Factual correction for text in skeleton analysis blogpost

[GenevieveBuckley]

There is a factual error in my old skeleton analysis blogpost.

The blogpost shows a violin plot of the euclidean-distance measurement from skan, and says this is the skeleton branch thickness.

We can see that there are more thick blood vessel branches in the healthy lung.

That is incorrect. I misunderstood, and the error wasn't caught by Juan's otherwise excellent review (Juan is the author of the skan library).

Instead, it is correct to call this "the straight line distance from one end of the branch to the other".

Juan says:

The first thing I’ll say is that euclidean distance is not the thickness — it is the straight line distance from one end of the branch to the other

See this comment and this comment in an image.sc forum discussion for the full context. That discussion is from last year, but I only became aware of it today.

[GenevieveBuckley]

I should add, it's not a fatal error.

The conclusion I'd drawn from that plot was "This is consistent with what we might expect, since the healthy lung is more well developed than the lung from the hernia model." which is still supported by the correct interpretation of the plot - that there is a longer end-to-end distance for most blood vessels in the healthy lung. This makes sense, as the lung with the hernia is squashed into a smaller volume than the healthy one.

[GenevieveBuckley]

It would be better if the violin plot show was one for branch-distance (the total length of the branch).

Alternatively, the violin plot can stay the same (showing euclidean-distance), and the text updated to reflect the fact this is really measuring the end-to-end distance between branch start and end points, as if you'd pulled a string taught between them. That makes slightly less sense to show, when you could otherwise just show the branch distance, but it doesn't require re-analysing the data to generate a new plot.

Related: jni/skan#197

[GenevieveBuckley]

Alternatively, the violin plot can stay the same (showing euclidean-distance), and the text updated to reflect the fact this is really measuring the end-to-end distance between branch start and end points, as if you'd pulled a string taught between them. That makes slightly less sense to show, when you could otherwise just show the branch distance, but it doesn't require re-analysing the data to generate a new plot.

Update, we're taking this approach. It's the easier way out. I had a quick look at my analysis repo here and it doesn't seem super quick to re-run and generate a new plot - I'd have to find the original data and make some changes to the notebook first.