The aim of this study was to evaluate a commercial dosimetry workstation by comparing its performance with that of our current dosimetry solution, first using a body phantom and then in patients undergoing 177Lu-DOTATATE treatment.
First, due to the uncertainties associated with planar imaging, as described by Garkavij et al. (39), we decided to implement a 3D SPECT/CT dosimetry protocol. Indeed, DTK allows various imaging scenarios (e.g. whole body and SPECT/CT scans) (33), but limited to five time points. Conversely, PLANET®Dose accepts unlimited time points, but only SPECT/CT acquisitions. In both cases, the reconstruction step was performed with the GE application “Preparation for Dosimetry Toolkit” because PLANET®Dose does not include this functionality. However, it accepts reconstructed data supplied by others workstations, unlike the “Dosimetry Toolkit” application. The used reconstruction parameters were determined in a previous study (31).
In a clinical dosimetry study, the preliminary step of calibration is crucial to obtain accurate activity quantification (40, 41). According to the manufacturer’s recommendations, a 2D CF is preconized for DTK, whereas the calibration procedure is left to the physicist’s discretion in PLANET®Dose. In the phantom-based study, the same 3D methodology was followed to obtain the CF for each software package. As described by Gustafsson et al.(42), SPECT segmentation is essential to determine the activity concentration. Different segmentation methods are proposed: manual, automatic or semi-automatic (from the technically very easy, such as fixed threshold, to the most complicated, such as the Fourier surface method). For our phantom-based study, we selected a fixed volume threshold method based on an automatically drawn isocontour around the bottle with a volume of 200 mL. This volume was segmented on the SPECT images acquired at the first time point and rigidly propagated to the others. This step is directly affected by the accuracy of the rigid registration of all SPECT/CT images (43).
CFs are expressed in counts.s− 1.MBq− 1 by DTK and in Bq.count− 1 by PLANET®Dose. This implies that the CF must be modified with the acquisition duration for PLANET®Dose. This option is available in PLANET®Dose in which different CFs can be entered at different time points. In the literature, with gamma cameras similar to ours (40), the CF in counts.s− 1.MBq− 1 is usually two times higher than the one we obtained. This observation is explained by the use of the application “Preparation for Dosimetry Toolkit”. The crucial recommendation at this step is that the conditions used in clinical studies in terms of acquisition and reconstruction parameters have to be similar to those used for calibration.
On the other hand, we showed a negligible variation of CF values over time, from T = 0 to 216 h. This means that the same CF can be used for each time point. This observation is particularly interesting for DTK in which a single CF must be entered. With the body phantom, the mono-exponential function provided by each software package to fit time activity curves showed a slight difference especially at the Y-intercept. Indeed, in PLANET®Dose, the mean CF used for each time point was 67 Bq.count− 1 and 63.4 Bq.count− 1 for the first time point, which could imply an activity overestimation at this point. This could partly explain the difference of residence time obtained between platforms. Nevertheless, differences from the theoretical residence time below 6% are acceptable.
Concerning the absorbed dose estimation on phantom, OLINDA/EXM® V1.0 calculated the mean values by considering an “organ” of 200 g (density of 1 g.cm− 3). With PLANET→Dose, tissue density heterogeneities can be corrected using a method similar to that described by Dieudonne et al. (30). The DK method with density correction showed results close to those obtained with OLINDA/EXM® V1.0. The values calculated by OLINDA/EXM indicate that more than 98% of the absorbed dose comes from short-range beta particles of 177Lu. Sandström et al. (44) argued that the absorbed dose for kidneys is mainly due to the self-absorbed dose because only a minor proportion originates from the cross-absorbed dose due to gamma radiation from the surrounding organs. Our results showed a very small difference, below 2.5%, between LDM and DK for the phantom-based and the patient dosimetry evaluations. Therefore, as shown by Pasciak et al. (26) for selective internal radiotherapy with 90Y, the LDM method is suitable also for 177Lu. Surprisingly, the values obtained with LDM, which considers only the kinetic energy from beta particles, were higher than those obtained with the DK approach that considers both beta particle and photon energy deposition. This could be explained by the LDM principle that takes into account all the energy from beta particles deposited in the voxel, but excludes any spill-out to neighbours, a phenomenon that the DK approach models exactly. This could generate an overestimation of the absorbed dose calculation in the voxel.
In the patient study, the masses obtained with the two packages for liver were more similar than those obtained for smaller organs, such as kidneys and spleen, and the variability increased with the OAR decreasing size. The residence time supplied by PLANET→Dose was generally higher than that calculated by DTK, as shown with the phantom-based study. The Bland-Altman plot showed that PLANETⓇDose (LDM with density correction) slightly underestimated the absorbed dose compared with DTK + OLINDA, particularly for kidneys and spleen. The comparison of the dosimetric outcomes calculated by PlanetⓇDose and by OLINDA/EXM® V1.0, using the same residence times provided by PLANETⓇDose, showed that the calculation method has only a minor influence compared with all the previous steps (i.e. registration and segmentation). For accurate registration, position reproducibility during the four SPECT/CT acquisitions is crucial. Thus, patient set-up and immobilization devices are strongly recommended.
Our comparison methodology, using similar parameters in terms of registration, segmentation with constant volumes over time and mono-exponential function to fit the time activity curves, has some limitations. Indeed, using DTK, the segmented volume was maintained, but adjusted by translation or rotation at each time point when necessary. However, this step is not available in PLANET→Dose. To be as close as possible to the DTK approach, in PLANETⓇDose we chose to perform an organ-based registration followed by rigid propagation. Nevertheless, for some patients organ segmentation did not fully match at the other time points. Thus, the delineation was moved to organs with different density. This implies an important deviation when calculating the absorbed doses with density correction. For instance, part of the spleen in Patient 31 was moved to the left lung at other time points and the absorbed dose was overestimated because of the density correction. This observation could explain the important differences obtained for some patients and the observed significant differences. We are aware of the limitations of our methodology in this work, but we considered these deviations as clinically acceptable.
The concordance evaluation (Lin’s coefficient value of 0.99) highlighted an excellent agreement between methods. Moreover, the dosimetry results obtained using PLANET→Dose (absorbed dose to liver, kidneys and spleen of 0.45 ± 0.50 Gy/GBq, 0.45 ± 0.13 Gy/GBq and 0.62 ± 0.17 Gy/GBq respectively) are in agreement with those of the literature (45). The high standard deviation obtained for liver is due to the presence of hepatic lesions in some patients. These lesions were not excluded from the liver because of the difficulties in segmenting them using DTK; Bolean operators were not available in the used version. For this reason, tumours were not analysed in this study.
As proposed by Gear et al. (46) in a practical guidance paper, the uncertainties at each step of the dosimetry analysis should be determined to express the accuracy of the dosimetry results. PLANET→Dose does not offer such tool, but time activity curve fitting can be evaluated by the given contribution (in %) of the time points to the chosen function model.
Due to the limitations of the mean organ-level absorbed dose approach, especially the impossibility of accounting for the heterogeneous activity distributions in organs, the choice between voxel-based dosimetry and the mean absorbed dose approach is currently debated (47). Grimes et al. (25) showed that the mean absorbed dose to an organ is in good agreement with the voxelized absorbed dose calculated using the Monte Carlo method. They also demonstrated that the voxel S value method gives 3D absorbed dose distributions nearly equivalent to those obtained with Monte Carlo simulations. PLANET→Dose calculates the absorbed dose at the voxel level and then rescales them at the organ level to provide the mean absorbed dose to the segmented organ. The software is not validated currently for clinical use to produce a fully 3D dosimetry, allowing the edition of absorbed dose-volume histograms and isodose curves (16, 48).
From a qualitative point of view, PLANET→Dose is a user-friendly commercial solution that proposes a wide range of tools for segmentation, and several analytic fit functions. The time necessary for a dosimetry analysis is significantly reduced. Therefore, considering the good agreement between our reference dosimetry method and PLANETⓇDose, the concordance of the dosimetry results with the literature, the added value of this software in terms of easy contouring, wide choice of time activity curve fitting models, time saving, and the fact that the observed differences were explicable and clinically acceptable, we think that the PLANET®Dose software can replace our current dosimetry package without any correction for dosimetry analysis.