{"entity":{"id":"idea-ai-lesion-segmentation-dosimetry","kind":"idea","name":"AI segmentation to make voxel-level radioligand dosimetry practical at scale","aka":[],"tldr":"Working out how much radiation a patient's kidneys and tumours received after a radioactive treatment normally means a specialist spending hours tracing organ outlines by hand on a scan. Software that draws those outlines automatically using artificial intelligence can do the same job in minutes, making personalised dose tracking available everywhere, not just in a few specialist centres.","summary":"Personalised, voxel-level dosimetry after radioligand therapy requires segmenting organs and tumours on serial SPECT/CT or PET/CT scans, a step that traditionally takes a trained physicist or physician several hours per patient using manual contouring. Deep-learning segmentation models, typically convolutional neural networks such as U-Net or newer transformer architectures, have been developed to automate this step. A 2025 comparison of AI-based against fully manual kidney segmentation for single-time-point 177Lu-PSMA dosimetry, using 8 treatment cycles from 5 patients, found no statistically significant difference in mean absorbed dose between the two methods, with a Dice similarity coefficient of 0.898 between AI and manual contours, though the underlying kidney volumes did differ significantly between methods.\n\nA broader review of automated dosimetry pipelines reports that combining CNN-based organ segmentation with automated registration and dose-rate estimation can cut total per-patient processing time from 4-6 hours to 30-45 minutes, a 70-90% reduction, while achieving Dice similarity coefficients of 0.88-0.95 against expert manual contours for organ segmentation specifically. Performance for tumour lesion segmentation, as opposed to organ segmentation, is less mature: an external, held-out-centre test of an nnU-Net lesion segmentation model for PSMA PET/CT and post-therapy SPECT/CT reported lower Dice scores of 0.76-0.78, and the model's PET-derived tumour volume estimates were described as unstable in that external test.\n\nAn independent 2026 evidence review of AI in oncologic theranostics concludes that, despite this technical progress, no AI segmentation or dose-prediction model for PSMA or peptide receptor radionuclide therapy dosimetry has yet been validated in a prospective, multicentre trial demonstrating that AI-guided activity adaptation (continuing, delaying or stopping treatment based on the AI-derived dose) actually changes patient outcomes; the review explicitly calls this the priority next study for the field.","asOf":"2026-09-22","links":[{"label":"Single time-point kidney dosimetry in 177Lu-PSMA therapy: AI-based vs manual segmentation (2025)","url":"https://he01.tci-thaijo.org/index.php/bulletinAMS/article/view/279571"},{"label":"Radiomics and Artificial Intelligence in Radiotheranostics (PMC review, 2024)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10814892/"},{"label":"Towards Routine AI-Based PET/CT and SPECT/CT Lesion Segmentation and Tracking in PSMA Theranostics (arXiv, 2026)","url":"https://arxiv.org/abs/2603.22666"},{"label":"Artificial intelligence across oncologic theranostics: evidence review (Frontiers in Nuclear Medicine, 2026)","url":"https://www.frontiersin.org/journals/nuclear-medicine/articles/10.3389/fnume.2026.1935247/full"}],"tags":["artificial intelligence","dosimetry","segmentation"],"related":["radioligand-dosimetry","in-vivo-dosimetry","radiomics","mim-software","quantitative-imaging-biomarkers","lu-177"],"indications":[],"sections":[],"technologies":[],"targets":[],"drugs":[],"companies":[],"institutions":[],"pathways":[],"terms":[],"trials":[],"people":[],"bottlenecks":[],"keyPapers":[],"journals":[],"dependsOn":[],"notes":[],"hypothesis":"Replacing manual contouring of organs and tumours on post-therapy SPECT/CT and PET/CT with deep-learning segmentation will cut per-patient dosimetry processing time from hours to minutes without a clinically meaningful loss of accuracy, making voxel-level personalised radioligand dosing practical to run routinely rather than only in specialist research centres.","rationale":"A direct comparison of AI-based versus manual kidney segmentation for 177Lu-PSMA dosimetry found no significant difference in mean absorbed dose (p=0.964) with Dice concordance of 0.898, and a broader review reports 70-90% reductions in total processing time using automated pipelines with Dice similarity of 0.88-0.95 for organ segmentation. However, tumour lesion segmentation specifically performs less well in external testing (Dice 0.76-0.78, unstable volume estimates in one held-out-centre study), and an independent 2026 evidence review states plainly that no prospective, multicentre, decision-impact trial has yet shown AI-guided dosing actually improves outcomes.","test":"A prospective, multicentre equivalence trial comparing AI-automated against expert-manual dosimetry-guided treatment decisions (continue, delay, or stop a treatment cycle) in patients receiving 177Lu-PSMA-617 or 177Lu-DOTATATE, with the primary endpoint being agreement between the AI and manual absorbed-dose estimates against a Monte Carlo reference standard, and a secondary endpoint of agreement in the resulting clinical decision, in roughly 100-150 patients across several centres.","maturity":"being-tested-at-scale","actor":"industry","cost":"small","horizonYears":4},"route":"/ideas/idea-ai-lesion-segmentation-dosimetry/","neighbours":{"technology":[{"id":"in-vivo-dosimetry","kind":"technology","name":"In vivo dosimetry and patient-specific quality assurance","route":"/technologies/in-vivo-dosimetry/"},{"id":"quantitative-imaging-biomarkers","kind":"technology","name":"Quantitative imaging biomarkers (RECIST, PERCIST, SUV, ADC)","route":"/technologies/quantitative-imaging-biomarkers/"},{"id":"radioligand-dosimetry","kind":"technology","name":"Radioligand dosimetry","route":"/technologies/radioligand-dosimetry/"},{"id":"radiomics","kind":"technology","name":"Radiomics","route":"/technologies/radiomics/"}],"company":[{"id":"mim-software","kind":"company","name":"MIM Software (GE HealthCare)","route":"/companies/mim-software/"}],"isotope":[{"id":"lu-177","kind":"isotope","name":"Lutetium-177","route":"/isotopes/lu-177/"}],"idea":[{"id":"idea-open-dosimetry-data-commons","kind":"idea","name":"An open data commons linking patient-level radiopharmaceutical dosimetry to outcomes","route":"/ideas/idea-open-dosimetry-data-commons/"}]}}