Skip to main content

Table 2 Criteria of the predefined visual scoring system for simulated segmentation variations

From: An evaluation of performance measures for arterial brain vessel segmentation

Score Combined error severity Segmentation quality Scoring criteria
1 Low High • Minor errors with a typically low number of false positive or false negative voxels with minimal deterioration of segmentation quality and/or
• Minor to moderate boundary errors
2 • False-positive labelling of a low number of voxels not associated with any anatomical structure and/or
• False-positive labelling of parts of an anatomical structure and/or
• Parts or arterial segments of the arterial tree are represented without major errors and/or
• Moderate to severe boundary errors
3 Moderate Moderate • False-positive labelling of at least one defined anatomical structure and/or
• Parts or arterial segments of the arterial tree are missing and/or
• Severe boundary errors
4
5 • False-positive labelling of anatomical structures in multiple locations and slices and/or
• Major parts or arterial segments of the arterial tree are missing
6
7 High Low • False-positive labelling of anatomical structures in multiple locations and slices significantly compromising segmentation quality and/or
• Major parts or multiple arterial segments of the arterial tree are missing
8
9 • No/failed discrimination between vessels and other anatomical structures and/or
• Major parts or multiple major artery segments of the arterial tree are missing
10
  1. An error severity score was assigned to each simulated segmentation variation based on visual assessment. Higher scores indicate higher combined severity of errors in the segmentation and therefore lower quality of the segmentation. For example, simulated segmentation variations with a score of 7 to 10 are considered low quality and receive a high severity score. Due to the “and/or” criterion one point from each category is enough to assign a score. In higher error severity scores criteria from lower scores can also be fulfilled. For instance, a segmentation with a score of 9 can contain a severe boundary error but this criterion is not listed again under the criteria for score 9 since it is mentioned previously within the criteria of score 3