MICCAI’12 Workshop

Post-workshop update:

During the challenge, participants ran their algorithms on the Test2 dataset. Ranking of the teams was done on the results obtained during the onsite challenge.
  • Best automatic endocardium segmentation:
Maria A. Zuluaga, M. Jorge Cardoso and Sébastien Ourselin, Centre of Medical Image Computing University College London, London UK
  • Best semi-automatic epicardium and endocardium segmentation :
Wenjia Bai, Wenzhe Shi, Haiyan Wang, Nicholas S. Peters, and Daniel Rueckert, Biomedical Image Analysis Group, Department of Computing, Imperial College London, UK & National Heart and Lung Institute, St Mary's Hospital, Imperial College London, UK
  • Best semi-automatic endocardium segmentation (2 ex-aequo):
Oskar M O Maier, Daniel Jimenez Carretero, Andres Santos Lleo, and Mara J Ledesma-Carbayo, Escuela Tecnica Superior de Ingenieros de Telecomunicacion, Universidad Politecnica de Madrid, Madrid, Spain. Wenjia Bai, Wenzhe Shi, Haiyan Wang, Nicholas S. Peters, and Daniel Rueckert, Biomedical Image Analysis Group, Department of Computing, Imperial College London, UK & National Heart and Lung Institute, St Mary's Hospital, Imperial College London, UK Photos from the workshop are available here. Results obtained on Test1 dataset are available here Workshop proceedings are available at the end of this page.
 

Workshop program (October 1st, 2012)

The workshop will start with the presentation and distribution of the onsite testing datasets and the start of the challenge. For the remainder of the morning session, participants will run their algorithms on this data and concurrently a poster session will take place. The organizers will evaluate the results with reference data during the lunch break. After lunch, we intend an invited lecture and a set of selected talks (based on the pre-workshop results). Following these presentations, we will have a poster session, where the participants can explain their algorithm, possibly including on-site demos. The afternoon session will conclude with a presentation of the challenge results, and a town hall discussion. Tentative program
Morning 9.00 -12.30 Presentation and distribution of the on-site testing datasets
On-site challenge: participants will run their algorithms on this data
Include a coffee break (10.30 - 11.00)
Lunch 12.30-14.00 Lunch break (included in the registration)
Concurrent poster session
Evaluation of the on-site challenge results by the organizers
Afternoon 14.00-15.00 Guest speaker lecture by Prof. J.C.H. Reiber (Medis, the Netherlands) "Innovations in diagnostic imaging technologies: basic principles and clinical perspectives of cardiac CTA and MRI"
15.00-15.30 Presentations from both challenges
15.00-15.15 “Automatic detection, quantification and lumen segmentation of the coronary arteries using two-point centerline extraction scheme”, Rahil Shahzad et al. (BIGR, Rotterdam, NL)
15.15-15.30 “Multi-Atlas Based Segmentation with Local Label Fusion for Right Ventricle MR Images”, Wenjia Bai et al. (Imperial College London, UK)
15.30-16.00 Coffee break Concurrent poster session The participants can explain their algorithm and possibly include on-site demos
16.00-16.30 Presentations from both challenges
16.00-16.15 “Accurate Stenosis Detection and Quantification in Coronary CTA”, Brian Mohr et al. (TMVSE, Edinburgh, UK)
16.15-16.30 “Automatic Right Ventricle Segmentation using Multi-Label Fusion in Cardiac MRI”, Maria A. Zuluaga et al. (University College London, UK)
16.30-17.30 Presentation of the challenge results Prizes giving Closing remarks

List of accepted papers

A Simple and Fully Automatic Right Ventricle Segmentation Method for 4-Dimensional Cardiac MR Images, Ching-Wei Wang, Chun-Wei Peng, and Hsiang-Chou Chen, Graduate institute of biomedical engineering, National Taiwan University of Science and Technology, Taipei, Taiwan
Right-Ventricle Segmentation with 4D Region-Merging Graph Cuts in MR, Oskar M O Maier, Daniel Jimenez Carretero, Andres Santos Lleo, and Mara J Ledesma-Carbayo, Escuela Tecnica Superior de Ingenieros de Telecomunicacion, Universidad Politecnica de Madrid, Madrid, Spain.
Multi-Atlas Based Segmentation with Local Label Fusion for Right Ventricle MR Images, Wenjia Bai, Wenzhe Shi, Haiyan Wang, Nicholas S. Peters, and Daniel Rueckert, Biomedical Image Analysis Group, Department of Computing, Imperial College London, UK & National Heart and Lung Institute, St Mary's Hospital, Imperial College London, UK
Automatic Right Ventricle Segmentation using Multi-Label Fusion in Cardiac MRI, Maria A. Zuluaga, M. Jorge Cardoso and Sébastien Ourselin, Centre of Medical Image Computing University College London, London UK
Right ventricle segmentation by graph cut with shape prior, Damien Grosgeorge, Caroline Petitjean, Su Ruan, Jérôme Caudron, Jean-Nicolas Dacher, LITIS EA 4108, Université de Rouen, France
Multi-Atlas Segmentation of the Cardiac MR Right Ventricle, Yangming Ou, Jimit Doshi, Guray Erus, and Christos Davatzikos, Section of Biomedical Image Analysis (SBIA), Department of Radiology, University of Pennsylvania
Rapid Automated 3D Endocardium Right Ventricle Segmentation in MRI via Convex Relaxation and Distribution Matching, Cyrus M.S. Nambakhsh , Martin Rajchl, Jing Yuan, Terry M. Peters, Ismail Ben Ayed, GE Healthcare, London, Ontario, Canada & Western University, London, Ontario, Canada & Robarts Research Institute, London, ON, Canada
 

Guest speaker

This year, the workshop organizers are honored to welcome Prof. J.H.C. Reiber, a world class expert capable of providing deep insight into the latest developments in cardiac CTA and MRI. Biography of Prof. Dr. Ir. J.H.C. Reiber Johan H.C. Reiber received his M.Sc. EE-degree from the Delft University of Technology in 1971 and his M.Sc and Ph.D. from Stanford University, USA in 1975 and 1976, respectively. In 1977, he founded the Division of Image Processing (LKEB) at the Thoraxcenter, Erasmus University in Rotterdam, and continued these activities from 1990 at the Department of Radiology, Leiden University Medical Center (LUMC) in the Netherlands. Since 1995 he has been Professor of Medical Image Processing at the LUMC, and from 1995 until 2005 also a professor of cardiovascular imaging at the Interuniversity Cardiology Institute of the Netherlands (ICIN) in Utrecht. In 2000 he became a member of the Royal Netherlands Academy of Arts and Sciences (KNAW), Physics faculty, section Technical Sciences. He is (co)-author of more than 625 scientific papers, and co-author/editor of 15 books. He is editor-in-chief of the International Journal of Cardiovascular Imaging, and serves on the Editorial Board of several other journals.  In 2004 he became an IEEE Fellow for his contributions to medical image analysis and its applications. Other fellowships include those of the European Society of Cardiology (1988) and the American College of Cardiology (2010). He is also co-founder and CEO of Medis medical imaging systems bv in Leiden, a global provider of software packages for the quantitative analysis of medical images, in particular of the cardiovascular system.

Prizes

During the on-site workshop, the Test2 dataset will be processed by the participating teams. There will be 3 categories:
  • Fully automatic segmentation methods of the endocardium only
  • Semi-automatic segmentation methods of the endocardium only
  • Fully automatic and semi-automatic segmentation methods of the endocardium and epicardium
The winners in each category will be rewarded by a life-size human heart model, sponsored by Pie Medical Imaging. Participants which segment both the epicardium and endocardium may also enter the "endocardium only" categories. A winner team ranked 1st in several categories will only get one prize, so that the prize will be transfered to the 2nd ranked participant.

How to participate?

Update (October 7th, 2012): Participation to the challenge as a MICCAI'12 workshop is now closed. You may still download and use the data, please follow this link.   Below are the instructions for the MICCAI'12 Challenge participants:
  1. Download, fill in and send back the registration form (please see below) to Caroline Petitjean (caroline.petitjean@univ-rouen.fr). This registration is solely intended to the RVSC organizers. It does not compel to the registration to the MICCAI’12 challenge, but is a mandatory step before downloading data and submitting results to the challenge.
  2. Upon reception of the registration form, you will be allowed to download the Training Set (16 patients). The Training database includes DICOM images, a list of images to be segmented, and manually segmented epicardium and endocardium contours. ED and ES phases are already defined, as well as basal and apical slices.
  3. On June 4th, the Test1 Set (17 patients) will be released. For submission, each participant may send an email to Caroline Petitjean, attached with the segmentation results of endocardium and epicardium, or endocardium only. Within 48 hours, full details of the method’s performance analysis will be sent back to the participant.
  4. Challengers may submit a paper summarizing their method and results (4-8 pages max, LNCS format) to the RVSC organizers.
  5. On the day of the workshop (Oct 1st), the Test2 Set (17 patients) will be released for on-site challenge. There will be 3 categories:
  • Fully automatic segmentation methods of the endocardium only
  • Semi-automatic segmentation methods of the endocardium only
  • Fully automatic and semi-automatic segmentation methods of the endocardium and epicardium
The collation paper will be compiled only from participants who submitted their results on the day of the workshop.  

Deadlines

Training data ready for download March, 19th 2012
Testing data ready for download June, 4th 2012
Submission deadline July 5th 2012 (instead of June, 29th)
Notification of acceptance to participants July, 9th 2012
Camera-ready paper July, 30th 2012
On-site testing data ready for download October, 1st 2012
Workshop
  Our challenge guidelines have been established by following examples of the MICCAI’09 LV segmentation challenge, the MICCAI’11 STACOM Cardiac Left Ventricular Segmentation Challenge, and the MICCAI’12 Rotterdam Coronary Artery Algorithm Evaluation Framework.