The submission is related to the junction analysis of tubular objects which is quite challenging and, if manual, it could be very time consuming. The authors' approach is based on analysis of connected components in forking area. The code implementation is very good and ease to follow.

The authors provided the code, principle description, "how-to use it" manual, and test images. So the user can run the code on the images and see the results.

The authors do provide the source code along with the input images. After the code being run, the result could b compared to what the authors presented in the paper.

I personally downloaded the code and compiled it. The code was unfortunately oriented towards the MS Windows users and in some places did not follow the template name convention resulting in compiling errors on Linux/UNIX platform. After cosmetic corrections though, the code could be compiled on Linux/UNIX platform as well. Running the code produces the same results reported in the paper.

The authors provide their own experience and full information needed for prospective users. One disadvantage in the implementation is the lack of parallelism which is included by the authors in the future work.

The paper clearly describes the idea and how to use the code. The code is written in a way which easy to follow. Since I am a Linux guy, it took a while to change the code so it could be compiled on the Linux machine. Besides that, it does not take too long to start using the tool.

The code was easy to read and implemented in a modern style. Since it is oriented to MS Windows users, It needs just a little bit change to make it multi-platform.

Data is of high quality and it is very easy to use.

Any field where the junction analysis is required. Some examples are vasculature, airways, river basins, etc.

Very useful tool. On my laptop ( Intel CPU 1.73GHz, 2GB)  using the authors' image COWinput.mha with parametrs 2.0 3.0 the program spent time below.
real    65m8.260s
user    59m40.164s
sys    0m15.741s

That would be nice to compare the performance of this code with similar ones from mesh and/or point cloud analysis.

This paper describes a method for locating junctions in tubular structures. The method is demonstrated in 2D and 3D.

The paper comes with the full source code, as well as input and output images.

This paper is a perfect example of Open Science. It provide all the material required for replicating the work. The document describes the basic principles of the algorithm and then explains how to use the provided method.

The work presented in the paper can be fully replicated.

I downloaded the code, compiled it and ran it.

Compilation required several minor changes to the code due to the absence of the "typename" keyword in several locations. This is understandable, since the authors probably developed this code in a platform whose compiler doesn't provide full support for typenames (e.g. Visual Studio 6.0).  After repairing the typenames, the code compiled fine whith gcc 4.3.2 in an Ubuntu 8.10 GNU/Linux installation.

I ran the 2D example. (when compiled for Release it took 537seconds to run).

I'm still running the 3D example...

The description of the use of the filter is very clear, and the parameters required for it are related very directly to the geometry of the junctions, so selecting reasonable parameters is very easy.

This is a great example of a reproducible paper.

The authors use ITK as a base for implementing their junction detection method, and they use VTK for visualization of the results.

The paper doesn't elaborate on the use of open source but certainly provide a vivid example of how new contributions can build upon the material available in open source tools.

The code provided by the authors is in very good shape.

Authors describe very clearly how to use the code.

In general lines it satisfies the ITK coding style.

Minor changes were required to compile it under GNU/Linux with gcc.

I was able to use the code in about 30 minutes (due to the typename issues), otherwhise it would have been a matter of a couple of minutes.

Authors provided their source code.

The code is easy to read and well formatted.

Minor changes were required in order to compile it on Linux, but after them I would anticipate that the code is fairly portable.

Architecturally the GenerateMethod() could be split into several smaller methods, to follow the Single Responsibility Principle, which will make easier to maintain the code for the long run.

The quality of the provided data is very good.

The provided input segmentation of the Circle of Willies is a great dataset and will prove useful for other research.

The segmentation of the traquea is a binary image and may need an anti-aliasing for being reused for other purposes.

That being said, these are "input" datasets, therefore they don't diminish in any way the quality of the paper.

The authors provided the output images that result from running the algorithm, and they are consistent with the ones I generated when replicating their work.

This work is extremely useful, and there is a great variety of areas in which people in the community can take advantage of this contribution.

In particular:

• Vascular studies


• Segmentation of airways


• Analysis of roads in teledetection images


• Analysis of neurons in microscopy

The authors have made a great contribution to the ITK community.

The size and form of this paper is perfectly suited for the Insight Journal. It is a concise paper, it goes to the point, describes the principles clearly and explain how to use the software.

It it the ideal type of contribution, from which any reader will be able to build upon in a matter of minutes.

Great Work !