Data

 

    In our institution we have collected 45 cardiac cine-MR datasets with expert contours, as part of the clinical routine. Cine steady state free precession (SSFP) MR short axis (SAX) images were obtained with a 1.5T GE Signa MRI. All the images were obtained during 10-15 second breath-holds with a temporal resolution of 20 cardiac phases over the heart cycle, and scanned from the ED phase. Six to 12 SAX images were obtained from the atrioventricular ring to the apex (thickness=8mm, gap=8mm, FOV=320mm320mm, matrix= 256256). In these SAX MR acquisitions, endocardial and epicardial contours were drawn by an experienced cardiologist in all slices at ED and ES phases, following the convention of including papillary muscles and endocardial trabeculations in the ventricular cavity. All the contours were confirmed by another cardiologist. The manual segmentations will be used as the ground truth for evaluation purposes.

    All the data will be in anonymized DICOM format following HIPAA regulations, and permission for the described open database has already been granted by our institution’s ethics review board. The imaging data will be provided without any pre-processing after reconstruction, as it is typically presented to physicians for analysis. The datasets will be randomly divided into three groups: 15 for training, 15 for testing and 15 for the on-line contest. Teams registered with a letter of intent can download both the training and testing datasets from the workshop website.  The training datasets are provided for participants to optimize their algorithms, which have both images and ground truth segmentation contours. The testing datasets are presented to blinded participants, in that the data will not include ground truth segmentations. Participants will send in their segmentation results in standard format based on the testing datasets for evaluation by a given deadline.  The evaluation results are then sent back to the participants for preparation of their workshop papers. The on-line test datasets will be released before the start of the on-line competition during the MICCAI conference.  Segmentation results on these datasets are required to be submitted in three hours for evaluation.  All submitted segmentation results will be assessed using an evaluation software, which can also be downloaded from the workshop website for teams registered with a letter of intent. Qualitative results will be provided only for the testing results, not the on-line portion of the contest. The evaluation methods will be described in the next section.