Summary:

The paper suggests an implementation of a vessel detection and enhancement function based on the article of Manniesing.

Hypothesis:

The work makes a hypothesis of detection of tubular structures in 3D images and improvement of their network and mean intensity.

Evidence:

The authors provide tests, test and baseline images. Their results are reproducible. Regression tests were not run.

Open Science:

The paper is devoted to the concept of Open Science. The authors provide a rather explicit source code, input, baseline images and enough information for the results to be reproduced.

Reproducibility:

The code was successfully compiled and run. And the results were reproduced.

Use of Open Source Software:

Authors use ITK, but they don't describe much their experience with it.

Open Source Contributions:

The authors provide the complete working code, which is rather clear. It took me some time to understand the flow of the algorithm and find the error mentioned in the earlier review by Luca Antiga on the single-scale diffusion enhanement.

Code Quality:

The code quality is acceptable. The authos suggest a use of 3D Diffusion Tensor, which complicates the extension of the function to ND images.

Applicability to other problems:

The program can be used to detection and enhancement of any linear, tubular objects.

Suggestions for future work:
A flowchart or a more explicit explanation of the functions can be suggested for an easier understanding.

Requests for additional information from authors:

In the Dart test file the suggested scales are not the same as in the article and the ones used for saving the processed images. Scales [0.4, 5.0] are used instead of [0.5, 4.0].

Additional Comments:

NA

In the paper, the authors provide an ITK implementation of the Vessel Enhancing Diffusion algorithm presented by Manniesing. The algorithm was implemented by defining four new ITK classes: a new vesselness function class, a multiscale filter class and new vesselness-driven anisotropic diffusion function and filter class. The paper is completed with source code, examples and testing datasets.

Download, compilation and execution of the code were successful. 

Both regression tests didn't succeed. Although they produced plausible results, the output images didn't match the baseline images submitted by the authors. (Mac OSX 10.4, ITK 3.2.0).

The reviewer extended some of the code presented in this paper, in particular the itkMultiScaleHessianSmoothed3DToVesselnessMeasureImageFilter class. in a paper submitted to the IJ (http://insight-journal.org/dspace/handle/1926/576), which also provide a ND generalization of Frangi's vesselness. The extended multiscale filter can be used with all Hessian-based vesselness formulations.

In addition, following similar arguments used in the cited paper, the work presented here could be generalized to N dimensions, by properly formulating the smooth term in the vesselness function. Ultimately, this could be also used for ND anisotropic diffusion.

It would be nice to integrate the present work with some of these generalizations.

The discrepancies between the baseline images and the ones produced by the tests should be clarified.

The bug reported above should be fixed.

Just a few typos: 

par 1.1, line -1 wrt Eq 1: plate-like structure -> plate-like structures.

page 3, line +1 wrt Eq7: Where -> where

page 3, paragraph 2.2, line 1: follow -> follows

page 6, first paragraph, last line: double comma

Summary:
The authors provide an open source implementation of vessel enhancing diffusion (VED) algorithm proposed by Manniesing et al. The implementation is validated on example images and VED algorithm is compared to Frangi vesselness filter. Authors explain their implementation and provide example test cases to reproduce the results described in their work.

Evidence:
The results presented by the authors are well in coherence with the results shown in the original work by Manniesing et al. Especially, shrinkage effects resulting from Frangi vesselness filters can be clearly seen in the figure 3. As expected, VED algorithm performs much better on the same input data.

Open Science:
The work is an excellent example of open science. References to the original work are given in the abstract and the algorithm can be easily compared to the original work. Authors provide their source code and the data they used in their work.

Reproducibility:
I downloaded the code and compiled it successfully on my 64-bit linux system. I run the tests provided by the authors using cmake. It was straightforward to reproduce their results. The images produced by VED algorithm were similar to the images in figure 4. However, I could not get the same images as in figure 3, although I observed similar effects. The slice output by the test might be different than the one in the paper.

Use of Open Source Software:
The implementation makes of ITK.

Open Source Contributions:
The source code is provided with the submission. It was very easy to use the code and reproduce the results in the paper. For a user familiar with ITK, it is easy to use the code.

Code Quality:
The code is well documented and easy to read. Overall the code was in excellent shape and was at the standard of the code available in ITK release packages.

Requests for additional information from authors:
It might have been better if the authors explained their experience with the parameters. A discussion of the parameters would be valuable for the readers. Nonetheless, the original work by Manniesing et al. can be easily checked for that purpose.

Additional Comments:
In their work, the authors provide an excellent implementation of VED algorithm and provide their source code and data. The comparison to vesselness filters are a plus. Overall, this work will be a nice contribution to ITK.

Summary:
In this work, authors describe an implementation of the VED algorithm under ITK. They explain how the algorithm works and how they have made the implementation. Experiments are presented and readers are encouraged to build and run the code.

Hypothesis:
They assume a smoothed version of the Frangi's vesselness, and the value of the algorithm's parameters.

Evidence:
NA

Open Science:
This work is adhered to the concept of Open Science. Authors provide the source code and the data used for the experiments. Also they provide the output data of their implementation, and they provide some details to replicate the results (it would be nice to specify the number of scales used for Frangi's vesselness filters).

Reproducibility:
I've compiled the code and reproduced the author's work. I played with parameters to get similar results to the authors. I think that it would be nice to specify the number of scales used for Frangi's vesselnes filters and the value of the parameters by default.

Use of Open Source Software:
They implement this work under ITK. They describe the experiments that they have done, but I think that it would be nice to explain a little bit the interpretation of the algorithm parameters (i.e epsilon, w and S). I would avoid to use "S" again for a parameter, because it has been used before to denote the square sum of eigenvalues (eq. 4).

Open Source Contributions:
The code is easy to use and compile, but it would be nice to put the ITK version at the CMakeList.txt file. I tried to compile it under ITK-2.4 and I had problems until I used ITK-3.2.

Code Quality:
The code follows the codification style used under ITK. It is easy to read and easy for ITK's users.
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.]
NA
Requests for additional information from authors:
In subsection 1.1 the cite to the reference 4 has parenthesis "()". Parameter "c" at the equation (5) is not explained or named. There is a double comma "," in the first paragragh of the page 6.

Additional Comments:
My rating is based on the IJ reviewer guidelines.