Predictions for Image Quality < 1

For each slice, we get either the prediction mask for the left or the right. The algorithm will also make predictions for slices with image quality < 1, how will those predictions be treated during evaluation? Additionally, the annotation is sparse so the algorithm will also give prediction for the intermediate slices for which the annotation is missing.

Please clarify the evaluation criteria using dice score for slices with image quality < 1 or with missing annotations.

Thanks in advance.

• The slices with image quality < 1 will not included in the evaluation phase.
• The annotation format of the test set is similar to that of the training set, i.e., part of the slices in the test set are labeled. During the evaluation phase, only the slices with ground-truth annotations will be used for the calculation of quantitative metrics, e.g. DSC, 95 HD. We will not specify the annotated slices in advance, therefore, participants should provide results for each slice.