Comparison of Salient Point Detection Methods for 3D Medical Images

Please use this identifier to cite or link to this publication: http://hdl.handle.net/1926/49
Salient points are used for various applications, such as medical image registration, tracking, stereoscopic matching. The purpose of this paper is to compare two commonly used methods to extract salient points in 3D medical images. We give an interpretation of the methods and validate their performance empirically based on criteria derived for the task of image registration, displacement measurement and tracking in medical images.
Data
minus 1 File (222Kb)
Code
There is no code review at this time.

Reviews
minus Nice validation of salient point detectors and contribution to open source software by Tom Fletcher on 09-17-2005 for revision #1
starstarstarstarstar expertise: 3 sensitivity: 5
yellow
Summary:
This paper reviews and compares two methods for computing salient points in 3D medical images. A set of validation metrics are devised and used to compare the performance of the two methods on 13 MR images of the brain.

Hypothesis:
Not applicable, as this paper is a comparison of existing methods.

Evidence:
The results of the comparison of the two corner detection methods are given on 13 MR images of the brain.

Open Science:
The source code is available as an extension of ITK in the NA-MIC SandBox. As far as I can tell, the MR images used in the paper are not made available.

Reproducibility:
I was able to download, compile and run the author's source code with relative ease. The test programs ran without errors, and I ran the corner detectors on a provided synthetic binary image of a 3D box. It would be nice if the MR images from the paper were also provided, which would make the results of the paper fully reproducable.

Use of Open Source Software:
The software to compute the salient points is provided as an open source addition to ITK in the NA-MIC SandBox. There is little documentation on the usage of the programs, but they are simple command-line programs that are easy to figure out how to use. It might be useful to add in the top-level readme file a description of the two command line tools.

The validation metrics and the framework used to compile the statistics comparing the methods is not provided (at least I didn't find them). This would be nice to have, as others might be interested in comparing yet other salient point detectors.

Open Source Contributions:
The salient point detectors are provided, and they are a very nice addition to ITK. Using the code in new software should be rather straightforward as the methods are implemented as ITK image filters.

Code Quality:
The code looked clean and easy to read. There was sufficient comments I think to understand what the code does, especially when combined with the detailed description of the methods in the paper. One nit-picky suggestion is to make the top-level CMakeLists.txt file also compile the two example command-line programs.

Applicability to other problems:
This code will be very helpful to researchers that want to find salient points in images, most likely for image registration.

Suggestions for future work:

Requests for additional information from authors:
Again, it would be nice to have the MR images used and the framework for the validation.

Additional Comments:
I would like to see just a little discussion/analysis of why the comparisons turned out the way they did. For instance, why is it that the correlation measurement showed better reproducability than the curvature measurement? My feeling is that this is because the correlation measurement is an expectation value in a small neighborhood, which might make it more robust to scale changes and registrations. Also, why is the curvature method more correlated to the entropy?

One point of the paper that I am confused with is whether the images used in the results are of the same subject or different subjects. You mention in the description of repeatability that salient points should be robust to images "scanned at different times", which makes it sound like the same subject scanned multiple times. And since the registration used is rigid, it appears that it is a single subject. Rigid registration of multiple subjects would not do a good job of aligning salient points, and thus it would be a poor measurement of repeatability.
minus Good study on salient points by Julien Dauguet on 08-19-2005 for revision #1
starstarstarstarstar expertise: 3 sensitivity: 5
yellow
Summary:
[Short description of the paper. In two or three phrases describe the problem that was addressed by the authors and the approach they took to solve it.]
Salient points are used in image analysis to extract informative and reliable regions. This paper describes two different methods to extract salient points and compares them on thirteen MR images.

Hypothesis:
[If Applicable: Describe the assumptions that the authors have made and they hypothesis of their work, note that not all papers will fit the model of hypothesis driven work, for example, the description of an image database, or the description of a toolkit will not be driven by an hypothesis, in which case, please simply write : “Non Applicable” in this field or delete the subtitle.]
The authors chose two methods based on derivatives of intensity to extract salient points: one using the Gaussian curvature and the other using the correlation matrix..

Evidence:
[Describe the evidence that the authors provide in order to support their claims in the paper. This is a key component on Open Science, opinions that are not supported by evidence should be labeled as “speculations” or “author’s opinion” while. The same rule applies to the text of the reviews: claims should be supported by evidence]
These methods are more adapted to low signal images than those using edge detection process.


Open Science:
[Describe how much the paper and its addendums adhere to the concept of Open Science. Do the authors provide the source code of the programs used in their experiments? Do the authors provide the input images that they used? Or are those images publicly available? Do the authors provide the output images that they show in the paper? Do the authors provide enough details for you to be able to replicate their work?
The methods proposed in this paper were implemented as open source using and extending Insight ToolKit.


Reproducibility:
[Did you reproduce the authors’ work? No
Did you download their code? Did you compile it? Did you run it? No
Did you managed to get the same results that they reported?
Were there information missing from the paper, that was necessary for you to reproduce the work? Suggest improvements that will make easier for future readers to reproduce this work.]

Use of Open Source Software:
[Did the authors use Open Source software in their work? Do they describe their experience with it, advantages and disadvantages? Do they provide advice for future users of those Open Source packages?]
They used open source ITK.

Open Source Contributions:
[Do the author’s provide their source code? Is it in a form that is usable? Do they describe clearly how to use of the code? How long did it take you to use that code?]
Yes, they provide their source code on the Web.

Code Quality:
[If the authors provided their source code: Was the code easy to read? Did they use a modern coding style? Did they rely on non-portable mechanism? Was it suitable for multiple-platforms?]
Not read.

Applicability to other problems:
[Do you find that the authors methods can be applied to other image analysis problems? Suggest other disciplines or even other specific projects that could take advantage of this work]
Yes, as said in their introduction, salient points can be used in all sort of image registrations but also in tracking, stereoscopic matching.

Suggestions for future work :
[Suggest to authors future directions for improving their methods, or other domains from which they could learn technique that could help them advance in their research.]
Comparison with other types of salient points extraction. Tests of such methods for image registration.

Requests for additional information from authors :
[Did you find that information was missing from the paper? Maybe parameters for running the tests? Maybe some images were missing? Would you like to get more details on how the diagrams, or plots were generated?]
A diagram for statistics could have been interesting.

Additional Comments:
[This is a free-form field]
Good study on two methods for the extraction of salient points: it is a very interesting tool to obtain reliable displacement fields.
Add a new review
Quick Comments


Resources
backyellow
Download All

Statistics more
backyellow
Global rating: starstarstarstarstar
Review rating: starstarstarstarstar [review]
Code rating:
Paper Quality: plus minus

Information more
backyellow
Keywords: salient points, corner detection, landmarks
Export citation:

Share
backyellow
Share

Linked Publications more
backyellow
An implementation of the minimization of region-scalable fitting energy levelsets An implementation of the minimization of region-scalable fitting energy levelsets
by Tang H., Hameeteman R., Gelas A., Van Walsum T.
An Architecture Validation Toolset for Ensuring Patient Safety in an Open Source Software Toolkit... An Architecture Validation Toolset for Ensuring Patient Safety in an Open Source Software Toolkit...
by Gary K., Kokoori S., David B., Otoom M., Blake M.B., Cleary K.

View license
Loading license...

Send a message to the author
main_flat
Powered by Midas