bImPy is a multi-dimensional image viewer and annotation tool built with Python using a Qt interface. bImPy should be pronounced 'bimpee' but you can call it whatever you want.
We are currently using bImPy to visualize and edit vascular tracing that are semi-automatically generated by Vesselucida(MBF) and DeepVess.
bImPy will eventually replace our flagship spine tracking software Map Manager and PyMapManager.
The roadmap is that by providing highly interactive 3D image volume (and time-series) visualization integrated with precise editing of annotations, bImPy will be able to leverage the ever growing automatic tracing that uses artificial intelligence (AI). Although AI is great, it always comes with an unknown error rate. For many biological image analysis tasks, error is not acceptable and the analysis must approach the ground truth. With bImPy we hope to achieve this.
See https://cudmore.github.io/bImPy/
We are developing bImPy in the open but the project is still in an early alpha stage and there will be breaking changes with each release. Please follow along as we develop this exciting new software and feel free to try it out and contribute ideas and code.
Figure 1. Stack window with a vascular tracing. (a) Control image contrast with sliders and a pixel intensity histogram. (b) The current slice and interface to toggle different aspects of the tracing. (c) Image (grayscale) and annotation/tracing interface. This is a 3D .tif stack looking at one image slice with the tracing overlaid. The tracing includes vessels (cyan), branch points (red), a vessel selection specified by the user (yellow), and a single point selection with an automatically drawn line to measure diameter (megenta). (d) Each point in the tracing has an automatically generated perpendicular line (megenta) and the line intensity profile is used to automatically generated best fits for diameter (blue). (e) Feedback on mouse position and pixel intensity as well as the last annotation edit.
Figure 2. Stack window using the 3D Napari viewer. bImPy is also using the excellent 3D visualization tool Napari. Here is a Napari 3D view of the image volume (green) with tracing (cyan) and a user selection (yellow).
Figure 3. Stack window using interactive and sortable annotation lists. Because tracings usually get large with more than 10,000 individual points and many hundreds of vessel segments, it is often useful to have a list of annotation objects. In this screenshot, we have a list of branch points (1st list) and vessel segments (2nd list). Selecting an annotation from a list or in the image will highlight that annotation in all other views. Each annotation in the lists is analyzed and the list can be sorted by, for example, vessel length. Here, we have selected a longer vessel with length of 314.26 (Len 3D) in the list and it is selected in the image (yellow).
We assume you have the following installed on your computer:
git clone https://github.com/cudmore/bImPy.git
This will make a folder named bImPy, we will use that for the rest of the install.
Make sure you change into the bImPy directory
cd bImPy
If you are using macOS or Linux, use our pre-made install scripts.
./install
Assuming you have venv and pip
cd bImPy
# create a virtual environment in folder 'bImPy_env'
python -m venv bImPy_env
# activate the virtual environment (macOS)
source bImPy_env/bin/activate
# activate the virtual environment (Windows)
bImPy_env/Scripts/activate.bat
# install requirements
pip install -r requirements.txt
# install bImPy, The '-e' allows editing of the installed package code
pip install -e .
./run
# activate the virtual environment (macOS)
source bImPy_env/bin/activate
# activate the virtual environment (Windows)
bImPy_env/Scripts/activate.bat
# run
python bimpy/interface/bStackBrowser.py
Qt for Python is a very confusing ecosystem. We are following the lead of the Napari developers and using PySide2 and QtPy. We are not using PyQt5? Relevant entries in pip freeze should look like this:
napari==0.2.11
PySide2==5.13.2
QtPy==1.9.0
If using macOS Sierra (10.12), PyQt==5.13.0 needs to be used.
bImPy is currently designed to load .tif files only. We are working on a version that will load commercial microscope formats and have started implementing this for Olympus .oir files.
This makes heavy use of Python bio-formats which in itself is excellent and is used by Fiji. Regrettably, bio-formats uses Java and requires the Python package javabridge. If you run into problems installing or using bImPy, it is probably due to your Java installation.
Sometime you need to force a particular version of PyQt. Usually PyQt==5.13.0. Once you have done one of the above, downgrade PyQt as follows
# activate the virtual environment
source bImPy_env/bin/activate
# downgrade PyQt
pip install PyQt==5.13.0 --upgrade
# downgrade PyQtChart
pip install PyQtChart==5.13.0 --upgrade
Please email Robert Cudmore at UC Davis if you are interested. See https:robertcudmore.org.