Update September 30, 2019

The challenge is closed. Due to the low submissions, no results will be published. 

Update March 20, 2019

The dataset is released, see downloads after registration to grand-challenge.org. Soon via the journal Medical Physics** **the paper: "Accuracy of registration algorithms in subtraction CT of the lungs: a digital phtanom study" wil be published online. This paper shows the methodology of how the images are obtained and what the accuracy is of subtraction CT algorithm. There is a trainingset which will not be used for the results and not be checked by the organisation. Let us know if something is not working. 

Update November 14, 2018

At the moment we are working on the dataset. The difference will be that at the moment the images only have noise from a water phantom CT scan, while in future (approximately within a few months) we will obtain images that are CT simulated, which includes realistic noise/streak artifacts. When the images are available, we will uploadthe images. We hope that this gives a more accurate result of the accuracy of the registration algorithm.

Aim

The aim of this challenge is to determine the accuracy of registration algorithms, using an anthropomorphic digital phantom, in registering unenhanced and enhanced computed tomography (CT) chest images. The challenge calculated the voxel-by-voxel difference between the ground truth and the algorithm-estimated deformation in the lungs. 

Abstract/Motivation

Distribution of intravenously iodinated contrast in the lungs can be used as a marker for pulmonary perfusion, for instance for assessing perfusion defects related to pulmonary embolism, tumor compression, bronchopathy. Or perfusion caused by lung masses. With an unenhanced and enhanced CT scan an temporal subtraction CT imaging can be obtained. This includes the removal of vessels, filtering and adding of a appropriate colour scale, but more important is the registration of the unenhanced towards the enhanced. The geometry should be identical to subtract the two images from each other. There are already other challenges for lung registration, but the difference of this challenge is the unenhanced registering to enhanced scans and the fact that each lung voxel is compared for their displacement.   

Dates

Webpage launched                                             February 22, 2018

Training data released                                        March 28, 2018         

Test data released                                               April 5, 2018        

Submission                                                          April 30, 2018

New data released                                               March 20, 2019