Smoothing Images

[Anonymous]

Recently I have noticed that a lot of MALDI ion images that have been presented at conferences etc have been smoothed to make them look prettier.

Personally, I originally come from a microscopy background (specifically scanning tunnelling microscopy or STM) and the smoothing of images was often done but was scientifically frowned upon.

At the end of the day, smoothing images is manipulating your data, but not in a meaningful or scientific manner. Be proud of your pixels!

Anyone want to disagree or have an alternative view?

I know this is not the most obvious subject to start a discussion on, but having just discovered this forum the one thing it seems to be missing is actual discussion about issues relating to MALDI ion imaging and so this is my (perhaps poor) attempt to kick-start something.

Cheers,

Dan

[Markus Stoeckli]

How do you define the term ‘smoothing’? For displaying our images, we often use the interpolate function, as show in the example below:

[img]http://www.maldi-msi.org/components/com_joomlaboard/uploaded/images/raw.jpg[/img]
raw pixels

[img]http://www.maldi-msi.org/components/com_joomlaboard/uploaded/images/interpolate.jpg[/img]
interpolated

Is this by your definition also a form of smoothing which one should not apply to images?

[i]In BioMap, one can switch between the two modes by right-clicking on the window, clicking on Properties and selecting the DisplayMethods in the Display tab[/i]

markus

Post edited by: net-climber, at: 2007/01/08 20:16

Attachments:

• interpolate.jpg

[Anonymous]

Hi,

Yes, that kind of interpolation is what I am talking about. I will certainly agree that it makes images ‘prettier’ but I don’t think it is a scientifically valid procedure, particularly if no details are given on how the image was processed.

In the extreme case, imagine you have a 3×3 pixel Maldi ion image. The central pixel produced no signal for a particular ion mass (and so is coloured black), yet all the surrounding pixels registered a strong signal (and so are coloured white). If I were to now smooth/interpolate this image the central pixel will be transformed into a shade of gray suggesting that some signal was obtained from this position.

Perhaps in 90% of cases the ion of interest really was present and for some reason wasn’t successfully detected and so smoothing makes the image appear a little more like it ‘should’. But what about the 10% of cases where the ion of interest really wasn’t present? It seems to me that this is exactly the sort of information that you are trying to determine with imaging and by smoothing images you are manipulating your data and potentially losing this information.

I appreciate my example is a little extreme, but this doesn’t make the point any less valid. Although I will happily be proved wrong. 😀

Dan

[BL]

Hi Dan,

Thanks for your comment. I will never use interpolate again. We are looking at ratios of myosin light chain and myosin light chain phosphate peaks in contracting smooth muscle. We seem to be losing information with the interpolate function.

I am completely open to borrowing from the experience from other fields. Do you have any insight on base line correction and stuff like that when comparing images? Barbara

[Author]

Posts