[log in to unmask] wrote:
> On Wed, 8 Oct 2008, John Oreopoulos wrote:
>
>> My apologies again if this discussion thread becomes heated. I
>> thought I'd pass on another news bit about image manipulation, this
>> time revolving around brightness and contrast:
>>
>> http://www.ctv.ca/servlet/ArticleNews/story/CTVNews/20081008/stemcell_study_081008/20081008?hub=Health
>>
>>
>> Is this not something the journal should specify and the reviewers
>> should be looking for in the first place? It seems to me again that
>> all of this could be avoided if the authors simply state and describe
>> all image manipulations when submitting for publication
>>
>> John Oreopoulos
>>
>
> From the story you reference, the researcher did some minimal image
> processing that in no way altered the conclusions of the paper, and
> was hit on a religious objection that has no practical basis.  Doesn't
> sound to clever to me.
>
> The position these purists are taking is simply silly.  Were the same
> criteria in place before digital imaging, it would be "unethical" to
> produce prints from negatives -- or for that matter to even *develop*
> negatives at all -- since  all development and all printing
> *necessarily* involve "image processing."  When was the last time you
> created a print without affecting contrast and brightness?  Never? 
> Hmmm....
by this logic, the only "ethical" way to include images would be as a
table of sensor readouts in the supplementary material. I wonder who
would be happy about that.

my stand point here is firm, any image manipulation is allowed. it is
better by *default* to assume image processing. but the original image
should always be made available in that case. the method description
should come in form of a script to redo the operation using an open
source package. I think this is how biologists should work with their
data anyway because it allows them to redo the operation very easily on
other images. as an additional advantage, checking correctedness can be
almost automatic, the journal simply reruns the script. the only thing
left to argue about is the choice of operations, left to the reviewers.

currently we have too many black boxes; deconvolution operations is one
group of very important but hard to describe algorithms (the number of
biologists here who has implemented it, raise your hand). I would not in
any way be satisfied with a method description "was deconvolved with
XXX" because I most likely do not have the package. you cannot expect a
reviewer to suddenly shell out 10k usd just to verify a picture.

/Johan

-- 
--
------------------------------------------------
Johan Henriksson
MSc Engineering
PhD student, Karolinska Institutet
http://mahogny.areta.org http://www.endrov.net