Skip to content

CHANGELOG.md

Latest commit

 

History

History
983 lines (723 loc) · 38.6 KB

CHANGELOG.md

File metadata and controls

983 lines (723 loc) · 38.6 KB
Raw

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[Unreleased]

[0.32.3] - 2026-08-05

Changed

  • Update from dependabot: cryptography
  • Add frozen affine to dependencies

[0.32.2] - 2026-07-23

Changed

  • Update from dependabot: soupsieve, Pillow
  • Allow two-row affine transform matrix upload (add vstack)

Fixed

  • Scalebar proportion on canvas change when using non-default pixel ratio #181

[0.32.1] - 2026-06-24

Fixed

  • Hover-tip info for stitch now includes CosMX
  • Update from dependabot: pyarrow, flask-cors

Removed

  • diskcache and imctools dependencies

[0.32.0] - 2026-06-01

Added

  • Stitching for CosMX Anndata ROIs with set uns identifiers
  • Warning modal if user supplies stitch image dimensions >50000 pixels

Changed

  • Output quantification CSV as zip with timestamp notification
  • Update dependency uv
  • Increase default spatial radius size to 6

[0.31.1] - 2026-05-20

Changed

  • Update dependency idna

[0.31.1] - 2026-05-11

Fixed

  • Mask option change when apply mask is off still updates cache #172
  • Better check for annotations in geojson to tiff CLI converter
  • Update dep based on dependabot: urllib3

[0.31.0] - 2026-04-28

Changed

  • Use image blend and mask fill cache to cache layers for faster blending
  • More informative tooltips for hover template options for performance #169
  • Add .obs export for CosMX preprocessing example

Fixed

  • Mask boundary recompute on gating to match gate selection #171

[0.30.2] - 2026-04-24

Added

  • CosMX preprocessing markdown notebook in examples #168

Fixed

  • Proper parsing of lazy-load files with unique partial name overlap (i.e. fov_1, fov_11) #166

Changed

  • Update dependabot vulnerabilities: lxml

[0.30.1] - 2026-04-20

Fixed

  • DGE table will now re-compute on both ROI and overlay change
  • Make the ROI switch coordinates (for placement) blank on ROI switch

[0.30.0] - 2026-04-15

Added

  • Transfer stitched image to WSI view (low-res preview remains)
  • Updated Visium HD export with masks in examples
  • Modal for differential gene expression for Anndata ROIs with categorical overlay

Changed

  • Add better permission on one-directory standalone builds
  • Update dependabot vulnerabilities: requests, crytography, Flask-HTTPAuth, uv
  • Cache mask boundaries when first created for ROIs that are sufficiently large
  • Set dataset gallery down-sampling default to True

[0.29.0] - 2026-03-17

Added

  • Created, edit, delete stitched images + add in canvas by coordinate #159
  • Input to manually set WSI zoom level
  • Better channel name parsing from OME

Changed

  • Updated dependencies based on cryptography and pillow vulnerabilities
  • Export canvas tiff as zip if sufficiently large

Fixed

  • Change user blend selection in dataset gallery to state variable
  • Check for unique condition in mod menu reset
  • Index mismatch for 10x spatial mask rasterisation
  • Spatial object id matching to handle spatialdata/Anndata sub-setting

[0.28.0] - 2026-02-03

Added

  • Toggle for down-sampling ROI gallery thumbnails
  • Toggle to inverse WSI transformation matrix #157

Changed

  • Auto-detect updates to WSI tiles in osd view #155
  • Extend canvas to osd coord transfer to non-Anndata ROIs
  • Local persistence on WSI scaling factor

Fixed

  • Skip any corrupted files on ROI gallery generation

Removed

  • update_poetry bash script

[0.27.0] - 2026-01-07

Added

  • CLI script for headless additive blending of files using existing JSON output #150

Fixed

  • Unique zip directory for every invocation of mask download #153

[0.26.0] - 2025-12-04

Added

  • Current blend channels highlighted in panel table #144
  • Ability to upload and choose from multiple WSI transformation matrices #145
  • Expand WSI file extensions accepted: jpg, jpeg

Changed

  • zarr parsing now accepts parent directory of multiple zarr subdirectories
  • increase bottom margins in channel and dataset gallery
  • Change toggle label for filters #143
  • WSI dropdown now doesn't use filetype extension, just stem base name

Fixed

  • Modal popup warning on quantification format error
  • Better gallery proportions for all ROIs #146
  • Don't reset panel table on subsequent uploads #147
  • Handle old global filter label, migration to new label #149

[0.25.0] - 2025-11-11

Added

  • Ability to parse anndata.obs as cluster projection frame #140
  • Output segmentation masks for Visium HD
  • Expand accepted tiff file extensions
  • Toggle to enforce single marker lazy for every ROI

Changed

  • Default parse for non 10x spatialdata iterates all spatial tables
  • Slide identifier and index in dataset strings
  • Toggle canvas compression format and level for sufficiently large ROIs
  • Use CLI array store type for .h5ad imports
  • Default spatial marker radius of 4
  • Remove limits on timeout and file number for server side file cache system
  • Exclude the smallest micron bin size for Visium HD from spatialdata
  • Set default port of 8080
  • Subset UMAP points while retaining coordinate order #119

Fixed

  • Include current session mask uploads when parsing zarr
  • Set category size limit when importing cluster projections
  • Check for existing cluster annotations on all import methods
  • Trailing slashes in zarr store parsing

[0.24.0] - 2025-09-08

Now requires Python 3.10, 3.11, or 3.12

Added

  • Expand WSI file types to include .ndpi
  • Ability to parse spatialdata zarr store
  • Ability to export current canvas shapes as JSON w/ re-upload #132
  • Console statement for user for where rakaia is running #133
  • Parse channel metadata if tiff is from ImageJ or has a page name

Changed

  • fully convert to pyproject.toml for PEP compliance
  • auto-select if single ROI is uploaded #134
  • Modal alert for panel mismatch on non-consecutive uploads
  • Single channel parsing from mcd now uses readimc subset indices

Fixed

  • Better dropdown menu proportions for choose WSI
  • Consistent UI styling for tooling + radio selections #135
  • Empty parsing if metadata keyword in ROI identifier
  • Annotations now compatible with modified shapes (redrawn/drag)
  • Nomenclature for mask overlay projection #137

Removed

  • dash-draggable due to conflicts with newer dependencies
  • deprecated files in envs: Singularity + pipenv

[0.23.0] - 2025-07-15

Added

  • openseadragon with custom JS to view co-registered H & E: supports tiff, btf, and svs images with JS or local filepath import
  • Ability to transfer zoom coordinates for Visium (V1, V2, HD) & Xenium to OSD WSI viewer
  • Key event listener for down arrow to switch between canvas and WSI viewer
  • Error warning in modal on ROI change for dropdown and dimensions
  • Ability to upload mask cluster annotations for multiple ROIs by CSV
  • Input to specify minimum number of query objects per ROI to be included in dataset gallery #129
  • Download region statistics table

Changed

  • make waitress production mode the default
  • handle concurrent user sessions better with clientside session ids used for static Flask routes and serverside stores
  • Loading widget for channel gallery now converted to non-fullscreen
  • Use UMAP channel subset in plugins/models #124
  • UI enhancements to import bar, channel gallery, toolbar #126 #127 #128
  • Vectorized mask filling from cluster annotations to improve performance
  • Vectorized spatial marker/spot rendering for marker arrays for faster rendering w/ max position overlap
  • Compute mask derivatives (hover, boundaries) on fly instead of storing
  • smaller pixel slider steps for lower intensity channels
  • Extra checks + change parsing rules for steinbock output directories with .txt files
  • single marker lazy loader and region thumbnail now uses array store type from CLI
  • Update autoscale CLI to process TXT files

Fixed

  • now supports quant results from steinbock with duplicate ROI indices #122
  • Fix UMAP categorical sub-setting to allow de-selecting of first category #123
  • Quantification naming conventions from in-app quantification w/ multiple files
  • Channel gallery tiles: use original image if dimensions are sufficiently small
  • pixel histogram will collapse if marker array is not available (i.e. due to lazy loading)
  • improved logic for single marker lazy loading based on total ROI pixels
  • proper ROI query order from quantification across multi-ROI files
  • Allow channel label to remain white if already user set with autofill colours turned on
  • Import cluster assignments from db/config if other ROIs have existing clusters
  • Bad UMAP menu proportions on Firefox #130

[0.22.0] - 2025-01-31

Changed

  • Single marker lazy loading for very large ROIs (both large dimensions + many markers, auto-detected): includes spatial datasets and extended to mcd & tiff
  • Channel gallery warning when tiles aren't rendered (spatial or large ROI as above)
  • Update dependencies to address dependabot and safety scan vulnerabilities
  • prettify UMAPs #115
  • change setuptools requirements explicitly to requirements.txt
  • UMAP/heatmap channel selection to modal
  • Avoid UMAP patching when updating coordinates (to be compatible with UMAP gallery selection)

Fixed

  • Using prev or next ROI before a key listener has been initiated blocks the callback
  • Allow y-axis overflow for multiple gating selection dropdown
  • pymongo client uses certifi.where() for SSL certificate to be compatible with pyinstaller
  • None pixel intensity lower and upper bounds for channels on slider checked separately
  • Gallery autoscaling when using single channel view
  • Do not allow key event triggers from appending children to ROI gallery
  • specific order for filter/hist callbacks w/ autofill colours

Added

  • module docstrings
  • hot pixel filter (hpf) difference for region summary statistics as additional column per channel #116
  • auto-detect frozen application (support for pyinstaller)
  • Input to apply or reset gating annotation to all quantified ROIs
  • UMAP min distance input from 0 to 1
  • Standalone build docs converted to pyinstaller from pyapp
  • UMAP gallery modal for UMAP min distance plots from steinbock snakemake pipeline output
  • Import scaling JSON when reading pipeline directory (local path)
  • Spectrum filters on channel tiles (rainbow, blue gold)
  • Show gating objects within current zoom level

[0.21.0] - 2024-11-21

Added

  • Ability to transfer annotations between ROIs
  • Option to export channel tile and ROI query galleries as HTML while retaining resolution
  • Support for reading spatial transcriptomic datasets from Anndata (.h5ad)
  • Custom input for setting custom spatial spot/marker radius (radius parsed on default)
  • Ability to generate a visium spot mask in-app
  • Adaboost tree classifier for plugin/quant models
  • Toggle for turning on/off masks in ROI gallery (default is True)

Fixed

  • Import filenames as sets (do not import duplicate filenames)
  • Improved scaling tick markers for smaller dynamic ranges and sparse matrices
  • Explicit sorting (descending) on value counts for quantification distribution tables
  • Quantification only on markers in memory to be compatible with spatial
  • Tick mark bug when using custom range slider max and low values
  • Check new cluster subtypes by category on a new upload and update only if different subtypes
  • Fix reset cluster subtype reassignment because of numerical to string casting
  • Pixel histogram close on ROI change and not blend update

Changed

  • Increase limits on raw image uploader
  • No default database connection string
  • Favicon for browser tab #114

[0.20.0] - 2024-09-19

Added

  • plugin ecosystem for quantification results: leiden clustering, random forest (existing annotation, object mixing)
  • Keyboard Event listener for mask toggle (up arrow)
  • Ability to read all outputs from a steinbock directory using the local filepath

Changed

  • grouped clustergram for categorical UMAP projections instead of heatmap
  • set auto open tab to default in CLI
  • revamped landing page

Removed

  • Basic HTTP authentication for the landing page

[0.19.0] - 2024-08-29

Added

  • script to install using pyapp (UNIX systems)
  • Cluster projection distribution table with selected cluster sub-categories
  • Image prioritization using similarity scores computed from a cluster proportion

Changed

  • slight modification to RegionAnnotation pydantic implementation to be compatible with pyapp
  • glasbey palettes and better categorical identification for UMAP overlay variables

Fixed

  • update from db or JSON for consecutive panel changes
  • multi cluster color assignment import with multiple ROIs
  • string casting for cluster category distributions
  • UMAP figure patch now considers categorical overlays based on number of unique values
  • quantification to cluster transfer for old pipeline (imcPQ) syntax
  • proper selection of objects through UMAP categorical legend if initial zoom hasn't been used

[0.18.0] - 2024-08-09

Added

  • Ability to transfer quantification columns to cluster projection per ROI
  • Median and standard deviation to region summary #107
  • Ability to toggle side placement of canvas tools sidebar #108

Fixed

  • Check for signal retention for channel gallery thumbnails after Image.Resampling.LANCZOS down-sampling
  • Don't enable zoom on channel-wise gallery due to varying ROI dimensions
  • Add panel length check in addition to key match for JSON, db config updates
  • More robust checks for importing files when ROIs already exist in the current session

Changed

  • Recursively go through channel gallery children on label changes instead of re-rendering all components
  • Do not update channel gallery on zoom automatically: use button trigger to update
  • Canvas mode bar moved to center #108

[0.17.0] - 2024-07-23

Added

  • sphinx docstrings for classes
  • Dropdown to enable multiple cluster annotation import and toggle

Changed

  • Rename (rakaia) and code clean (pylint compatibility)
  • modify cache clearing conditions on new app initialization
  • Set gating parameter slider steps smaller by half (0.005)

Fixed

  • Proper mass deletion of annotations across multiple categories
  • Make ROI querying from quantification work across multiple MCDs

[0.16.0] - 2024-07-09

Added

  • Exception thrown when the file parser doesn't successfully parse any ROIs
  • Named blend configurations can be saved and toggle using a dropdown menu, to rapidly change the canvas among sets of blended channels
  • Mask name to ROI matching from steinbock pipeline naming conventions
  • conversion parser for intensity h5ad from steinbock to internal dataframe
  • Mask input options now saved in JSON and database #104
  • Animation to mantine color swatch buttons on click
  • Ability to gate a custom list of mask IDs provided as a string input
  • Set steinbock mask dtype os environment variable on app startup, related to this

Changed

  • Major UI revamp prior to public release
  • Improved shape region parsing for redrawn/edited shapes
  • Better generalizations for conditions for tiff and HTML download
  • Switch to glasbey for palette generation
  • Switch to steinbock scipy.ndimage implementation for quantification and region props
  • add local persistence to mask-related inputs
  • Switch to scipy.ngimage to compute mask object overlap on drawn annotations
  • Bump import max for dash-upload image components (up to 50gb)
  • Channel gallery thumbnails now use smaller add icon without text
  • Modify heatmap to show down-sample by default with reporting total object counts
  • Run UMAP on current channels in heatmap
  • refactor API module naming for wider release
  • Use partial functions as switch statements for file reading in RegionThumbnail, FileParser
  • dedicated category dropdown for click annotations
  • JSON panel update now requires all keys to match, not just length
  • Cap length of frequency distribution table in quant tab

Fixed

  • Prevent callback on gallery channel view when not enabled
  • Handle overflow for long ROI names in ROI data table using scroll

Removed

  • Undo last annotation for regions (annotations table has selectable deletion)
  • Preset hover (converted to static input + button)

[0.15.0] - 2024-05-17

Added

  • warning message for ROI query when images or quant results are missing. #101
  • CLI script for autoscaling ROIs from mcd
  • Min amd max dimension limits (optional) + keyword searching for ROIs in gallery query with improved recursive searching to match the query number
  • Percentage distribution for UMAP categories

Fixed

  • missing callback output for opening in-browser quantification modal
  • fully refresh channel selection on JSON/db update
  • Proper color updates to the blend hash when using autofill #102
  • Re-import exported custom metadata works as expected
  • proper str casting for cluster annotations as circles
  • Upper bounds of 0 from JSON style imports are no longer reset to the default percentile
  • Change default CSV annotation value to Unassigned to avoid empty parsing by pandas on re-upload

Changed

  • ROI selection menu moved to main canvas to facilitate more consistent use. #97
  • mask boundaries using skimage now use the innermost pixels as opposed to outer
  • Better legend and scalebar x-axis range placement for varying ROI dimensions
  • Specify Ubuntu version 22.04 in Docker
  • UMAP channel overlay now uses dash.Patch for faster re-render (only numeric to numeric)
  • Move PanelMismatchError import warning into a modal alert

[0.14.0] - 2024-04-18

Added

  • Tab for marker correlation: set a target and baseline marker with masking to get proportion of marker overlap between mask/image and target to baseline within mask objects + Pearson pixel correlation (uses current channel blend parameters for filter and threshold) (compatible with zoom)
  • Button to autofill single mask upload names with the current ROI identifier
  • Mask object counter in the quantification modal when a mask is enabled
  • Exception thrown when lazy loading doesn't work for the current ROI, possible due to the string delimiter
  • Generalize the mongoDB database access by exposing the connection string as an input: Allows custom configuration of mongoDB instances
  • Option to select specific cluster categories for mask projection
  • Checklist toggle to bulk annotate all current canvas shapes (without, defaults to most recent)

Fixed

  • Update canvas shape filtering on freeform draw to clear errors caused by plotly 4462
  • Fix switch trigger to disable ROI switching on keyboard
  • Mask objects in ROI thumbnails are now all represented by the same intensity (255)
  • Do not reset channels to quantify on ROI changes
  • Proper ROI querying with tiff files from UMAP
  • Proper blanking of the UMAP plot when the quantification results are updated in-browser
  • Freeform rectangle coordinates now parsed from either left or right dragging
  • Fixed bug where annotating from UMAP overwrites the annotation hash instead of the quantification sheet
  • PDF writer now includes all annotations in loop

Changed

  • Enforce ROI change can occur only on the canvas tab
  • Edit the uirevision on shape clearing to resolve plotly 2741: modify the uirevision variable while maintaining truthy value
  • Better visibility for channels set to white in the dash ag grid (label set to black)
  • minimize canvas compression level to speed up image generation
  • More lenient cluster sheet import (now only verifies required column names)
  • Wrap gallery tiles in loading screen
  • Do not read ROI image for adjusting canvas size/dimensions

[0.13.0] - 2024-03-12

Added

  • Custom string delimiter input to change the string representation of ROIs within the session
  • Data refresh button for refreshing ROI selection in case of a PreventUpdate exception in the canvas callback
  • Add poetry files for development option
  • Custom exception for data reading error at likely callback point where disk storage errors occur
  • Added integrated (total) signal to channel summary statistics
  • single or multichannel gating from quantification results: gating can be applied to a canvas mask and combined with clustering labels (mask and circle)
  • Add annotation from gated cell list to quantification results or CSV object list output
  • removing the most recent annotation will also apply to the quantification frame annotations: will set any annotated objects in quantification to the default None
  • Delete annotations by choice from the annotation preview table #89
  • Toggle switch for enabling ROI change using left and right arrow keys
  • Toggle button for select/deselect all for in-browser quantification
  • Toggle button for disabling in-session messages (Default is to enable for safety)
  • Button to re-import ROI annotations into the quantification frame when the quant frame is re-generated

Changed

  • Annotation dictionary entries internally represented by pydantic base model
  • Use only unique aliases in canvas legend: allows for channels representing the same measurement to be combined with the same colour and represented only once for simplicity
  • JSON and mongoDB documents now contain cluster colour annotations if they exist on export
  • JSON export includes aliases
  • MCD import now includes the acquisition ID in the ROI name in addition to the acquisition description to ensure unique ROI naming
  • modify UI theme
  • change debug command line option to production-mode: wraps debugging and production-level WSGI server into one command for concurrent deployments
  • image downloads will now include the ROI name in the file name output #92
  • Annotation preview table now contains the preview column which describes either the bounds of the annotation or the objects contained within in a short string

Fixed

  • Added additional checks for panel length mismatches when files are not imported at the same time
  • Add ability to retain cluster assignments for individual ROIs when toggling
  • Cluster annotations can now be added to the canvas legend via toggle: can replace the channel labels when clusters are applied
  • readimc updated to 0.7.0 to include reading imc files with strict=False for leniency on corrupted data
  • Add CLI option for session cache storage with a default generated by tempfile
  • fixed improper array index type for quantification in the browser
  • fix attribute errors on HTML canvas export that block the download
  • ROI selection deletion now reflected dataset preview table

[0.12.0] - 2024-01-26

Added

  • mongoDB database persistence: Users can use a registered a mongoDB account for the rakaia-db Atlas instance to and/remove current and saved session configurations (blend parameters, global filters, channel selection, naming/aliases) (Note: uses an 0 Sandbox version for initial testing)
  • Custom exception class for panel length mismatches on file parsing: verifies panel lengths both within single uploads and among all files in multi-upload
  • Optional integer input to down-sample the heatmap for proper rendering
  • Ability to export objects inside region annotations without having a matching quantification sheet: object ids (such as cells) can be exported in CSV format if the annotation has a corresponding mask

Changed

  • toggleable child class of dash_extensions Serverside: ability to add string keys to Serverside objects to allow overriding of previous disk caches to reduce storage requirements and eliminate duplicate data writing for the same dcc.Store objects. Default is enabled, can be turned off in CLI options. Enables concurrency in shared instances #85
  • JSON I/O supports the values for global filters
  • Apply global filter to ROI query thumbnails
  • Modal alerts now use pydantic basemodel
  • Session downloads are now removed from the collapsible and use individual dcc.send_file prompts instead of an html.A hyperlink
  • custom scalebar value now sets the length of the scalebar by the value input and combines with the pixel ratio
  • ROI thumbnail queries can now be generated from tiff and txt files
  • Download for the UMAP projection moved into the UMAP configuration modal
  • Annotation mask downloads (tiff in zip) will include the original mask name used to generate the annotations, if provided #88
  • Make the metadata import more generalizable based on columns present: look only for the label column and length match
  • Remove redundant cell outline computation on single mask import (boundary is computed by default)

Fixed

  • imported JSON configurations will now match the proper length of datasets imported from h5py
  • Annotating the UMAP from the UMAP plot will trigger the recent annotation to be applied to the visualization
  • Fix improper range slider tick markers and maximum for channels with a max below 1
  • Add global filter variables to annotation PDF writer
  • Fixed storage step of underlying image in CanvasImage to allow export to tiff with mask or grid mods
  • Fixed overwriting previous annotation shapes when annotation categories are changed: shape parsing for annotations will now consider only the most recent shape drawn
  • freeform drawn annotations will be filtered to include only indices within the boundary of the image #87
  • Fixed incorrectly formatted function for importing custom metadata table (channel labels and names)
  • Fixed proper setting of debugging mode for both Flask and dash server components. By default, debugging mode will be enabled for troubleshooting, with the possibility of switching the default in subsequent versions
  • Fixed improper shape clearing on canvas layout adjustments
  • Fixed improper cluster imports: casting to str and fixing skimage.measure regionprops ordering for circles

[0.11.0] - 2023-12-21

Changed

  • Added persistence variable logging to select variables (session + canvas appearance) that are cached in the browser for future sessions based on user preferences
  • Additive image blending for channels is now handled through numexpr instead of numpy for faster performance (applies to additive blending, recolouring, and intensity thresholds)
  • Modify pattern matching for mask names to ROI names in datasets and quantification sheets to include partial match of mask to ROI
  • Move the canvas tiff download out of the download dropdown to prevent using Serverside storage for every canvas image (saves tmp disk space): now uses dcc.send_file instead of an html.A hyperlink

Added

  • Toggle feature to cast the canvas legend as horizontal instead of vertical (horizontal = all channel names on one line)
  • Added ability to add custom colour swatches through CLI as a comma separated string #81
  • Added ability to upload cluster assignments in CSV format to annotate an applied mask by colour (auto-generated or default random) #82
  • Ability to re-import point annotations and populate shapes in canvas #83
  • First draft: ability to read in quantification sheets from anndata/h5ad #54
  • CLI toggle option to toggle the backend array store type from 32 byte float (default) to unsigned 16 byte integer, with a tradeoff of precision vs. memory and speed

Fixed

  • Fixed extraneous callback on the canvas redraw the channel order is triggered but the order has not changed
  • Fixed improper intensity histogram max range due to sub-setting

[0.10.0] - 2023-11-27

Changed

  • Internal API changed: new directory structure to accommodate future data classes for the internal data stores
  • quantification: the expression bar plot is now replaced with a heatmap that changes dynamically on UMAP region selection. Channels can be added/removed using a checklist
  • The gaussian filter is now changed to GaussianBlur from the cv2 package.
  • Loading screens for the data import and data switch are now toggle-able with the CLI option -dl to disable load screens when importing or changing to new data. Allows for smoother transitions between ROIs #73
  • Next and previous ROI buttons are grayed out when the next or previous ROIs are not available, respectively. #75
  • mask object outlines are now computed with skimage to drastically increase speed and avoid pixel iteration
  • Imported imaging filenames are now sorted using natsort by default (feature can be disabled in the session configuration modal)

Added

  • Added a sigma value input for the gaussian filter as above. Default value is 1 with steps of 0.1. Used in the new gaussian blur filter
  • ROI queries from the quantification tab will apply the mask and fill in the cells show in the UMAP with white
  • Ability to auto assign colours from the swatches #72
  • Added functionality for keyboard-directed switches between ROIs using the left and right arrow keys #79

Fixed

  • Fixed channel filter values being reset to a default when a channel does not have a filter. Now, the values will remain from the last channel to get a filter applied
  • Fix improper shapes clearing and malformed shapes that cause erroneous "half-shapes" to appear on the canvas
  • Check for very long ROI names or ones containing a space that may interfere with proper separation in the dropdown menu. #71
  • Fixed UMAP clearing when in-app quantification updates the heatmap: UMAP plot will be hidden until new UMAP projections are computed
  • Fixed improper mask projection over canvas subset when regions are exported to PDF
  • Fixed improper conversion of cell mask to boundaries when numpy floats are not used in the mask array

[0.9.0] - 2023-11-07

Changed

  • Click annotations now have a circle surrounding the pixel by default (radius of 4 pixels) that is editable with adjustable radius size
  • Channel image gallery is now clickable through pattern matching: channels can be added to the canvas through a gallery click
  • Reset the mask option to None on an ROI switch to avoid the assumption that the current mask should be applied to a different image
  • Cache the quantification measurements CSV and UMAP coordinates as Serverside objects to speed up interactive UMAP for large datasets
  • Annotations linked to quantification sheets are parsed to identify which column should be used to match the annotation to the ROI
  • Set the default drag mode to zoom on a canvas export to HTML (prevents spurious shapes from being drawn offline)
  • Cast all return values for the quantification frame as Serverside objects to speed up manual annotations for large data frames
  • Current zoom is retained on ROI change if the dimensions of the new ROI match exactly. Otherwise, a canvas refresh is made
  • Annotation categories can now be created in both the canvas side tab and the quantification option tab
  • Improved heuristics for matching the quantification sheet columns to the current ROI name using either description or sample. #69

Added

  • Dataset query: mcd files can now be queried across ROIs to get clickable thumbnails for different ROIs that can then be loaded into the canvas
  • ROI name is now visible in the session outside the dropdown menu #68
  • Button to reset current channel percentile scaling to the default of the 99th percentile. #67
  • Additional session configuration variables in top right corner: can now switch on scroll zoom on the canvas (default is not enabled)
  • Ability to overlay grid on canvas with white lines to assist in cell counting
  • Link between sub-setting cells on the UMAP and sending a query to the query tab to display relevant ROIs by descending cell count
  • Pattern matching on mask and ROI names to automatically find the proper mask on an ROI change
  • Users can now annotate quantification sheets from interactive UMAP subsetting in addition to annotations on the canvas. Annotation category options are the same as the options listed for canvas-derived annotations

Fixed

  • Edits to the pixel histogram and range slider for images with pixel ranges between 0 and 1
  • Added additional logic to parse for ROI names in quantification sheet #69
  • Edited callback triggers for the expression bar plot to update properly on a mode change when a subset is being used

[0.8.0] - 2023-10-12

Changed

  • Region statistics table now contains separate rows for each region drawn with a shape (rectangle or freeform). Previously, all metrics were averaged across drawn regions
  • Permit the canvas to be resized when zoom or drawn shapes are applied
  • Allow bulk upload of masks, using default base names as the labels when more than one mask is uploaded at once #59

Added

  • Ability to click a pixel on the canvas and annotate it with a cell type and export point annotations in CSV format
  • Ability to sort channels alphanumerically in the channel selection dropdown (default is original channel order) #61
  • Add port input as CLI option (default is 5000)
  • Ability to export annotations as masks (one tiff per annotation class/grouping)
  • Input for setting ratio of pixel to distance #62
  • Ability to invert annotations placement on x-axis
  • Ability to subset ROI based on freeform draw on export #65
  • Set range on pixel intensity slider to new maximum to make it easier to adjust pixels for channels with a very large intensity range #60
  • Buttons for previous and next ROI. #63
  • Output click point annotations in CSV format with cell ID (if a mask is applied) and ROI ID #64

Fixed

  • Fixed error in gallery zoom when the zoom box was outside of the subset dimensions
  • Fixed histogram tick marks for input channels with a max intensity below 3 (divisor for the tick marks)
  • Fixed shape removal on new canvas modification when layers are added
  • Fix missing tick marks on range slider when switching to channels with lower bounds or numbers not easily divisible by 3

[0.7.0] - 2023-09-18

Changed

  • Ensure scalebar value is always centered over the scalebar line
  • Changed nested hash tables for arrays and blend parameters to top level hash tables for better hierarchical representation #34
  • Remove any previous rakaia cache dirs in the tmpdir prior to execution
  • Channel modification menu is automatically updated with the latest channel in the blend when the layers are updated. If a layer is removed, then the last channel in the queue is transferred to the mod menu #57

Added

  • CLI input option for local file dialog with wxPython #38
  • Input to add multiple annotation quantification columns #54
  • Frequency table for current UMAP selection #54
  • First draft of output report for annotations #53 (static)
  • Add version in header banner, right justified
  • Individual toggle buttons for the scalebar and legend to be toggle visible individually #55
  • Toggle ability to read either local filepaths or read all files in a directory matching the appropriate image file extensions
  • Ability to export the panel blend parameters as JSON and re-import (drag and drop)
  • New warning on file import if multiple file type extensions are detected
  • Message after file import listing the file obtained

Fixed

  • Fixed improper canvas resizing on layer change
  • Fixed improper update of intensity slider from h5 due to improper string to int conversion
  • Fixed error in writing Nonetype to h5 datasets
  • Fixed proportion/aspect ratio of interactive canvas export to HTML

[0.6.0] - 2023-08-21

Changed

  • Auto sizing of the canvas now happens automatically on an ROI dimension switch instead of manually

Added

  • Ability to add the rakaia cell type annotation to a quantification data frame from a region annotation: supports zoom, rectangles, and open form (non-convex) shapes #48
  • Added slider to change the size of the annotations (scalebar value and legend text)
  • Add toggle for custom hovertext over a mask, adding the mask/cell ID to the hover #51

Fixed

  • Edits to callbacks on channel switching to avoid running callbacks when the channel blend parameters haven't changed (happens when switching between channels in the mod menu)
  • Fix the centering of the scalebar value over the scalebar and edit default scalebar size and ratio of scale value to legend size

Changed

  • More explicit warning on quantification upload if the dimensions are not compatible

[0.5.0] - 2023-08-10

Changed

  • Major updates to the UI, particularly for the image annotation/blending tab: multiple inputs converted to being held in bootstrap offcanvas tabs (one offcanvas for inputs and one for advanced canvas settings)
  • Edit the canvas div to center the canvas by default and on resizing
  • Make the pixel histogram collapsible and by default have it hidden to speed up channel switching
  • Add font awesome icons to various buttons and remove default styling

Added

  • Add default argparse argument to open the browser window automatically on CLI run
  • Add colour box swatches from dash mantine colour picker for preferred colours

Fixed

  • Make sure the minimum of the default upper bound for pixel scaling by percentile is at least 1 #49

[0.4.0] - 2023-08-01

Added

  • ability to import multiple masks as single channel tiffs, either with or without cell boundaries with range slider for mask opacity
  • ability to import a measurements CSV with defined columns for marker expression by segmented cell
  • UMAP plotting for the marker measurements as above by cell
  • toggle scalebar and legend in the canvas
  • add button to remove an ROI from the session #41
  • ability to custom sort the channels in a dash ag draggable grid #36

Fixed

  • ability to perform multiple rounds of dataset import from drag and drop #16
  • retain the currently selected channel in the mod menu when the ROI is changed #43

[0.3.0] - 2023-07-10

Added

  • first draft of zoomed in coordinate navigation #26
  • add ability to add the pixel intensity values to the hover template #19
  • percentile scaling #22 #45

Fixed

  • proper pixel blending for overlapping pixels
  • better pixel scaling (linear from lower to upper bound)
  • callback separation for speed improvements
  • remove dcc.Stores from dcc.Loading by default to improve app smoothness

Changed

  • convert the pixel intensity changer to a range slider instead of box on the histogram

Removed

  • local dash diskcache background manager (conflict with pickle on certain Windows machines)

Security

[0.2.0] - 2023-06-19

Added

  • Lazy loading for mcd files using dataset preview by ROI
  • Ability to read IMC datasets from multiple sources including single and multi-page TIFF, mcd, .txt file

Fixed

  • #18
  • #21
  • #23
  • #25
  • #30
  • #31
  • #32
  • #28
  • #20
  • #24
  • #37
  • #40

Changed

  • Output from each session will be the currently loaded ROIs depending on the type of file input
  • Modal warnings will appear for improper uploads on dataset and custom metadata

Removed

Security

[0.1.0] - 2023-05-29

Initial beta/development release (internal)