Automated Dosimetry: Reducing Radiation Exposure Through Precision

Radiopharmaceutical therapy - treatment using radioactive agents that concentrate in tumor tissue - is one of the fastest-growing areas of oncology. Lu-177 PSMA for prostate cancer, Lu-177 DOTATATE for neuroendocrine tumors, Y-90 microspheres for liver cancer, and the emerging Ac-225 agents for alpha-emitter therapy all depend on the same fundamental principle: delivering sufficient dose to tumor while sparing organs at risk. The challenge is that "sufficient dose" and "safe dose to organs" vary substantially from patient to patient based on body composition, renal function, tumor burden, and prior treatment history. Fixed-activity dosing, the current standard in most centers, ignores this variation entirely.

Why Fixed-Activity Dosing Is a Clinical Problem

The standard approach to Lu-177 PSMA treatment administers 7.4 GBq per cycle regardless of patient-specific biodistribution. This protocol was established based on population-average safety data from clinical trials, designed to keep the 95th percentile patient within bone marrow and kidney dose limits. In practice, this means most patients receive substantially less than the activity that their organs could safely tolerate - and some patients with below-average renal uptake receive dose to their kidneys that approaches or exceeds the threshold for late nephrotoxicity.

A 2022 analysis of Lu-177 PSMA treatment data from a large German academic center found that individualized dosimetry-based dosing would have permitted dose escalation in 43% of patients and required dose reduction in 12%, compared to the fixed 7.4 GBq protocol. The dose-escalated cohort had meaningfully higher PSA response rates. The 12% requiring reduction had renal dosimetry that would have placed them at elevated nephrotoxicity risk on the standard protocol. Fixed dosing served approximately 45% of the population optimally.

The reason individualized dosimetry is not standard practice is operational, not clinical. Performing a full dosimetry calculation requires serial SPECT/CT imaging at 24, 96, and 168 hours post-injection; organ segmentation on each time point; activity quantification after SPECT scatter and attenuation correction; fitting a time-activity curve to each organ's data; and numerical integration using S-values or Monte Carlo simulation to convert absorbed dose in gray. This workflow takes 4-8 hours of physics and technologist time per patient - time that most centers do not have for every patient in a growing RPT program.

What Automated Dosimetry Actually Does

Automated dosimetry software addresses the bottleneck at multiple points in the workflow. Organ segmentation - historically a manual process requiring a medical physicist to contour kidneys, liver, spleen, and tumors on each SPECT/CT time point - is the most time-intensive step and the first target for automation. Deep learning segmentation models trained on multi-timepoint SPECT/CT data from Lu-177 studies achieve Dice similarity coefficients of 0.88-0.93 for kidneys, 0.91-0.95 for liver, and 0.82-0.89 for spleen - sufficient accuracy for clinical dosimetry use.

After segmentation, activity quantification requires recovery coefficient correction for partial volume effects, particularly important for small tumors and kidneys with heterogeneous uptake. Automated systems apply scanner-specific recovery coefficient tables calibrated to the institution's SPECT/CT system, eliminating a manual lookup step that is inconsistently performed in practice. Time-activity curve fitting uses Levenberg-Marquardt or similar nonlinear least-squares algorithms to fit mono-exponential or bi-exponential curves to the 3-4 time point data, with automated outlier detection for imaging data points that are inconsistent with physiological constraints.

The output is organ-level absorbed dose in gray for each time point combination: kidneys, bone marrow (estimated from blood samples or vertebral body VOIs), liver, spleen, and each segmented tumor. This calculation, which requires 4-8 hours manually, takes under 4 minutes in automated systems after image reconstruction is complete.

Kidney Dosimetry: The Critical Constraint

Renal toxicity is the dose-limiting consideration for Lu-177 PSMA and Lu-177 DOTATATE therapy. The accepted kidney dose constraint for Lu-177 RPT is 23 Gy per treatment cycle (cumulative threshold ~40 Gy), derived from external beam radiotherapy kidney tolerance data and adapted for the different dose rate of radiopharmaceutical delivery. However, this threshold assumes uniformly distributed dose within the kidney - an assumption that is frequently violated when renal cortical uptake is heterogeneous.

Automated voxel-level dosimetry tools, which calculate dose at individual voxel scale rather than organ-average, provide more accurate kidney dose estimates in patients with focal cortical retention of Lu-177 PSMA. In a comparison study from Memorial Sloan Kettering, voxel-based dosimetry identified patients at elevated nephrotoxicity risk that organ-average dosimetry classified as safe - a discrepancy that has direct implications for long-term renal function in patients expected to receive multiple treatment cycles.

Bone Marrow Dosimetry and Hematological Safety

Bone marrow is the dose-limiting organ for several RPT agents, including I-131 used in thyroid cancer treatment. Absorbed dose to active bone marrow (red marrow) is difficult to measure directly because bone marrow is distributed throughout the skeleton and inaccessible to direct volume measurement. Two approaches are used in clinical practice: blood-based dosimetry, which estimates marrow dose from blood sample activity measurements using the MIRD pamphlet methodology, and image-based dosimetry, which uses VOI measurements in lumbar vertebrae as a surrogate for marrow activity.

Automated blood-based dosimetry integrates with laboratory information systems to retrieve blood sample activity measurements at scheduled time points, applies decay correction, fits the time-activity curve, and calculates marrow dose using published S-values. For I-131 thyroid ablation, this workflow allows individualization of ablation activity to achieve a target tissue dose while keeping marrow dose below 2 Gy - an approach that is standard at major thyroid cancer centers but underused at community hospitals due to the calculation burden.

The Argument for Prospective Dosimetry

The clinical benefit of individualized dosimetry is most compelling when used prospectively to guide the second and subsequent cycles of multi-cycle therapy, rather than retrospectively to assess completed treatment. A patient who receives cycle 1 at standard 7.4 GBq and has dosimetry calculated from the cycle 1 imaging data can have cycle 2 activity adjusted based on their actual pharmacokinetics - increasing activity if organ doses were well below limits, or reducing if they were near threshold. This is the precision oncology model applied to RPT.

The QUANTEC data for kidney tolerance and the emerging evidence from the DOSISPHERE-01 trial for Y-90 radioembolization both support dose-escalated treatment yielding improved tumor response when dosimetry confirms safety. The operational barrier to implementing this at scale has been the calculation time. With automated dosimetry completing in under 4 minutes post-reconstruction, the decision to adjust cycle 2 activity can be made at the same day as cycle 1 follow-up imaging - changing the treatment planning timeline from days to hours.

Regulatory and Reimbursement Context

AAPM Task Group 186 provides the current technical standard for quantitative SPECT dosimetry used in RPT dose calculation. Clinical dosimetry reports must include uncertainty estimates for each organ dose, accounting for uncertainties in SPECT quantification, segmentation accuracy, and time-activity curve fitting. Automated dosimetry software must propagate these uncertainties through the calculation chain and present them in the output report for the supervising medical physicist to review.

CMS reimbursement for dosimetry in RPT is currently limited, creating a financial disincentive for widespread adoption. Several specialty societies including SNMMI are working with CMS on dosimetry code development. In the meantime, academic centers with active RPT programs are absorbing dosimetry costs as part of clinical development, and the cost-per-patient decreases substantially when the calculation is automated rather than requiring 4-8 hours of physics time.