Hi
In the picture below, there are a CT slice (on the left) and its lung lobes segmentation from TotalSegmentator (on the right)
The picture shows that, on the right lung, the lobes segments border does not lie over the lung fissure (yellow arrow shows fissure, red arrow shows lobes segments limits) so that could be improved. At the same time, the left lung fissure is very dim, but TotalSegmentator identifies it well.
Do you think there could be any chance of getting better results training an AI with TotalSegmentator dataset just for lung lobes segmentation? If so, how?