With this in mind, researchers from Lund University and the National Physical Laboratory have undertaken a detailed study of the measurement uncertainties associated with such dosimetry. In particular, they examined the renal absorbed dose from 177Lu-DOTATATE, which is used to treat disseminated neuroendocrine tumours, and (as 177Lu emits both beta particles and gamma photons) can be imaged using SPECT (Phys. Med. Biol. 60 8329).

"Renal absorbed doses are currently used to tailor 177Lu-DOTATATE treatments for individual patients, and to investigate relationships between absorbed dose and toxicity," explained Lund University researcher Johan Gustafsson. "For both purposes, knowledge of the uncertainty in absorbed dose values is fundamental; we believed there was a gap in the knowledge of how different sources of uncertainty combine for these measurements."

The dosimetry process

SPECT/CT-based 177Lu-DOTATATE dosimetry involves several aspects: gamma camera calibration; acquiring CT-derived density images; SPECT imaging and reconstruction; volume-of-interest (VOI) delineation; calculation of renal absorbed-dose rate; and calculation of the absorbed dose and biologically effective dose (BED). Each of these steps could introduce uncertainty into the final reported values.

Gustafsson and colleagues used Monte Carlo analysis to investigate the propagation of uncertainties throughout this dosimetry process (the MC pipeline). To do this, they used three anthropomorphic computer phantoms coupled to a pharmacokinetic model of 177Lu-DOTATATE.

Monte Carlo-simulated SPECT images of each phantom were input into the MC pipeline, which processed the image data and estimated the renal absorbed dose and BED to the left and right kidneys. To determine the combined uncertainty in absorbed dose and BED, the researchers repeated this process 256 times, with each parameter in the pipeline varied slightly each time to reflect realistic uncertainties.

For all three phantoms, the standard deviation (SD) of the mean absorbed dose estimates was approximately 6% – about the limit of what is achievable for currently used dosimetry schemes. The uncertainty in BED was similar.

To examine the impact of individual sources of uncertainty, the researchers repeated the calculations (for the left kidney of phantom 1) with the uncertainty sources fixed one by one and observed the decrease in SD when removing each source. The reduced models excluded variability in: the recovery coefficient (RC, used in correction of partial volume effects) only; and the RC plus gamma-camera sensitivity, VOI delineation, 177Lu radionuclide data, density map generation, imaging time point and noise in SPECT projections.

They found that variability in RC was a major contributor to the overall uncertainty: fixing this value reduced the SD from 0.21 to 0.11 Gy. Applying an RC of unity (as opposed to 0.84 used in the reference calculations) resulted in underestimation of absorbed dose by an average of 15%. Omitting the variability in gamma-camera sensitivity reduced the SD further, to 0.06 Gy.

Removing variability in the other sources of variability had minimal impact on absorbed-dose estimates, implying that RC and gamma-camera calibration were the most important sources of variability.

Physical phantom

To benchmark the simulation results, the researchers conducted an experiment using a physical anthropomorphic phantom with a liver insert filled with 0.5 MBq/g 177Lu-DOTATATE. They performed SPECT/CT seven times, between one and nine days after filling, and used the images to determine activity in the insert for each acquisition.

Cumulated activity concentrations were determined for different combinations of four data points selected from the seven acquisitions, giving 35 values in total. The experiment gave similar deviations to those obtained in the model when RC variability was excluded, with a mean deviation in cumulated activity concentration from the reference value of 1.7% (minimum and maximum of 0.6% and 2.9%).

The researchers note that as this experiment calculated cumulated activity concentration, rather than absorbed dose, parts of the MC pipeline involving calculation of absorbed-dose rate from activity are not covered. However, they suggest that the deviation value of 1.7% indicates that the method captures the most important sources of uncertainties in the radionuclide dosimetry chain.

"For us, the most important finding is the perspective of the dosimetry chain as one single system," Gustafsson told medicalphysicsweb. "We believe that this perspective gives insight into how different sources of uncertainty affect the estimation of kidney absorbed dose."

Related articles in PMB
Uncertainty propagation for SPECT/CT-based renal dosimetry in 177Lu peptide receptor radionuclide therapy
Johan Gustafsson et al Phys. Med. Biol. 60 8329
Pharmacokinetic digital phantoms for accuracy assessment of image-based dosimetry in 177Lu-DOTATATE peptide receptor radionuclide therapy
Gustav Brolin et al Phys. Med. Biol. 60 6131
Factors affecting the repeatability of gamma camera calibration for quantitative imaging applications using a sealed source
N Anizan et al Phys. Med. Biol. 60 1325
On the biologically effective dose (BED)—using convolution for calculating the effects of repair: II. Numerical considerations
Johan Gustafsson et al Phys. Med. Biol. 58 1529

Related stories

• Alpha therapy targets metastatic disease
• Simplified dosimetry for 89Zr-immuno-PET
• Radio-immunotherapy: attacking cancer from within
• Optimizing peptide receptor radiotherapy