Python-based software for generalized object instance segmentation from (cryo-)electron microscopy micrographs/tomograms. The software package is built on a general workflow where predicted semantic segmentation is used for instance segmentation of 2D/3D images.

Features

Robust and high-throughput semantic/instance segmentation of all microtubules:
- Supported file formats: [.tif, .mrc, .rec, .am]
- Supported modality: [ET, Cryo-ET]
- Supported Å resolution: [any best results in 1-40 Å range]
- 2D micrograph modality microtubule segmentation will come soon!
Robust and high-throughput semantic/instance segmentation of membranes:
- Supported file formats: [.tif, .mrc, .rec, .am]
- Supported modality: [EM, ET, Cryo-EM, Cryo-ET]
- Supported Å resolution: [all]
High-throughput semantic/instance segmentation of actin [Beta]
Fully automatic segmentation solution!
Napari plugin [Coming soon]
Cloud computing [Coming soon]

Citation

DOI [Microscopy and Microanalysis]

Kiewisz R., Fabig G., Müller-Reichert T. Bepler T. 2023. Automated Segmentation of 3D Cytoskeletal Filaments from Electron Micrographs with TARDIS. Microscopy and Microanalysis 29(Supplement_1):970-972.

Link: NeurIPS 2022 MLSB Workshop

Kiewisz R., Bepler T. 2022. Membrane and microtubule rapid instance segmentation with dimensionless instance segmentation by learning graph representations of point clouds. Neurips 2022 - Machine Learning for Structural Biology Workshop.

What’s new?

Full History

TARDIS-em v0.2.5 (2024-05-22):

Added actin segmentation
Improvement from Microtubule and Membrane prediction with updated models
Added option for scripting TARDIS predictions
Added visualization for semantic and instance predictions
TARDIS build in results visualization
Bug fixes
Documentation tutorials
Pypi and Conda releases
Re-trained DIST model using simulated datasets
Build 2 model for:
- filaments and general 2D structures
- 3D objects like membranes mitochondria LiDAR data etc.

Quick Start

For more examples and advanced usage please find more details in our Documentation

Install TARDIS-em:

pip install tardis-em

or

conda install tardis-em -c rrobert92

Verifies installation:

tardis

Filaments Prediction

3D Actin prediction

Full tutorial: 3D Actin Prediction

Usage:

recommended usage: tardis_actin [-dir path/to/folder/with/input/tomogram]
advance usage: tardis_actin [-dir str] [-out str] [-ps int] [-ct float] [-dt float]
                         [-pv int] [-px float] ...

2D Microtubule prediction

TBD

3D Microtubule prediction

Full tutorial: 3D Microtubules Prediction

Example:

Data source: Dr. Gunar Fabig and Prof. Dr. Thomas Müller-Reichert, TU Dresden

Usage:

recommended usage: tardis_mt [-dir path/to/folder/with/input/tomogram]
advance usage: tardis_mt [-dir str] [-out str] [-ps int] [-ct float] [-dt float]
                         [-pv int] [-px float] ...

Membrane Prediction

2D prediction

Full tutorial: 2D Membrane Prediction

Example:

Data source: Dr. Victor Kostyuchenko and Prof. Dr. Shee-Mei Lok, DUKE-NUS Medical School Singapore

Usage:

recommended usage: tardis_mem2d [-dir path/to/folder/with/input/tomogram] -out mrc_csv
advance usage: tardis_mem [-dir str] [-out str] [-ps int] ...

3D prediction

Full tutorial: 3D Membrane Prediction

Example:

Data source: EMPIRE-10236, DOI: 10.1038/s41586-019-1089-3

Usage:

recommended usage: tardis_mem [-dir path/to/folder/with/input/tomogram] -out mrc_csv
advance usage: tardis_mem [-dir str] [-out str] [-ps int] ...

Usage:

Documentation: