Magellanmapper

Changes

GUI
- Atlas Editor uses command-line specified alpha values

CLI
- Option to load microscope and atlas profiles through YAML files
- Script with sample commands for various common tasks

Atlas refinement
- Option to turn off negative labeling used to distinguish hemispheres
- The origin will update after atlas cropping
- Cleaned up profile settings for label edge extension and mirroring

I/O
- Fixed image stack export to use the new sub-image settings
- Applied workaround for exporting image labels where the first labeled plane is empty

Dependency Additions

- PyYaml (already included with other dependencies)

Installation
- These release notes are now included in the repository

GUI
- Setting transparency in overlaid images is more configurable

CLI
- Multiprocessing supported cross-platform by adding `spawn` support for Windows, which is also available and the new default in Mac; currently implemented for blob detection and image downsampling
- Fixed shebang for most Python files

I/O
- File naming simplifications: removed series string (eg `00000`), fixed some extensions (eg NPY instead of NPZ for single value archives), clearer suffixes (eg `blobs` instead of `info_proc`)
- Distinguished sub-images (specify by `--subimg_offset` and `--subimg_size`) vs ROIs (`--offset` and `--size`), which fixes many issues with saving/loading image and blob subsets

Server pipelines
- Tile stitching script uses `JAVA_HOME` for Java by default
- Fixed attempting to upload to S3 when directory is not set

GUI
- Theming capabilities, including default and dark themes
- Workaround for error if 3D points contain invalid values
- Workaround to show labels when the first image displayed has no labels
- Detection sphere slider is larger by making the max value compact
- Fixed settings for overlaid multichannel images, allowing scalars in addition to sequences
- Larger radius to select each ROI Editor draggable circle
- Customize the number of serial 2D plot columns using the `--plot_labels layout` option
- ROIs without any blobs can be saved
- Truth blobs in verified plots are shown with their actual radius rather than as a fixed size and placed below detection blobs

CLI
- Added "".py"" extension to run script to allow cross-platofrm launch by double-click
- Fixed detecting and activating Conda and Venv environments in the run script
- Explicitly set truth database takes priority over the default path

Atlas registration
- No longer specifies left/right for each label since the laterality may change with mirroring
- Test profile without any registration iterations
- Alternate output paths can be given as directories in addition to full paths
- Fixed regression in generating a new atlas

Nuclei detection
- Simplified setup for varying values at a single index during a grid search of a hyperparameter as an array
- Total variation denoising weight is configurable

I/O
- Default zoom and magnification values are now 1 rather than -1, and total magnification is always shown as a non-negative value

Server pipelines
- Option to specify the plane orientation

Python stats and plots
- Shift ROC legends to the lower right to reduce the chance of obscuring data points

Changes

GUI
- Open multiple Atlas Editors at the same time, including synchronized annotation updates
- Fixed inability to open an ROI Editor when a non-editable window is open
- Fixed error when attempting to show 3D blob locations without any blobs
- Fixed error when showing some ROI Editor overview images

Nuclei detection
- Grid search hyperparameters groups have been reorganized into selectable profiles
- Fixed running grid searches/ROC curves with Pandas 1.0
- Unsharp filtering and erosion can be turned off during image preprocessing
- The `vmin` saturation settings is now configurable, similar to the `vmax` setting
- The lower threshold factor for max scaling is now configurable to reduce false detections in low signal areas
- Microscope profiles for minimal preprocessing and low resolution images

Additional Assets

Sample 3D microscopy nuclei-stained region at cell resolution and downsampled whole brain are included here as separate assets

Changes

Installation
- No longer requires Git, C compiler, or preinstalled Java
- Startup script attempts to activate an environment if necessary, allowing the script to work from a double-click
- Moved installation docs to a separate file and added table of specialized dependency requirements
- Fixed Venv environment setup script
- Fixed SimpleElastix Windows build script to use the Conda environment Python executable if available

GUI
- Performance enhancement when viewing labeled images, especially for images with many colors
- Shows path of image loaded during startup
- Allows loading the ROI Selector without any image
- Fixed VTK error window display on Windows

CLI
- Removed unnecessary pipelines script options
- Fixed shutdown error on Windows and explicitly shut down the JVM if necessary
Atlas refinement:
- Label fill function now interpolates more smoothly and no longer adjusts the originally edited planes

Nuclei detection
- Fixed error when no blobs are found during verification against truth sets

I/O
- Applies metadata after importing image to make it immediately available for use
- Fixed regression in loading registered images

Dependency Updates

Python Dependency Changes

- OpenJDK 8 is installed through the Conda pathway
- Uses prebuilt Javabridge

Server dependency Changes

- Git is no longer required since not accessing Javabridge from Git
- C compiler is no longer necessary as long as prebuilt Javabridge works
- Java is no longer needs to be preinstalled if following Conda pathway

Changes

GUI
- show the file path in the Atlas Editor window
- fixed loss of label visibility after setting the opacity to 0 and changing planes

Atlas refinement
- profile for Allen CCFv3
- use the resolution of the fixed image during registration if the initial registration fails to start

I/O
- use FFMpeg for export to MP4 files for smaller files that can be opened by QuickTime
- fixed TIFF import to match only files with TIFF extensions (eg `.tif` or `.tiff`)
- fixed loading images by SimpleITK to select a single channel

Python stats and plots
- option for image padding size