Abstract
The subject of the paper is to introduce a new class of 4D filters to itk.While it is very easy to load a 3D volume stored as a series of DICOM images with ITK, some convenience functions that are already available for 3D volumes are not yet available for the 4D case. Especially for MR data (and to a certain limit also for CT data) it is very important to look at the full 4D series, since the radiologist usually does not make a diagnosis based on a single volume but on the full time series of volumes which shows the temporal flow of the contrast agent. As an example, for breast imaging the radiologist typically examines parameter maps like MIPt (maximum intensity over time), TTP (time to peak) or WI (Wash-In). These parameter maps are the 3D result of post-processing the entire 4D time series. The goal of proposing the denoted 4D filters is to provide ITK with functionality that enables a postprocessing job to handle the entire 4D series as one entity and not as a loose collection of 3D volumes.
Keywords
Source Code and Data
Reviews
Mathieu Malaterre
Monday 27 August 2007
Summary:
How to user ITK when dealing with 4D dataset
Reproducibility:
[Did you reproduce the authors' work?
yes
Did you download their code? Did you compile it? Did you run it?
Compilation went fine on linux debian/gcc 4.2 (some warnings). It did run fine.
Did you managed to get the same results that they reported?
I did get the exact same results
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.
Paper is cristal clear.
Open Source Contributions:
Author did provide the source code. It is fairly easy to read.
Code Quality:
The code is simple and fairly well written.
What need to be done for inclusion in ITK:
- Add concept checking (Input image is 4D, output is 3D)
- The filter is describe as being multithreaded in the documentation, but implement GenerateData in the txx file (single threaded version). The filter needs to be rewritten to implement the correct threaded version. This is highly important for such massive amount of data (4D dataset).
- Fix all the warnings
Applicability to other problems:
The filter is very generic and can be used in multiple situation involving most actual data extraction from 4D dataset.
Suggestions for future work:
See code quality
Additional Comments:
I would highly urge the community to push this filter in ITK for the next release.
The document also adress a current issue in ITK to deal with 4D dataset and DICOM. Most manufacturer are providing 4D dataset for MR as a set of 2D images. Unfortunately ITK ImageSeriesReader are only able to reader N into N+1 image. Whereas we would need a N into N+2 ImageSeriesReader.