20 COVID-19 CT scans are included, with left lung, right lung, and infection regions labeled by two radiologists and verified by an experienced radiologist. The dataset was created by Jun Ma of Nanjing University of Science and Technology to support annotation-efficient deep learning methods. It is designed for three benchmark segmentation tasks using limited COVID-19 scans, existing non-COVID-19 lung CT data, and heterogeneous datasets.
Use Cases
- Benchmarking annotation-efficient segmentation models based on the described three-task framework.
- Training models to segment lung infections in CT scans based on the expert-labeled infection masks.
- Developing models for left and right lung segmentation using the provided anatomical labels.
- Studying transfer learning from non-COVID-19 lung disease datasets to COVID-19 cases as mentioned in the description.
Strengths
- 20 CT scans with expert annotations from multiple radiologists.
- Labels are verified by an experienced radiologist, suggesting quality control.
Limitations
- Row count and dataset size are unknown, which may limit suitability assessment.
- Column-level documentation is absent; field semantics must be inferred after download.
- Last update date is unknown; freshness unverified.
Provenance
- Source
- Nanjing University of Science and Technology
- Collection Method
- CT scans were labeled by two radiologists and verified by an experienced radiologist.