- Submitted (semi)-automatically obtained segmentations on the provided test images are evaluated according to the description on the Evaluation page.
- Please note that only segmentations of the non-contrast enhanced images are evaluated, as these images are clinically used for calcium scoring. If your method provides segmentations in the contrast-enhanced image, please make sure to map these to the non-contrast enhanced image.
- Obtained segmentations may be submitted for evaluation up to two times.
- Submitted segmentations should comply with the following criteria. For each test image, one short-typed image with results is required. Every voxel in this image should have one of the following values:
- 0: The voxel does not belong to a coronary calcification. Background value.
- 1: The voxel belongs to a coronary calcification in the left anterior descending (LAD) coronary artery. Please note that in this challenge, calcifications in the left main coronary artery are labeled as LAD calcifications.
- 2: The voxel belongs to a coronary calcification in the left circumflex (LCX) coronary artery.
- 3: The voxel belongs to a coronary calcification in the right coronary artery (RCA).
- In case your method does not make a distinction between coronary calcifications in the different major coronary arteries, an additional label should be used:
- 4: The voxel belongs to a coronary calcification, the coronary artery was not determined.
- Connected groups of voxels (3D 6-connectivity) with intensities greater than 130 HU should present one calcification and therefore all voxels in that calcification should have the same label. When this is not the case, the evaluation software will assign the majority label to all connected voxels.
- Please note that your results must be submitted in MetaImage format (mhd+raw file). This format consists of a text file ending in .mhd which contains ASCII header information and a binary file ending in .raw that contains the voxel data. ITK supports this format. For more information see this page.
- Please name your segmentation files by replacing the 'CTI' in the CT image filename with an 'S'. E.g., the segmentation file for CT image 'TEV1P1CTI.mhd' should be named 'TEV1P1S.mhd'.
Results of your segmentations will only be published on the website when a document (e.g. a journal paper) describing the method is provided. An exception will be made in case commercially available software is evaluated. Such teams need to specify the software and the exact version that was used.
Documents are limited to maximum 8 pages (suggested length 4 pages).
Please include the following details in your document:
- Specify whether the method is automatic or semi-automatic. Please specify what user input is required.
- Specify whether the method identifies in which coronary artery each calcification is located?
- Specify whether the method uses only the CT images or both the CT and CTA images?
- Provide the overall structure of your algorithm and describe each step. If post- and preprocessing are used, please specify these steps as well.
- Specify whether the algorithm was trained with data other than that provided in this challenge. If so, please carefully describe additional data.
- In case the algorithm has been tested on other databases, please consider including those results as well.
- Specify the average runtime of your algorithm and the system on which this was achieved.
- List limitations of your algorithm. Did the algorithm work on all images? Or were there vendor-specific or risk-group specific differences?
File upload is available to registered users only. Please build an archive (RAR or ZIP) containing your obtained segmentations and a document describing your method and upload this archive file.