Where’s Wally?

Many ‘stripy nanoparticles’ articles have essential claims which are not supported by any experimental evidence.

One such example is a 2012 Nature Materials article which is based on 6 types of nanoparticles, 3 ‘stripy’ and 3 ‘not stripy’ but where STM images had been provided for only one of the six samples. In his data request, Philip Moriarty had asked for the 6 data sets underpinning this article.

Some data have been finally released yesterday, but, as far as I can see, the data related to the Nature Materials 2012 article only include 3 data sets, the ones which are supposed to be stripy. It is of course critical to be able to examine the other ones too as the article is based on the idea that a difference in conductivity stems from a difference in monolayer structure. 

Still it is worth looking at the raw data shared yesterday (I discuss here only the data related to this Nature Materials 2012 article, more posts will follow). First, here is the published image.

Reproduced from Fig 1b, of Nat Mater 2012, DOI:10.1038/NMAT3406; STM image of HT=EG2 NPs

Reproduced from Fig 1b, of Nat Mater 2012, DOI:10.1038/NMAT3406; STM image of HT=EG2 NPs

Can you spot those particles in the image below (unprocessed raw data, LUT chosen to be similar to the above)?

Unprocessed, image produced with Gwyddion, filename my_eunseon_eg2_110608.022 downloaded 17/05/2013 from http://sunmil.epfl.ch/

Unprocessed, image produced with Gwyddion, filename my_eunseon_eg2_110608.022 downloaded 17/05/2013 from http://sunmil.epfl.ch/

Did you find them?

Here is the same image, but with some contrast adjustment (but no filtering/processing).

2-redish LUTcontrast adjusted

Can you find them now? I can crop…

2-redish LUTcontrast adjustedCROP

Zoom in:

Same as above, but zoomed in (3 times, no interpolation)

Same as above, but zoomed in (3 times, no interpolation)

If I zoom in with a bilinear interpolation (thanks ImageJ),

it looks ‘better’…

zoom with bilinear interpolation

zoom with bilinear interpolation

Clearly these are the particles shown in the article. I think it is the right scan too but I may be wrong (there are two other scans of the same area in the folder shared yesterday but this one seems to be the closest). Here is a challenge: what image processing steps do you need to get from the raw data to the image presented in Nature Materials 2012?

Update 1 (20/05/2013): Fair points by ‘Commentator’ below. I insist though that I have found Wally the right set of nanoparticles. See below.

Image from published figure with mask from raw data.

Image from published figure with mask from raw data (white rectangle is where the zoomed-in particle hides the rest of image in the published figure)

Advertisements

5 comments

  1. I would note that such image manipulation is certainly not allowed in biology journals (though it happens too frequently, see the litany of posts on Retraction Watch). The Journal of Cell Biology takes a very dim view of such practice. It runs every image submitted through a series of software tools to determine whether such manipulation is likely to have been performed. There are also tools at ORI which, I believe, allow an assessment of image manipulation. In any event, one might enlist the help of the Journal of Cell Biology if this proves too challenging.

    Like

    1. In cell biology there are a number of well identified agreed unlawful image manipulations. It is more difficult to determine the threshold of the acceptable here.

      Like

  2. It’s a small point, but if it is in fact the correct set of NP, this is clearly not exactly the same image since the “slicing noise” (in contrast to the striped feedback noise contrast) is different between the two scans (most evident on the lower center particle). It may be from the same sequence, but it is not the same picture. The differences in contrast are probably mainly due to differences in flattening, color thresholding and interpolation / effective smoothing by zooming in. The problem with the data is not that they have been processed ‘overly’ as you seem to be suggesting, its rather, they are cropped very selectively, and, of course, that the interesting features themselves are an imaging artifact (!)

    Like

  3. I can sort of see stripes in the published image (and even nice curved arcs instead of straight up and down). That said, I sure can’t see them in the original data and the straight zoom. Very concerned about the image processing (putting pixels where there were none before).

    Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s