A growing phenomenon in the world of medical image processing is that of “Grand challenges”. The great advantage of such challenges is that all algorithms are evaluated on the same data and using the same evaluation procedure.
This year at the MICCAI conference the Biomedical Imaging Group Rotterdam of the Erasmus MC organized a challenge on the automatic detection and quantification of stenoses in coronary CTA data (http://coronary.bigr.nl/stenoses/).The challenge contained 48 multi-center multi-vendor CTA datasets. In total 11 teams from different institutes participated in the challenge. Teams had to (semi) automatically detect and quantify coronary lesions in the CTA datasets and submit their results to an online evaluation framework. Teams where ranked according to a number of criteria in three categories: detection (ability to detect significant stenoses), quantification (comparison with observer and QCA stenoses grades) and finally lumen segmentation (compared to observer drawn contours).
At the Division of Image Processing (LKEB), Department of Radiology at the Leiden University Medical Center (LUMC), the CTA group have developed and validated robust algorithms for extraction and segmentation of coronary CTA data. This work has largely been done in within the STW CADAStr project in close collaboration with the LUMC department of cardiology and Medis medical imaging systems. These algorithms have already contributed to a number of international high ranking publications both from inside and outside the LUMC.
On the 1st of October, at the start of MICCAI 2012 conference in Nice, France a team from the LKEB/Medis (Alexander Broersen and Pieter Kitslaar), ranked 1st place in the detection and 2nd place in the quantification and segmentation categories. Based on these ranks, they were the overall winners of the on-site challenge. Congratulations, job very well done!! The algorithms are now part of the QAngio CT RE analytical software product; commercialized by Medis Specials bv (www.medisspecials.com) a Medis BV company.