ChestPathCT5-S100: an open real-world chest CT dataset of five common thoracic pathologies with heterogeneous...

We present ChestPathCT5-S100, an open dataset of 87 real-world chest CT examinations spanning five common thoracic pathologies: rib fracture, pleural effusion, lung mass, pulmonary embolism, and pneumothorax. The dataset was assembled from a retrospective single-centre cohort over a ten-year period, intentionally preserving acquisition heterogeneity and concomitant findings representative of routine clinical practice. Cases include both contrast-enhanced (arterial and venous phase) and non-contrast examinations, drawn from emergency, oncologic, and trauma settings. All imaging volumes are provided in NIfTI format. Technical validation by two radiologists confirmed correct pathology category assignment, image integrity, and unambiguous visibility of the dominant pathology for each case. A binary co-occurrence matrix of concomitant findings is provided to support multi-label research designs. ChestPathCT5-S100 is publicly available on Zenodo under a CC0 1.0 license, permitting unrestricted use and redistribution. The dataset supports classification, detection, weakly supervised learning, and multi-task learning paradigms.