v2.1.1

Full Changelog: v2.1.0...v2.1.1

v2.1.0

What's Changed

• revert stain normalization function by @jlockhar in #19

Full Changelog: v.2.0.5...v2.1.0

v2.0.5

What's Changed

• Added support for Akoya .qptiff files
• Minor bugfixes

Thanks to Martin Alcaraz and Cluadio Scafoglio at UCLA for providing the example slide from the Akoya scanner.

v2.0.4

Fixed a small issue with spinning up parallel processing pools when analyzing multiple images in a row.

2.0.3

Bug fixes:

• Fixed issue when outputting stain normalized image caused by improper block sizes in the 'blockedimage' object.
• Fixed an issue that overestimated the number of cores that are available for parallel processing by removing 'startparpool' function and using default parameters.

v2.0.2

Change log

• Changed 'GLASS-AI Resources' directory to 'Resources'.
• Changed method for locating 'Resources' directory to use mfilename(), consistent with method used in GLASS-AI Annotation Editor app.
• Added link to GLASS-AI Annotation Editor GitHub repo.
• Added button to copy app to base workspace in non-deployed environments for debugging
• MATLAB working directory is now reset to invoking directory in non-deployed environment after app figure is closed.
• Added alert to show MATLAB paths (ctfroot, pwd, invoking directory) in a non-deployed environment.
• Removed real-time logging of file selection. Selected files are only recorded after 'Run' button is pushed.
• Removed real-time logging of output folder selection. Output forlder is only recorded after 'Run' button is pushed.
• Minor fixes/changes to run parameter logging to improve readability and parsing by parseGLASSAIlogfile.m

GLASS-AI app v2.0.0

This release is for version 2.0.0 of the standalone GLASS-AI app. No changes have been made to the underlying convolutional neural network in this release.

The main goal of this update was to improve the efficiency of running GLASS-AI. In comparison to previous versions, v2.0.0 can process the same images 2 - 3x faster without requiring significantly more computing resources. This improvement is largely due to broader usage of MATLAB's blockedImage data type and refactoring functions to take better advantage of parallelization on the CPU.

In addition, how input files are selected and output files are saved has been reworked to use more intuitive selection methods.

This release also includes a version compiled for computers using Apple Silicon chips. Mac users will need to ensure they download the correct installer for their CPU.

Release Notes

UI

• Changed input file selection to allow users to use system dialog rather than looking for matching file types in a directory chosen by the user. This should be more intuitive than the previous method, particularly when selecting multiple files.
• GLASS-AI now accepts .png and .jpeg files as inputs.
• GLASS-AI will now ask before overwriting existing output files (e.g., when rerunning the analysis on the same image). Users can approve each file individually or allow all files to be overwritten as needed. If declined, files will be saved with the suffix _new before the file extension.
• Added a Stop Analysis button that will interrupt analysis at the next possible step. Some steps can take several minutes to complete, but MATLAB does not provide a way to interrupt running commands
• Added a new splash screen and icon. The amateur design is on me, the atrocious resolution is MATLAB's fault.
• Added a version/update checker that will notify users when an updated version of the GLASS-AI app has been pushed to the main branch of this repo.

Analysis

• Rewrote GLASS-AI app to utilize blockedImages in more functions.
  • Input images are no longer broken into chunks before creation of patches used for semantic segmentation. Now patches are fetched directly from the blockedImage object for the input image.
  • As a consequence the Low memory mode option has been removed. This version of the GLASS-AI app may require more memory to run than previous versions using Low memory mode but the overall analysis time has is 2 - 3x faster.
• Reworked stain normalization to apply across the entire image before analysis rather than applying to individual image blocks (which are no longer used) by estimating stain vectors from a scaled down version of the entire image. This will improve consistency across the image and should avoid normalization issues that could occur when an image block was largely devoid of tissue.
• Reworked classification smoothing to apply across the entire image rather than applying to individual image blocks. This overcomes the limitation of previous version that were not able to smooth between image blocks, only image patches. Smoothing is now performed on a scaled-down version of the pixel classifications.
  • This may slightly alter results from the previous method, but this approach provides a tremendous increase in speed (particularly when using mode smoothing which would often take longer the semantic segmentation).
• Changed some of the default analysis settings
  • Classification smoothing window reduced from 200 to 160 pixels.
  • Minimum tumor area increased from 20 to 500 sq. micron.
  • Making a tumor segmentation map is disabled by default.
  • Saving the stain normalized image is disabled by default.

Other

• All of the files for the standalone application have been moved to the GLASS-AI app folder of the repo.
• Additional files used in the updated UI and support files for running the application are included in the GLASS-AI resources folder.
  • If you wish to use the GLASS_AI_APP.m or GLASS_AI_APP.mlapp files to run GLASS-AI on your own installation of MATLAB, this folder and its contents are expected to be in the same folder as the GLASS_AI_APP.m or GLASS_AI_APP.mlapp files.
• This version of the GLASS-AI app was written & compiled using MATLAB 2024a. If you have issues with running the GLASS_AI_APP.m or GLASS_AI_APP.mlapp files you may need to update your version of MATLAB (and the required support packages).

Future plans

• Further memory optimization
• Switch more functions to operate on blockedImages
• Make a secondary app for handling previewing output images in blockedImage format
• Allow users to specify output image format (e.g., .jpg or .png)

v1.1

What's Changed

• Add files via upload by @jlockhar in #1
• Delete versions/0.8r directory by @jlockhar in #2
• Delete versions/0.9r directory by @jlockhar in #3
• Update GLASS_AI_APP.m to v1.0 by @jlockhar in #4
• Merge dev into main by @jlockhar in #5

New Contributors

• @jlockhar made their first contribution in #1

Full Changelog: v0.8r...v1.1

GLASS-AI review release

Preliminary release for manuscript review