Tissue Microscopy Image Data Analysis Suite

T-MIDAS was created with a focus on the reproducibility of batch image processing and quantification

• Batch processing pipelines for image format conversion, preprocessing, segmentation, ROI analysis
• Executable with a simple, text-based user interface
• Runs on any low-grade workstation with a single GPU
• Modular and simple codebase with few dependencies for easy maintenance
• Supported imaging modalities: Confocal microscopy, slidescanner, multicolor, brightfield
• Logs all your workflows and your parameter choices to a simple CSV
• You can fork this repository to adapt the batch processing scripts to your own image analysis workflows
• Quick installation

T-MIDAS is built on established image processing libraries. All dependencies are listed here. See acknowledgements and citations for further information.

Text-based User Interface

Feature overview

More detailed information is provided via text-based user interface.

[1] Image Preprocessing   
    [1] File Conversion to TIFF
        [1] Convert .ndpi
        [2] Export series from .lif, .czi, ...
        [3] Convert brightfield .czi
    [2] Cropping largest objects from images using mobile SAM
        [1] Slidescanner images (fluorescent, .ndpi)
        [2] Slidescanner images (brightfield, .ndpi)
        [3] Multicolor image stacks (.lif)
    [3] Extract intersecting regions of two images
    [4] Sample random image subregions
    [5] Enhance contrast of single color image using CLAHE
    [6] Restore images with Cellpose (single or multiple color channel, 2D or 3D, also time series)
    [7] Split color channels (2D or 3D, time series)
    [8] Merge color channels (2D or 3D, time series)
    [9] Convert RGB images to Napari label images
    [10] Crop out zebrafish larvae from 4x Acquifer images (multicolor but requires brightfield)

[2] Image Segmentation
    [1] Segment bright spots (2D or 3D, time series) 
    [2] Segment blobs (2D or 3D, time series)
        [1] User-defined or automatic (Otsu) thresholding
        [2] Cellpose's (generalist) cyto3 model 
    [4] Semantic segmentation (2D; fluorescence or brightfield)
    [5] Semantic segmentation (2D; Segment Anything)
    [6] Semantic segmentation (3D; requires dark background and good SNR)
    [7] Improve instance segmentation using CLAHE

[3] Regions of Interest (ROI) Analysis
    [1] Heart slices: Add 100um boundary zone to [intact+injured] ventricle masks
    [2] Count spots within ROI (2D)
    [3] Count blobs within ROI (3D)
    [4] Count colocalization of ROI in 2 or 3 color channels
    [5] Get properties of objects within ROI (two channels)
    [6] Get basic ROI properties (single channel)

[4] Image Segmentation Validation
    [1] Validate spot counts (2D)
    [2] Validate blobs (2D or 3D; global F1 score)

[5] Postprocessing
    [1] Compress files using zstd
    [2] Decompress files using zstd 

[n] Start Napari (with useful plugins)

WIP

• AI for ROI detection in brightfield and fluorescence images
• Code stability

Installation (Ubuntu)

A prerequisite is the Conda package and environment management system. The minimal Conda installer miniforge is preferable for its simplicity and speed. After installing miniforge, you can use mamba in the Linux terminal. Now you need to download the T-MIDAS repository either using

git clone https://github.com/MercaderLabAnatomy/T-MIDAS.git

or by downloading and unpacking the ZIP. In your terminal, change directory to the T-MIDAS folder and type

python ./scripts/install_dependencies.py

This will create the T-MIDAS environment and install all its dependencies.

Usage

To start the text-based user interface in your terminal, change directory to the T-MIDAS folder and type

python ./scripts/user_welcome.py

Acknowledgements and Citations

This project relies on several open-source libraries and tools. We would like to acknowledge and thank the creators of these projects:

Core Libraries

• NumPy: Harris, C.R., Millman, K.J., van der Walt, S.J. et al. Array programming with NumPy. Nature 585, 357–362 (2020). DOI: 10.1038/s41586-020-2649-2

• CuPy: Nishino, R. O. Y. U. D., & Loomis, S. H. C. (2017). Cupy: A numpy-compatible library for nvidia gpu calculations. 31st confernce on neural information processing systems, 151(7).

• scikit-image: van der Walt, S., Schönberger, J. L., Nunez-Iglesias, J., Boulogne, F., Warner, J. D., Yager, N., ... & Yu, T. (2014). scikit-image: image processing in Python. PeerJ, 2, e453.

• tifffile: Gohlke, C. (2021). tifffile (Version X.X.X) [Software]. Available from https://pypi.org/project/tifffile/

• pyclesperanto-prototype: Haase, R., Royer, L. A., Steinbach, P., Schmidt, D., Dibrov, A., Schmidt, U., ... & Myers, E. W. (2020). CLIJ: GPU-accelerated image processing for everyone. Nature methods, 17(1), 5-6.

• Pillow: Clark, A. (2015). Pillow (PIL Fork) Documentation. readthedocs.

• pandas: McKinney, W., & others. (2010). Data structures for statistical computing in python. In Proceedings of the 9th Python in Science Conference (Vol. 445, pp. 51–56).

• OpenCV: Bradski, G. (2000). The OpenCV Library. Dr. Dobb's Journal of Software Tools.

• SimpleITK: Lowekamp, B. C., Chen, D. T., Ibáñez, L., & Blezek, D. (2013). The design of SimpleITK. Frontiers in neuroinformatics, 7, 45.

• PyTorch: Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., ... & Chintala, S. (2019). PyTorch: An imperative style, high-performance deep learning library. In Advances in neural information processing systems (pp. 8026-8037).

Image Analysis and Processing

• Cellpose: Stringer, C., Wang, T., Michaelos, M., & Pachitariu, M. (2021). Cellpose: a generalist algorithm for cellular segmentation. Nature Methods, 18(1), 100-106.

• napari: Sofroniew, N., Lambert, T., Evans, K., Nunez-Iglesias, J., Yamauchi, K., Solak, A. C., ... & Eliceiri, K. (2022). napari: a multi-dimensional image viewer for python. Zenodo.

• apoc: Haase, R., Royer, L. A., Steinbach, P., Schmidt, D., Dibrov, A., Schmidt, U., ... & Myers, E. W. (2020). CLIJ: GPU-accelerated image processing for everyone. Nature methods, 17(1), 5-6.

• aicsimageio: Mancini, M., Colon-Hernandez, P., & Amodaj, N. (2020). AICSImageIO: A Python library for reading and writing image data. Journal of Open Source Software, 5(55), 2584.

• readlif: Nimne. (2021). readlif: Python package to read Leica LIF files. GitHub repository.

• OpenSlide: Goode, A., Gilbert, B., Harkes, J., Jukic, D., & Satyanarayanan, M. (2013). OpenSlide: A vendor-neutral software foundation for digital pathology. Journal of pathology informatics, 4.

• cucim: RAPIDS Team. (2021). cuCIM: GPU accelerated image processing. GitHub repository.

• aicspylibczi: Allen Institute for Cell Science. (2021). aicspylibczi: Python bindings for libCZI. GitHub repository.

• python-bioformats: Carpenter, A. E., Jones, T. R., Lamprecht, M. R., Clarke, C., Kang, I. H., Friman, O., ... & Sabatini, D. M. (2006). CellProfiler: image analysis software for identifying and quantifying cell phenotypes. Genome biology, 7(10), R100.

Deep Learning Models

• torchvision: Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., ... & Chintala, S. (2019). PyTorch: An imperative style, high-performance deep learning library. In Advances in neural information processing systems (pp. 8026-8037).

• timm: Wightman, R. (2019). PyTorch Image Models. GitHub repository.

• MobileSAM: Zhang, C., Han, D., Qiao, Y., Kim, J. U., Bae, S. H., Lee, S., & Hong, C. S. (2023). Faster Segment Anything: Towards Lightweight SAM for Mobile Applications. arXiv preprint arXiv:2306.14289.

Additional Tools

• tqdm: da Costa-Luis, C., Larroque, S. K., Altendorf, K., Mary, H., richardsheridan, Korobov, M., ... & Trofimov, A. (2022). tqdm: A Fast, Extensible Progress Bar for Python and CLI. Zenodo.

• pytest: Krekel, H., Oliveira, B., Pfannschmidt, R., Bruynooghe, F., Laugher, B., & Bruhin, F. (2004). pytest: helps you write better programs. The pytest Development Team.

We are grateful to the developers and maintainers of these projects for their valuable contributions to the open-source community.