Utilize TSAI tiler to crop out FOVs

[alex-l-kong]

What is the purpose of this PR?

Closes #12.

How did you implement your changes

Using the .imzml file, and combined with the output from TSAI:

1. Assign each scanned coordinate on the slide to their respective core, as labeled by TSAI
2. Use the coordinates to crop out the FOVs

Remaining issues

Currently, generate_glycan_mask segments out all of the cores at once. The result of map_coordinates_to_core_name, however, can be used to segment out individual cores. We still need to update generate_glycan_mask to handle individual core segmentation.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[srivarra]

Just one small aesthetic thing, and a question about "Z" coordinates.

[coveralls]

Pull Request Test Coverage Report for Build 9131611699

Details

• 59 of 59 (100.0%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.9%) to 97.753%

---

Totals
Change from base Build 6777233943:	0.9%
Covered Lines:	213
Relevant Lines:	214

---

💛 - Coveralls

[srivarra]

Looks good to me!

[alex-l-kong]

@kxleow ran into a case where multiple slides need to be supported. That is currently not the case, as we assume the user has only one poslog file per run. This will need to be added into the pipeline.

[alex-l-kong]

Cropping now supports multiple poslog files.

[kxleow]

Alternative to using poslog and imzml for core cropping:

1. After generating binary mask, use imageJ to segment/split up cores into individual binary masks
2. Then use centroids in json to assign core name or positions to each mask
3. Output table - a column with mask ids and a column with corresponding core names
   Please create a new branch, perhaps a separate notebook for this function. Thanks!

Sub-issue: Find nearest neighbours if centroid not found

• Encountered a case where the centroid fell exactly on a point that was not acquired in a core region.
• Would be helpful to consider these cases by testing for a few neighbouring points if the centroid is not found in any cores
• to assist with troubleshooting, would also be useful to print which centroid the function failed to match

[kxleow]

Ran into a case with >4 TMAs combined and the order of TMAs did not follow a sequential order in the image (left-right or top-bottom). The order of TMAs were found in the warnings generated at the core cropping step.