comprehensive Integration: RESPAN uniquely integrates image restoration, axial resolution enhancement, and deep learning-based segmentation into a single, user-friendly application.
- 3-Dimensional Analysis: 3D information is efficiently utilized at all stages of the pipeline, ensuring improved performance over approaches limited to 2D or a combination of 2D and 3D techniques for quantification.
- In Vivo Spine Tracking: RESPAN has been demonstrated to successfully track spines autonomously across time in 3D in challenging in vivo two-photon imaging conditions.
- Increased Accuracy: By enhancing image quality prior to segmentation, RESPAN significantly improves the accuracy of spine detection and morphological measurements.
- User-Friendly Deployment and Interface: A ready-to-run application with a graphical user interface allows users without programming skills to perform advanced analyses.
- Built-in Validation Tools: RESPAN includes tools for validating results against ground-truth data, promoting scientific rigor and reproducibility.
- Model Training: RESPAN includes tabs in the graphical user interface that allow training of CSBDeep, Self-Net, and nnU-Net models, which normally require separate environments using Tensorflow and PyTorch, removing a significant barrier to training and utilizing custom models.
If you use RESPAN as part of your research, please cite our work using the reference below:
Sergio B. Garcia, Alexa P. Schlotter, Daniela Pereira, Franck Polleux, Luke A. Hammond. (2024) RESPAN: An Automated Pipeline for Accurate Dendritic Spine Mapping with Integrated Image Restoration. bioRxiv. doi: https://doi.org/10.1101/2024.06.06.597812RESPAN has been used in the following publications:
- Baptiste Libé-Philippot, Ryohei Iwata, Aleksandra J. Recupero, Keimpe Wierda, Sergio Bernal Garcia, Luke Hammond, Anja van Benthem, Ridha Limame, Martyna Ditkowska, Sofie Beckers, Vaiva Gaspariunaite, Eugénie Peze-Heidsieck, Daan Remans, Cécile Charrier, Tom Theys, Franck Polleux, Pierre Vanderhaeghen (2024) Synaptic neoteny of human cortical neurons requires species-specific balancing of SRGAP2-SYNGAP1 cross-inhibition. Neuron. https://doi.org/10.1016/j.neuron.2024.08.021.
Updating soon with information on how to create the environments for running RESPAN directly in Python.