A Whole-Body PSMA-PET/CT dataset with manually annotated tumor lesions.
Authors
Affiliations (7)
Affiliations (7)
- Department of Radiology, LMU University Hospital, LMU Medizin, Ludwig-Maximilians-Universität München, Munich, Germany. [email protected].
- Comprehensive Pneumology Center (CPC-M), Member of the German Center for Lung Research (DZL), Munich, Germany. [email protected].
- Munich Center for Machine Learning (MCML), Munich, Germany. [email protected].
- Department of Radiology, LMU University Hospital, LMU Medizin, Ludwig-Maximilians-Universität München, Munich, Germany.
- Munich Center for Machine Learning (MCML), Munich, Germany.
- University Hospital Tübingen, Department of Diagnostic and Interventional Radiology, Medical Image and Data Analysis (MIDAS.lab), Tübingen, Germany.
- Department of Nuclear Medicine, LMU University Hospital, LMU Medizin, Ludwigs-Maximilians-Universität München, Munich, Germany.
Abstract
We describe a publicly available, large, annotated dataset of 597 whole-body Positron Emission Tomography/Computed Tomography (PET/CT) studies with Prostate-Specific Membrane Antigen (PSMA)-targeting radiotracers ([18 F]PSMA and [68Ga]Ga-PSMA-11) from 378 male patients with suspected or diagnosed prostate carcinoma. Scans were acquired between 2014 and 2022 on three clinical PET/CT scanners. The imaging protocol consisted of PET and diagnostic CT acquisitions extending from the skull base to the mid-thigh. All PSMA-expressing tumor lesions were manually segmented on the PET images in 3D space using dedicated software. The dataset includes anonymized DICOM files of all PET/CT studies, corresponding DICOM segmentation masks, and a TSV file with patient age, PET/CT manufacturer and model name, PET radionuclide, and information on whether CT contrast agent was used. We demonstrate how this dataset can be used for deep learning-based automated analysis of PET/CT. Together with a previously published whole-body Fluorodeoxyglucose (FDG)-PET/CT dataset, this dataset was provided in the Medical Image Computing and Computer Assisted Intervention Society (MICCAI) registered autoPET III and IV Grand Challenges to enable the development of multi-tracer machine learning models for automated lesion segmentation in whole-body PET/CT.