A project led by Dr. Diedre Carmo has just released a set of CT images of COVID-19 subjects, along with expert annotations of anomalous regions, for use in developing automatic segmentation algorithms.

The field of supervised automated medical imaging segmentation suffers from the lack of annotated groundtruth data. This problem is even more noticeable when dealing with the segmentation of multiple types of lung findings in computed tomography with uncertain borders, such as opacities and parenchymal consolidation resulting from pneumonia. In this work, we make available the first public dataset of ground glass opacity and consolidation in the lung 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. We make available not only the final consensus segmentation, but also the individual segmentation from each annotator totaling 360 slices. This dataset can be used to train and validate new automated segmentation methods and to study interrater uncertainty in lung opacities' segmentation on computed tomography.

The data set is available from the Iowa Research Online website: https://doi.org/10.25820/data.007301