Summary:
[This paper presents a set of tools for the computation of 3D structural statistical shape analysis using the concept of spherical harmonics. ]

Hypothesis:
[NA]

Evidence:


Open Science:
[The source code is provided. The paper does provide a test study where the shape analysis of the caudate has been done.]

Reproducibility:
[The code did not build on Visual studios.net. It gives a link error for the shapealgorithm and the shapematrix project even after building the lapack lib.
I downloaded the code and compiled it on linux platform. It builds well on linux.

Use of Open Source Software:
[Based on ITK.]

Open Source Contributions:
[Yes the author provides the source code, and it is very simple to build on linux platform.]

Code Quality:


Applicability to other problems:
[3D shape analysis can not only be used in neuroimaging, but for different shape analysis problems. ]

Suggestions for future work:
[It would be of great help to have more statistical analysis incorporated in the SPHARM code.]

Requests for additional information from authors:
[The paper explains all the parts very well. The example helps to understand how to use SPHARM kit.]

Additional Comments:

Summary:
This is an extensive paper describing a methodology for shape analysis as well as the usage of a set of shape analysis tools. The premises of the methodology are desribed and then mapped to the workflow one would use to apply it to a particular problem.

Hypothesis:
Non applicable.

Open Science:
Source code is provided.

Reproducibility:
Code would not build on my system (Visual Studio .Net 2003).

The ShapeAlgorithms and SparseMatrixLib projects would not load into Visual Studio.

Use of Open Source Software:
Software is based on ITK. Workflow includes the use of ITKSnap

Open Source Contributions:
Source code is provided.

Code Quality:
The majority of the code (from the user experience) is command line executables. There are ITK filters, sources, mesh subclasses under the hood for doing Procrustes alignment, assessing false discovery rate, generating equal area meshes, assigning attributes to meshes.

While the above are derived from ITK classes, some modifications will be needed to adhere to ITK style guidelines.

Applicability to other problems:
This contribution can be used for a variety of shape analysis tasks. The Procrusted alignment will have applicability to wide set of applications. The Spherical Harmonics can also be used as a feature is detection. And False Discovery Rate analysis is a general statistical approach that can be re-used.

Suggestions for future work:
I'd like to see more the analysis and processing be represented as ITK filters and statistical algorithms.

Requests for additional information from authors:

Additional Comments:
Compilation issues need to be addressed. Please let me know if you do not have access to the compiler I used.

Summary:
This work presents a pipeline for statistical shape analysis of brain structures, using spherical parametrization, and spherical harmonic description for each surface, in order to compare similar structures. This work includes issues such as preprocessing, alignment, scaling, as well as a local and global testing procedure in order to show differences between groups of similar shapes, using a suitable metric (Hotelling T^2).

Hypothesis:
The SPHARM description of 3D structures is a good representation to perform significance tests, because it is a hierarchical, global and multi-scale boundary description.
The maximal degree for the SPHARM computation is chosen between 12 (hippocampus) and 15 (caudate, lateral ventricle) and the sampling level for the icosahedron subdivision in the SPHARM-PDM is chosen between 10 and 20.

Evidence:
The main hypothesis is supported by mathematical background, the experiments described and the figures illustrated in the paper, but those experiments and figures can not be repeated because the data is not available.

Open Science:
The authors provide the source code, but they do not provide the data to reproduce the results of the main text.

Reproducibility:
I was able to download, and successfully compile the code in a Linux workstation (using ITK 2.4.1).
However I could not execute any example, because it is not provided. Only some meshes and spherical parametrizations of the knee are provided.
In the main text there is documentation about the usage of the code, as well as a script to compute SPHARM-PDM descriptions, with the code provided.

Use of Open Source Software:
The authors used open source software (ITK and lapack), and part of the code belongs to a library called neurolib.
Connection with a graphical interface KWMeshVisu to visualize results is also justified.

Open Source Contributions:
The source code is completely included. There exist several independent command line executables:
SegPostProcess: extracts single label of segmented data, resamples the data to obtain good resolution, and ensures spherical topology.
GenParaMesh: extracts the surface mesh from the segmentation, and maps it to a sphere.
ParaToSPHARMMesh: computes the SPHARM-PDM representation.
StatNonParamTestPDM: Computes the significance tests.

Code Quality:
The code is written clear, in modern style, is commented, and in a first view it seems to be clear to understand and to use.

Applicability to other problems:
We agree with the authors that this tool can be also applied to other shape analysis problems with spherical topology, but I\'m not aware about any other examples.

Suggestions for future work:
I think that a fusion with the KWMeshVisu interface to facilitate the use of the whole software could be very helpful.

Requests for additional information from authors:

Additional Comments:
In page 10 \"Bonferroni correctection\" should be \"Bonferroni correction\"
Caption in figure 9, page 16 has typos: \"as well an vectorfield\" should be: \"as well as a vector field\"
In page 16 \"Fiigure 10\" should be \"Figure 10\"
In page 17 \"Significant\" should be lowercase.

The paper seems too long to me. Although the theoretic part is important for the understanding of the code, I think that it can be reduced to give a more direct and more general overview about the pipeline, pointing to the reference for the details.

My final comment is that the work presented here is quite promising, is well presented, and it can be very helpful to shape analysis applications, although no experiments can support my review.

Summary:
This paper proposed a framework to perform a shape analysis on a set of segmentated images. A PDM is generated from the set of segmentations. Finally the shape analysis is performed on the PDM.

Hypothesis:
N/A

Evidence:
N/A

Open Science:

Reproducibility:

What I have used are the excutables precompiled and provided by the authors. I didn\'t have a chance to look at the code. But I did run the excutables for our data.

Applicability to other problems:
I did run the excutable for our data. The result is under analysis and looks promissing. The tools did a very good job in constructing a PDM from segmented images. The parametrization were successive on 38 out 40 images.

Improvements:

The third line from the bottom on page 16, there is a typo "Fiigure 10", which is supposed to be "Figure 10".