Changes

• Improved support for reading OME-TIFF whole slide image files (single series, 3-channel RGB with LZW or JPEG compression)

Release binaries and installers can be downloaded below by pressing "assets". Make sure you have the required dependencies installed on your system before using the release: see the installation instructions for Windows, Ubuntu Linux and macOS

Changes

• Updated to OpenSlide 4.0.0 on Windows and Mac.
• Fixed bug with catching error when opening images with openslide
• Fixed issue with using h5py with pyfast on windows. This was done by statically linking the hdf5 library.
• Fixed bug in NIFTIImporter and added support for 32 bit integer images #205.
• Added tissue segmentation parameters to create method of TissueMicroArrayExtractor #204.
• Added vbeam example.
• Added Region and RegionList to pyFAST.
• Added arm64 Mac github action.
• Fixed multi-threading issues with RenderToImage.
• Added progress bar to RunUntilFinished.

Release binaries and installers can be downloaded below by pressing "assets". Make sure you have the required dependencies installed on your system before using the release: see the installation instructions for Windows, Ubuntu Linux and macOS

Changes

• Fixed issue with using h5py with pyfast on windows. This was done by statically linking the hdf5 library.
• Fixed bug in NIFTIImporter and added support for 32 bit integer images #205.
• Added tissue segmentation parameters to create method of TissueMicroArrayExtractor #204.
• Added vbeam example.
• Added Region and RegionList to pyFAST.
• Added arm64 Mac github action.
• Fixed multi-threading issues with RenderToImage.
• Added progress bar to RunUntilFinished.

Release binaries and installers can be downloaded below by pressing "assets". Make sure you have the required dependencies installed on your system before using the release: see the installation instructions for Windows, Ubuntu Linux and macOS

Changes

• Fixed issue with FAST not working on Ubuntu 23.10.
• Added a new python example showing how to create a rotating 3D gif: https://fast.eriksmistad.no/create_rotating_3D_gif_8py-example.html
• Implemented IsotropicResampler.
• Added 3 different normalization methods in ImageToImageNetwork.
• Implemented a very basic scalebar which can be enabled on the View.
• Colormaps: Implemented support for transparent colormaps, divergent colormaps, and added heatmaps Inferno and Fire and fixed some bugs with ApplyColormap.
• Implemented support for specifying compression in image pyramids and using neural networks for image pyramid tile compression.
• Fixed issues with input and output nodes not consistent in ordering from different OS. Nodes are now sorted on name.
• Implemented support for running TissueSegmentation on a regular image.
• Enabled PatchGenerator to run on any magnification, even if the magnification doesn't exist in the image pyramid.
• Implemented ImagePyramidAccess::getPatchAsImageForMagnification which enables users to extract a patch at any magnification.
• Implemented support for reading TIFF image pyramids with another data type than uint8.
• Fixed error with multi-input neural networks and batches, and an issue with batch size > 1 with TensorRT.
• Fixed bug with MetaImageExporter on windows when windows slashes were used.
• Fixed pyFAST bug with Pipeline::getPipelineOutputData and Pipeline::getAllPipelineOutputData.
• Expose fast::String data object to pyFAST.
• Fixed translation transform missing in UltrasoundImageCropper.
• Added setDisabled and isDisabled to ImageToImageNetwork.
• Removed invalid ASCII character from ImageFill.cl causing compile error on certain Mac machines.
• Set a default spacing for VSI files corresponding to 40X magnification.
• Some other minor bug fixes, documentation updates etc.

Full Changelog: v4.7.1...v4.8.0

Release binaries and installers can be downloaded below by pressing "assets". Make sure you have the required dependencies installed on your system before using the release: see the installation instructions for Windows, Ubuntu Linux and macOS

This is a patch release with mostly just bug fixes.

Changes

• Update Clarius Cast API to version 10.3.0.
• Fixed incorrect default port for Clarius.
• Enabled Clarius for arm64 on Mac.
• Fixed error in ImageResizer causing pixel shift.
• Added the missing TransferFunction to pyFAST.
• Fixed issue with runPipeline and datahub and additional command line parameters.
• Fixed issue with pyFAST and VertexTensorToSegmentation::create.
• Fixed bugs in PatchGenerator for 3D, and in 2D which didn't pad images at edges.
• Added resizeBackToOriginalSize argument to create method of SegmentationNetwork.
• Fixed NonLcalMeans was missing setModified in its set methods.
• Added possibility of converting from dB to grayscale in EnvelopeAndLogCompressor and in UFFStreamer when not scan converting.
• Fixed issue with ownership/deleting of OpenGL textures retrieved from Image object.
• Fixed bug with std::to_string rounding very small pixel spacing numbers causing shifts in high res WSI segmentation.
• Fixed bug in OpenVINO with index and multi input/output models.
• Added more documentation (python tutorials).

Full Changelog: v4.7.0...v4.7.1

Release binaries and installers can be downloaded below by pressing "assets". Make sure you have the required dependencies installed on your system before using the release: see the installation instructions for Windows, Ubuntu Linux and macOS

Changelog

• Updated ONNXRuntime to 1.14.0.
• Updated OpenVINO to version 2022.3.0.
• Updated clarius cast API to version 10.1.
• Made FAST compatible with Apple Silicone processors (arm 64).
• Added support for reading NIFTI files, including compressed NIFTI files.
• Fixed bug in 2D version of MeshToSegmentation.
• Implemented VertexTensorToSegmentation process object
• Fixed bug when multiple users try to create image pyramids on the same machine.
• Added support for string messages in OpenIGTLinkStreamer.
• Added ColorToGrayscale and GrayscaleToColor process objects.
• Fixed edge case bug with overlap in patch generator at bottom left and top right of image
• Added redistributable visual C++ runtime files to windows package to simplify installation on windows machines.
• Added 3D support to IntensityClipping.
• Implemented zero mean unit variance intensity normalization process object.
• Implemented SlicerWindow, a special window for visualizing 3D data.
• Fixed last frame bug in UFFStreamer.
• Added grayscale options to ImageFileStreamer and ImageFileImporter. ImageImporter grayscale is now default false.
• Fixed runPipeline to be able to run on headless machines.
• Implemented basic support for specifying Window in text pipelines.
• Implemented better support for neural networks with dynamic input shapes.
• Implemented SliderWidget and ButtonWidget objects to make it even easier to create GUIs with FAST.
• ++ several other bug fixes.

Full Changelog: v4.6.0...v4.7.0

Release binaries and installers can be downloaded below by pressing "assets". Make sure you have the required dependencies installed on your system before using the release: see the installation instructions for Windows, Ubuntu Linux and macOS

Main changes in this release are the addition of the ONNX Runtime inference engine for Windows which has DirectML/DirectX GPU support, as well as the new possibility of making streamers in python with pyFAST including a new GUI widget, PlaybackWidget, which can be used to play/pause/stop/loop/scroll through a data stream.

Changelog

• Added the ONNX Runtime inference engine with DirectML/DirectX support for Windows.
• Updated Clarius Cast API to v10.0.0
• Implemented the PlaybackWidget, a simple GUI element for controlling playback of random access streamers such as ImageFileStreamer.
• Implemented InterleavePlayback, a streamer which enables interleaved playback with the PlaybackWidget.
• Implemented possibility of creating streamers running in their own thread in Python.
• Fixed bugs in ImagePyramidAccess for tiff backend when reading arbitrary patches.
• Zooming in 2D now zooms towards mouse pointer.
• Panning in 2D can now be done using the arrow keys.
• Improved border radius setting of SegmentationRenderer, enabling border to be drawn outside of the segmentation as well on the inside.
• Added ImageFlipper process object which simply flips images.
• Implemented support for reading OME-TIFF files written by QuPath using the TIFFImagePyramidImporter.
• Fixed ImagePyramidPatchImporter to support color images.
• ++ Many bug fixes

Full Changelog: v4.5.0...v4.6.0

Release binaries and installers can be downloaded below by pressing "assets". Make sure you have the required dependencies installed on your system before using the release: see the installation instructions for Windows, Ubuntu Linux and macOS

Main change in this relelase is the introduction of FAST DataHub.

Changelog

• Implemented DataHub concept in FAST which enables easy distribution of data, models and pipelines, see https://datahub.eriksmistad.no/ for more info.
• Modified runPipeline to support --datahub command line option.
• Upgraded TensorRT to version 8.4
• Upgraded Qt to version 5.15 and SSL support.
• Implemented possibility of thick lines in LineRenderer for 2D using geometry shader.
• Added option to handle segmentation models that crop the input (overlap) in the model (patches are cropped) in PatchStitcher.
• Fixed issue with auto-detecting image dimension ordering in TensorRT engine, and enabled users to set image dimension ordering manually using text attribute.
• Fixed incorrect propagation of values in high-res segmentations in single-channel image pyramids.
• Added option to set a list of channels to ignore in output tensor of SegmentationNetwork and TensorToSegmentation
• More python tests

Breaking change

• If you use FAST with PySide2 you have to change the import order as such:

  import platform
  if platform.system() != 'Windows':
      import PySide2.QtSvg # Must import this before fast due to conflicting symbols
  import fast # Must import fast before rest of pyside2
  from PySide2.QtWidgets import *
  # ...

Full Changelog: v4.4.0...v4.5.0

Release binaries and installers can be downloaded below by pressing "assets". Make sure you have the required dependencies installed on your system before using the release: see the installation instructions for Windows, Ubuntu Linux and macOS

Changes

• Upgraded Clarius Cast API to version 9.3.
• Catch exceptions in ComputationThread and emit a signal (criticalError) which can be caught by a GUI.
• Enabled catching exceptions and their error messages in the ComputationThread and also potentially in streamer threads.
• Added signal when processing is done in ComputationThread.
• Fixed several multi-threading issues.
• Several improvements and bug fixes needed for V1.0 of FastPathology.
• Implemented function ImagePyramid::getLevelForMagnification to get the image pyramid level for a given microscopy magnification.
• Created the new ApplyColormap PO.
• Created the new RenderToImage PO.
• Fixed bug with incorrect spacing calculated from TIFF image pyramids.
• Fixed bug in ImageResizer causing edge artifacts.
• Write and read tensor spacing information in HDF5 tensor export and import.
• Started writing some python tests for pyFAST.
• Setup CI testing of both C++ and Python code for Ubuntu and macOS.
• Added some UFF data to the test data set.
• Many bug fixes ++

Full Changelog: v4.3.0...v4.4.0

Release binaries and installers can be downloaded below by pressing "assets". Make sure you have the required dependencies installed on your system before using the release: see the installation instructions for Windows, Ubuntu Linux and macOS

Changes

• Upgraded TensorRT to version 8.2
• Upgraded Clarius Cast API to version 9.0
• Upgraded OpenIGTLink to version 3.1
• Upgraded OpenVINO to version 2021.4.2
• Major refactoring of rendering system. Due to synchronized rendering, the computation thread was not able run faster than the rendering thread. This has now been fixed, and these two threads can run at their own speed. Rendering synchronization is now done in much cheaper (faster) way.
• Added ImageToImageNetwork convienience class for neural networks that perform image-to-image transformations.
• Fixed bug in which PatchGenerator created patches with another size than the one specified at the image boundaries. This is now handled by padding the patches.
• Fixed several issues in PatchGenerator related to offsets and patch overlap.
• Added image multiplication weight in NonLocalMeans filter
• Made it possible to make a minimal build of FAST without Qt.
• ++ bug fixes

Full Changelog: v4.2.1...v4.3.0

Release binaries and installers can be downloaded below by pressing "assets". Make sure you have the required dependencies installed on your system before using the release: see the installation instructions for Windows, Ubuntu Linux and macOS