DeepVess

DeepVess is a 3D CNN segmentation method with essential pre- and post-processing steps, to fully automate the vascular segmentation of 3D in-vivo MPM images of murine brain vasculature using TensorFlow.

Additionally, The topological loss directory has the code and model related to the Topological Encoding CNN paper.

How to use DeepVess

First, see Installing TensorFlow for instructions on how to install TensorFlow.

Second, run prepareImage in MATLAB. (See Help prepareImage)

>> prepareImage()

Third, run DeepVess in Terminal or Python. You can add the address of the output of prepareImage (e.g. ../image3D.h5) as the argument. Otherwise, code will ask you to input it later.

$ python DeepVess.py ../image3D.h5

Finally, run postProcess in MATLAB.

>> postProcess()

Note that prepareImage and postProcess accepts arguments to avoid input request. For more information look at their helps in MATLAB. The most important argument is saturated_prctile that depends on the micrscope.

>> help prepareImage
>> help postProcess

You can send me a sample image and I run DeepVess for you to see if this model works on your images.

Requirements

• Python 3 (It's compatible with Python 2 as well)
• TensorFlow 1.14+ (With older version you may use commit fee62a2)
• Matlab
  • Image Processing Toolbox (if using motion removal in prepareImage.m)
  • Bioinformatics Toolbox (if using Fix the path of centerlines to have straight centerlines in postProcess.m)

Publication

• Haft-Javaherian, M., Fang, L., Muse, V., Schaffer, C. B., Nishimura, N., & Sabuncu, M. R. (2019). Deep convolutional neural networks for segmenting 3D in vivo multiphoton images of vasculature in Alzheimer disease mouse models. PloS one, 14(3), e0213539. Open Access link.
• Haft-Javaherian, M., Villiger, M., Schaffer, C. B., Nishimura, N., Golland, P., & Bouma, B. E. (2020). A Topological Encoding Convolutional Neural Network for Segmentation of 3D Multiphoton Images of Brain Vasculature Using Persistent Homology. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (pp. 990-991). Open Access link.

  Contact

• Mohammad Haft-Javaherian mh973@cornell.edu, haft@csail.mit.edu

License

Apache License 2.0