Comment by Olga Bellon: answer

>Gestational Age estimation and Data Mining
>Is this review helpful?
>
>posted by Gavin Baker (newbie) on 08-31-2006 (revision 1)
>expertise: 3 | sensitivity: 4.6875 | helpfulness: NA
>
>Summary:
>This paper discusses a system for estimating Gestational Age (GA) in
>neonates. The method, called FootScanAge, aims to overcome the
>limitations of existing approaches; specifically: precision,
>invasiveness and limited validation.
>
>The introduction describes the system as "an Open Source and Java
>based Decision Support System", however no source was incldued in the
>submission, or available to obtain or access electronically.

Ok, I´ve explained this point, but there is no way to be more especific.

>
>Hypothesis:
>
>The paper's hypothesis is that FootScanAge is more accurate than the
>existing approaches cited, is less invasive, and is more thoroughly
>validated (especially with respecet to premature infants).
>
>Evidence:
>
>The results presented do not support the claims above. The accuracy
>was not compared to the other approaches mentioned. The image

Did you suggest me to test what? What kind of comparison do you
Want?

>acquisition process was not discussed, so it is unclear how the
>invasiveness of this approach can be evaluated. While the results are

You need just one image, that´s all. If you see the other methods
are not easy to have a score.

>based on 280 individuals, it does not discuss how many are premature,
>and this does not seem to be incorporated into the evaluation.

We´ll check this, but I results are based on premature newborns.

>
>Open Science:
>
>Not applicable (see below).
>
>Reproducibility:
>
>It was not possible to reproduce the results, as the source code was
>not supplied, nor was a binary application or refernce to further
>electronically accessible information. It would not be possible to
>implement the algorithms for independent validation based on the level
>of detail provided in the paper.
>
>Use of Open Source Software:
>
>The authors used the Java Advanced Imaging package to perform the
>image processing component. JAI is not strictly speaking Open
>Source/Free Software; the source is available for research use and
>non-commercial use, with some restrictions (such as maintaining ABI
>compatability with a test suite). There is no discussion about how
>JAI was used or integrated into the architecture. Other FLOSS
>projects are alluded to but not described.

Ok, we´ve just focus on the method, but we can improve this
explanation about who to use JAI. But we can easely convert
the code to C. It is not a big deal...

>Open Source Contributions:
>
>Not applicable (no source provided).

This is not fare! :)

>Code Quality:
>
>Not applicable (no source provided).

As I explain we can handle this issue.

>Applicability to other problems:
>
>This approach seems tailored to the specified application domain, and
>it is unclear (especially given the lack of detail provided) if these
>techniques could be applied elsewhere.

You get be kidding me.

>Suggestions for future work:
>
>Providing more details (as discussed below) would greatly strengthen
>this paper.

...as all scientific papers as you already know.

>Requests for additional information from authors:
>
>What is the software architecure of the system? How does the imaging
>module interface with the data mining step? How does the design
>incorporate JAI?

We can open a vídeo-conference to discuss this issue.

>More detail on the features used in the data mining is needed. How
>are the features selected, and what are they? How are they extracted?
>How are they combined to determine the GA score?

This is the main kernel of the method. We´we been working hard to
describe in detail all the ideas, but for now we´ve just the code
and some comments. I appreciate to explain the system in detail for
you.

>The tests use 280 instances (presumably photos). How was the data

No, the images were obtained from flatbed scanners and through
many image processing tools we obtaind the plantar surface region
as you see in the paper.

>validated? Presumably the training set had the GA determined by

No, we use a image-mining tool to compare with golden data.

>independent means? How accurate is the baseline data? How many were
>used for training versus classification?

The trainning data are based on golden standard cases, composed by
280 instances.

>
>What parameters were used to run the algorithms mentioned (eg. J.48,
>REPTree, M5P, and LMT)?
>

We can improve this.

>What do the ROC plots look like? What does this tell us about the
>performance of the classifier? How were the different algorithms
>compared against each other?

Hmmm! We just compare the results and the pediatric staff confirm.

>How does the cleaning filter work to remove the outliners?

100%

>How does the accuracy of this approach compare with the four
>approaches mentioned in the introduction?

See the number of weeks.

>Section 4 states that the specialist uses the results of the Image
>Mining Tool to "infer the final score". Are the results presented in
>section 5 from the system or the specialist? How does this imapct the

The specialist could change the score and the error is computed.

>accuracy? How are the estimates from the traditional means and
>specialist's values weighted and compared to produce the final result?

We are trying to create na automatic system...

>What is the significance that the use of all attributes in the GA
>estimation produces worse results than using fewer attributes?

The number of relationships are quite big to compute, and to reduce
the computation time we apply a classification + regression process
to obtain good searching spaces.

>
>Is the error rate of 1.4-2.08 weeks clinically significant? How does
>it compare with the other approaches?

With this we can provide a better treatment, and the result could be
much better in terms of quality of life to the pacient and so on.

>
>Additional Comments:
>
>Some of the text is unclear in the phrasing, and would benefit from
>proof-reading by an experienced English speaker.

Ok, we can check this.

>Figures 1 and 2 do not add a great deal of value to the paper. It
>could be far more informative to show screenshots of the application
>running, showing the real incoming data and the results of processing,

You can see the website with other screenshots
http://www.inf.ufpr.br/imago

>as described in Section 3. It would also be useful to see the images
>innotated with the features extracted.

We are developing and interative help to improve this.

>Figures 3 and 4 are of low quality (appear to be bitmap images), and
>would be clearer as EPS or PDF diagrams. The actual content of the
>diagrams is quite generic, and does not serve to illustrate the
>architecture as described in section 4.

We´ll improve this.

>It is unclear what Table 2 is representing; the caption and table qheadings
>could clarify.

Ok.

>The application focus clearly has merit, and there is great potential
>and benefit for providing accurate and non-invasive techniques for
>neonatal pediatrics.

Thank you very much for your time and suggestions.

Best,

Luciano
IMAGO Research Group

Comment by Olga Bellon: answer

>a java tool for estimating gestational age
>Is this review helpful?
>
>posted by Alice Villeger (newbie) on 08-31-2006 (revision 1)
>expertise: 2 | sensitivity: 4.66667 | helpfulness: NA
>
>Summary:
>This paper presents a tool written in Java implementing the FootScanAge method >to estimate gestational age. According to the results presented by the author, >gestational age can be determined simply and efficiently by using this tool. >However, this cannot be verified, as the FootScanAge method itself is not >described and (more importantly, perhaps) no source code is provided.

As mentioned in this message we are trying to improve the documentation and
to expand the project here in the University Hospital. We´ve been create
new opportunities to develop research in the “RUTE” project. And we hope
the code will be avaliable soon.

>
>Hypothesis:
>The FootScanAge method apparently assumes that a single image of the foot >provides enough data to estimate gestational age, contrary to other methods >currently used by clinicians, which involve more invasive procedures.
>
>Evidence:
>A feasability study was performed on a test database of 280 instances. >Unfortunately, no information is provided on the origin of the data: for >instance, the geographic origin is not given - and yet the author suggests it >might be a factor to take into account for further validation.

The data were obtained with a new technique developed in the University
Hospital that was is not avalible, the paper was submitted.
But, the 280 instances were computed and the results were alreadry
shared, see the references and web-site. With you like to receive a
copy of our results, please feel free to contact us.

>The evaluation shows an average error of 1 to 2 weeks in gestational age >estimation. It is said to be an improvement over other existing methods, but no >objective data (i.e. the usual precision of these other methods) is given for >comparison to support this claim.

See the references and comparisons about the gestational age estimation methods.

>pen Science:
>he software is said to be open source, but no link is given to download it.
>

Same as above.

>Reproducibility:
>he results could not be reproduced, as neither the software nor the data are >vailable.

It is not totally true, the method is described in detail in the literature,
But with you have a similar date you can reproduce most of the experiments.

>
>Use of Open Source Software:
>The software is written in Java and uses the Java Advanced Imaging API. This >framework is freely available, but not exactly open source, it seems:
>http://www.javalobby.org/forums/thread.jspa?threadID=17363
>According to that source, it is licensed under the JRL for non commercial use, >and under the JDL for commercial use.

We are na University if you undestand.

>
>Open Source Contributions:
>No source code

as above...

>
>Code Quality:
>No source code
>
>Applicability to other problems:
>Little details being given on the method, it is somewhat difficult to judge. >The application seems pretty specific, but maybe other clinical applications >involving data mining and image analysis could be found.

I agree, do you have any suggestion to make my paper better?

>Suggestions for future work:
>Interesting work, really. However, because of the Insight Journal's reviewers' >guidelines (which insist much on code availability and reproducibility), I >cannot give this paper a higher score.

OK. If someone else need the code to judge the paper, please let me know
and I´ll try to have the oportunitie to discuss the work.

>Requests for additional information from authors:
>A link to the software would be useful.
>
>
>Additional Comments:
>English is not my native language, but it seems to me the grammatical structure >of some sentences is a bit weird. Maybe you should have a native speaker proof->read this paper.

It´ll be done.

Comment by Olga Bellon: answer
>Gestational Age tool, source missing
>Is this review helpful?
>
>posted by Martin Styner (newbie) on 08-26-2006 (revision 1)
>expertise: 3 | sensitivity: 4.66667 | helpfulness: 4

Dear all,

Thank you very much for your time and review. I´ll try to explain the questions as best as possible. See as follows;

>Summary:
>The authors present a java-based system for the estimation of the gestation age >(GA) via the "FootScanAge" method. Overall, the manuscript does not describe >the FootScanAge method and no source code is given. It could very well be that >this tool provides an efficient way to estimate the GA, but this cannot be >verified.

The FootScanAge method is describe in the Academic Radiology paper too, see the references. The source code is not avaliable yet because we have the include a
good documentation about the software, system, management, update and upgrades.
I hope to finish this process soon, in order to put the whole code available
for you. We ar working in the “RUTE” project in Brazil, and I think that there
is a opportunitie to create a colaborative work in this workshop.

>Evidence:
>The manuscript presents a limited evaluation showing an average error of 1-2
>weeks in gestation age estimation. The evaluation is insufficiently described.

Did you already read the references about the traditional gestational age
Determination methods? None of those have a kind of precision.

>Especially outlier detection and rejection should be described in more detail.

I agree, we can improve that.

>Also, no details about the origin of the data are provided.

The data are from the University Hospital, obtained using a
common flat-bed scanner. We are working with a new digital device
to obtain the images... we ar trying to develop a new device, called
“FootScanner” to allow our research projects to be funding by
other agencies. We´we nice results comparing with other methods,
and we are planning to expand this idea to baby identification.

>Open Science:
>The authors claim that this adds an open source software tool to the community, >but no explicit source code, binaries or weblinks are given.

The web link is located at: http://www.inf.ufpr.br/imago
and the code will be available soon as mentioned early.

>Reproducibility:
>As there is no source code, binary, or other way to test the tools, its results >could not be verified.

I agree, but the published results show the results. Since you have the
code, could you provide us the dataset?

>Use of Open Source Software:
>The software is built on the Java Advanced Imaging framework >http://java.sun.com/products/java-media/jai , which is freely available, but I >am not sure whether it is open source.

What do you suggest?

>Comments to manuscript organization/quality:
>- Organization is ok in general
>- Figures 3 and 4 are of bad quality

What do you suggest us? This images were taken using low cost
Device to allow many department to use and so on. I hope to
be able to improve the image resolution as soon as possible.

>- Tables 1 and 2 need a meaningfull caption, no description is provided with >the tables.

Ok, I´ll try to improve this.

>- English grammar should be revised throughout the document

Any suggestion?