Document Type : Data Descriptor
Authors
1 Department of Medical Physics and Engineering, School of Medicine, Shiraz University of Medical Sciences, Shiraz, Iran
2 Research Center for Neuromodulation and Pain, Shiraz University of Medical Sciences, Shiraz, Iran
3 Department of Nuclear Medicine, Shiraz University of Medical Sciences, Shiraz, Iran
Abstract
Lung cancer is a leading cause of cancer-related mortality worldwide. [18F]-fluorodeoxyglucose ([18F]-FDG) Positron Emission Tomography/Computed Tomography (PET/CT) imaging is an advanced medical imaging modality used for lung cancer diagnosis. However, despite the availability of such an imaging modality, current methods of diagnosing cancer and interpreting images are prone to error. Therefore, Artificial Intelligence (AI)-based methods for medical image analysis are introduced and widely applied to enhance diagnostic accuracy. These methods, particularly deep learning techniques, require large datasets. Collecting image datasets, especially PET/CT images, is time-consuming, labor-intensive, and in some instances impossible due to the high costs and low availability of imaging centers. To facilitate further research in AI for lung cancer diagnosis, we present a publicly available 2D [18F]-FDG PET/CT lung image dataset comprising 3,061 image slices obtained from 94 patients. The dataset provides pathology-confirmed ground truth annotations together with patient-level identifiers, making it suitable for developing, validating, benchmarking, and comparing AI algorithms for lung cancer classification, localization, explainable AI, and other PET/CT-based computer-aided diagnostic applications.
Keywords
)