Manual segmentation of opacities and consolidations on CT of long COVID patients from multiple annotators

The field of supervised automated medical imaging segmentation suffers from relatively small datasets with ground truth labels. This is especially true for challenging segmentation problems that target structures with low contrast and ambiguous boundaries, such as ground glass opacities and consolidation in chest computed tomography images. In this work, we make available the first public dataset of ground glass opacity and consolidation in the lungs of Long COVID patients. The Long COVID Iowa-UNICAMP dataset (LongCIU) was built by three independent expert annotators, blindly segmenting the same 90 selected axial slices manually, without using any automated initialization. The public dataset includes the final consensus segmentation in addition to the individual segmentation from each annotator (360 slices total). This dataset is a valuable resource for training and validating new automated segmentation methods and for studying interrater uncertainty in the segmentation of lung opacities in computed tomography.