An open data commons linking patient-level radiopharmaceutical dosimetry to outcomes

Two open resources already exist but neither covers what is missing. OpenDose is a collaboration of 18 international teams that provides freely downloadable specific absorbed fractions and S-values for 1,252 radionuclides, generated and cross-checked using six different Monte Carlo radiation-transport codes; this is foundational physics reference data (how radiation deposits in standard geometries), not patient-level dosimetry from real treatments. The Cancer Imaging Archive (TCIA), and the NCI Imaging Data Commons that ingests it alongside other sources, hosts large public collections of clinical cancer imaging, including some radiotherapy dose and structure sets, under Creative Commons licences, but there is no dedicated TCIA-style collection pairing diagnostic and post-therapy radiopharmaceutical images with per-patient organ- and tumour-absorbed-dose calculations and clinical outcomes.

The practical consequence is that dose-response findings such as the LUMEN, LUTADOSE and University of Wisconsin dosimetry series described elsewhere in this graph are each built and validated on single-centre cohorts of a few dozen patients, because no shared, pooled, patient-level dataset exists to train or externally validate a dose-response or dose-toxicity model across institutions. This also limits how AI segmentation and dose-prediction models (see the AI dosimetry idea in this graph) can be externally validated, since most reported models are trained and tested within one centre's data.

No dedicated, open, patient-level radiopharmaceutical dosimetry-and-outcomes commons was found in this research; this is an idea rather than a project already under way, distinct from the isotope-supply observatory idea already in this graph, which tracks production and shortage data rather than dosimetry-and-outcome data.