A Repository of Annotated PSMA and FDG PET/CT Images for Algorithm Development in Staging of mCRPC for Treament with 177Lu-PSMA Therapy

A curated dataset of PSMA (Ga-68 PSMA-617 or 18 F-DCFPyL) and FDG PET/CT images for 100 cases staged prior to treatment with Lu-177 PSMA therapy. Image data are paired per subject with one PET/CT scan for each tracer and independent annotations of total disease burden for each of PSMA- and FDG-avid disease. 

PSMA PET/CT scans are annotated based on SUV threshold >= 3 with manual removal of areas of physiological tracer uptake. Threshold used for FDG PET/CT scans is liver-based, variable among individuals. Works describing the annotation methodology as applied to separate cohorts can be found in Ferdinandus et al. and Buteau et al.. 

Data is provided in a format (*.nii.gz) to facilitate algorithm development as part of the DEEP-PSMA challenge. All cases contain some presence of measurable disease on both PSMA and FDG images. CT and PET images are provided in their native resolution and Total Tumour Burden (TTB) contour is aligned to the PET image coordinates. The appropriate SUV threshold used for contouring is provided as well as a rigid registration in SITK format for course alignment between PSMA and FDG image sets. For more detail please review the Dataset Format page on the Grand Challenge site.

The best-performing methods along with link to algorithm description and source code repositories can be found on the Final Testing Phase Leaderboard.

This research work is supported in part by the Prostate Cancer Foundation through the Peter MacCallum Cancer Centre and the ProsTIC Centre of Excellence.

We would like to acknowledge the contributions of Arthur Dujardin and Paul Blanc-Durand to assist with review and update of several annotations from the training dataset after the challenge completed.