Alpha 0.10.7

What's Changed

• Filter intensity threshold can be zero integer edge case has been fixed by @paulshuker in #360

Full Changelog: 0.10.6...0.10.7

Alpha 0.10.6

Patch update for extract. Too much file reading for ND2 files has been fixed (#253).

changelog: https://github.com/reillytilbury/coppafish/blob/0.10.6/changelog.txt

Alpha 0.10.5

• Optical flow results are now saved as compressed zarr files (#246).
• OMP is faster (#240).
• OMP has better error messages and is fastest on the GPU (#240).
• Pytorch is now a default coppafish dependency.
• Coppafish now supports emailing users at the end of pipeline runs or crashes (see the documentation) (#219).
• New diagnostic plot Viewer2D added (see the documentation for usage).
• Minor Viewer bug fixes

Full Changelog: https://github.com/reillytilbury/coppafish/blob/0.10.5/changelog.txt

Alpha 0.10.4

• Updated view_tile_layout viewer

Full Changelog: 0.10.3...0.10.4

Alpha 0.10.3

• OMP crash fix

Changelog: https://github.com/reillytilbury/coppafish/blob/0.10.3/changelog.txt

Alpha 0.10.2

• Fix for a major OMP spot shape computation bug.
• OMP default config settings have been improved slightly.
• Package idna has been updated for security reasons.

Changelog: https://github.com/reillytilbury/coppafish/blob/0.10.2/changelog.txt

Alpha 0.10.1

This release includes large changes to the registration, both in terms of diagnostics and algorithms. We also have added support for blank tile, round, channels.

The changes can be summarised as:
Register

• z-smoothing is now much less harsh than yx smoothing in optical flow
• optical flow deals much better with edge cases where no good shifts have been found
• A new method of computing bg scales, which is based on a robust Theil-Sen regression. This replaces the old method which was sensitive to outliers and looked at data points with much more outliers present.
• A completely new viewer, which is refactored and optimised for faster viewing containing all necessary steps.

Bad TRC

• New way of dealing with bad tiles, rounds and channels. This is seamlessly integrated into code with the indexing package created by @paulshuker.
• Bad t,r,c are still extracted, but the filter step saves an array of 0s as the filtered image. The find spots step skips this trc, the call spots step deals with these bad trcs by calculating gene efficiencies only on tiles with no bad rcs. The probability approach simply takes the probability on bad tiles by removign the bad rounds.

Alpha 0.10.0

This is a major release from the previous version. To run this version, you cannot start from a <0.10.0 notebook. It is highly recommended to re-extract and re-filter all images again, i.e. delete the tiles directory and start from an empty output directory. Also, keep in mind that the output directory requires 28.5GB per tile (if each image is shaped 2304x2304x50) and 42GB per tile for the tiles directory. Dante_45_tiles.txt is a template config file used to run a dataset through successfully (note that the file extension should be changed to .ini).

The most significant changes are summarised below by section.

General:

• Coppafish will warn users when their version is outdated (#204).
• Jax is no longer supported (#212).
• Documentation is better and includes an advanced usage guide for users (#160). For example, users are advised on how to move their output or tile directory safely (#209).
• Various new and removed configuration parameters, defaults are set appropriately where possible.
• The pipeline.log now logs error tracebacks for easier debugging.
• Config parameters anchor_channel and dapi_channel are set to 27 and 0 respectively by default in basic_info.
• Config parameter dye_names is set by default in basic_info.

Extract:

• Pixel value histograms created in _extract.pdf fixed.

Filter:

• Default filtering better resolves microscope images and various bug fixes (#196, #171, #175).
• Filter is much faster.
• Pixel value histograms created in _filter.pdf fixed.

Register:

• ICP is better distance thresholded along the z direction, giving better registration results (#205).
• ICP now learns a single channel transform as opposed to one for each round.
• New optical flow registration added (#220).

OMP:

• OMP is faster.
• OMP spot scoring is better (#107).
• GPU pytorch memory crash fix attempt (#183).
• New _omp.pdf diagnostic.

Full Changelog

Alpha 0.9.4

• Background subtraction is now using the correct alignment parameters so it is actually working again
• Background subtraction blurring was turned off by a bug but this is no longer the case
• Background subtraction is only done if the background pixel is positive, which gives much better results
• Sobel filtering was being applied by default on alpha. If there was no parameter for sobel in Merry, then this would at least partially explain why it was taking longer Izzie
• Fixed the diagnostic to turn off the background image in the main viewer by clicking i
• Fixed the viewer to view the spots in the gene efficiency viewer
• All viewers are back in the traditional black layout so that we can see the white text on them again
• An OMP optimisation which reduces the amount of read time by a factor of 2. This speeds up the OMP step significantly.

Full Changelog: 0.9.3...0.9.4

Alpha 0.9.3

• New _register.pdf diagnostic PDF (#95).
• Detect spots memory crash bug fix.
• Gene efficiency plot in the Viewer bug fix.

Full Changelog: 0.9.2...0.9.3