Dosimetry software platforms
The characteristics of the dosimetry software platforms used in this study are presented in Table 2.
PLANET®Dose is a treatment planning system from DOSIsoft. The calibration procedure is left to the user’s discretion and a CF (in Bq.count-1) is required. This dosimetry platform does not reconstruct SPECT/CT data, but accepts reconstructed data in DICOM format from all devices. It provides multi-time point registration (rigid and elastic), organ segmentation (manual and automatic), and TIAC calculation with a wide choice of interpolation methods of the time-activity curve (linear, trapezoidal, mono-, X-, bi-, tri-exponential …). The mean AD can be calculated with or without media density correction, using either the local energy deposition (LDM) or convolution of dose voxel kernels (DK) (15,32,33) (Fig. 1b).
For this study, the segmentation, registration and TIAC steps were carried out on PLANET®Dose with restricted parameters, similar to those defined in DTK.
Dosimetry imaging protocol
All imaging acquisitions were performed with a SPECT/CT Discovery NM/CT 670 apparatus (General Electric [GE] Healthcare), including a BrightSpeed 16 CT scanner and a 3/8-inch NaI(Tl) crystal, according to the previously described acquisition protocol (29). Briefly, nuclear medicine images were acquired using a medium-energy general purpose parallel-hole collimator. A 20% energy window centered on the 208 keV photopeak and a 10% scatter correction window centered on 177 keV were applied. NM acquisitions were realized using a body contour option, rotation of 180° per detector, total of 60 projections and 45s each. For attenuation correction, CT images were acquired (120 kV, automatic mA regulation with a max at 200 mA, noise index at 6.43, slice thickness of 5 mm, rotation time of 0.8 s, pitch 1.375, 512x512 pixels matrix), with standard reconstruction.
The application “Preparation for Dosimetry Toolkit” was used for SPECT/CT image reconstruction for both dosimetry approaches. The Ordered Subset Expectation Maximization iterative reconstruction algorithm was used with 6 iterations and 10 subsets, attenuation, scatter, recovery resolution corrections and a Gaussian post-filter of 0.11 cm.
Calibration factor, Time Integrated Activity Coefficient
A NEMA IEC body phantom (Body Phantom NU2-2001/2007) containing two bottles of 250 mL filled with 200 mL of 82.2 ± 4.1 MBq [177Lu]Lu-DOTA-TATE was used. The intention was to obtain a geometry close to that of the kidneys. The background was filled with non-radioactive water. SPECT/CT images were acquired once at different time points to evaluate CF variations over time. The CF was estimated using one of the two bottles (Fig. 2).
Dosimetry Toolkit. A CT rigid registration centred on the full phantom was performed with “Preparation for Dosimetry Toolkit”. Then, using the “Dosimetry Toolkit” application, an isocontour representing a volume of 200 mL was automatically segmented on the first NM image and was replicated for the images at 24h, 72h, 120h and 216h. For each time-point, the segmented volume was kept constant, but its position was adjusted on the CT image by translation or rotation. To determine the CF in counts.s-1.MBq-1 at each time point, the number of events in the volume was divided by the acquisition time provided by the DICOM data (2700s) and by the activity. For radioactivity decay correction, a physical half-life of 6.647 days (34) was applied and the activity at each acquisition time was corrected from the phantom preparation time. The computed CF was the mean of the CFs at the different time points. To obtain the time-activity curve fitted by a mono-exponential function and the TIAC (h), information about the radionuclide and the previously calculated CF were entered in the appropriate interface.
PLANET®Dose. A CT rigid registration centred on the bottle was performed with PLANET®Dose. Similarly, an isocontour representing a volume of 200 mL was automatically segmented on the first functional image and was rigidly propagated to the other time point images. The segmented volume was maintained over time. To determine the CF in Bq.count-1 at each time point, the number of events in the volume was divided by the activity in Bq, by taking into account the radioactivity decay. The computed CF was the mean of the CFs at the different time points. A mono-exponential fitting function, similar to the DTK approach, was used to calculate the TIAC (h).
In the phantom study, as a mono-exponential activity decay occurred due to the radioisotope physical half-life, the TIA, , and the TIAC, , were estimated as follows (35):
where A0 is the injected activity, f is the bound activity fraction (1 in this case), and Teff = TR is 177Lu physical half-life (6.647 days) (34).
The theoretical τ of 230.15h was considered as the reference TIAC and was compared to the values obtained with the two dosimetry platforms.
Mean absorbed dose
The mean AD was calculated using 3 approaches presented in Fig 1:
- The GE DTK approach uses TIAC provided by DTK, entered in OLINDA/EXM® V1.0 to derive average absorbed doses to tissues/organs (Fig. 1a).
- The PLANET®Dose approach uses reconstructed images (Preparation for DTK) and full processing (registration, segmentation, TIAC and absorbed dose calculation using LDM and DK methods, with and without density correction) on PLANET®Dose (Fig. 1b).
- The third approach is similar to the former, but stops at the TIAC step to compute absorbed doses with OLINDA (Fig. 1c) for a kidney mass adjusted to 200 g.
Similarly, by knowing the theoretical TIAC of the radionuclide in the phantom, a theoretical absorbed dose was calculated using OLINDA/EXM V1.0 for a kidney mass adjusted to 200 g, and this value was compared to those obtained with the TIACs from PLANET®Dose and from DKT in OLINDA/EXM V1.0.
Patients and treatment
Twenty-one patients (5 women and 16 men; median age 68 years, range 41 to 82 years) with a neuroendocrine tumour and treated with [177Lu]Lu-DOTA-TATE, Lutathera® (Advanced Accelerator Applications, Saint Genis Pouilly, France) were evaluated (Table 3). The treatment consisted in 7.2 ± 0.2 GBq activity (four infusions in total) injected every 8 weeks. Amino acids (lysine + arginine) were administered concomitantly to ensure renal protection by reducing tubular reabsorption of the radiolabelled peptides. All patients were hospitalized in specialized radioprotection rooms for 24h after injection. Patients were then released and had to come back for further imaging sessions. OAR dosimetry (liver, kidneys and spleen) was performed after the first and second infusion of [177Lu]Lu-DOTA-TATE. Dosimetry data after cycle 1 were not evaluable in one patient and another patient died before the second infusion. In total, 40 dosimetry analyses were performed with each dosimetry platform.
The study was approved by the local ethical review board.
The dosimetry workflow for the two platforms is presented in Table 4. SPECT/CT images were acquired at 4h, 24h, 72h and 192h after infusion. For some patients, due to health problems, technical issues or calendar reasons, SPECT/CT images were acquired at only three time points after injection. As dosimetry for the first 2 cycles of the therapy is performed routinely in our department, currently with DTK+OLINDA, we performed a retrospective additional analysis with PLANET®Dose of clinical data already available.
Reference dosimetry method (Fig. 1a). For infusion 1 and 2, after the last SPECT/CT image acquisition at 192h, all SPECT/CT data were loaded on the “Preparation for Dosimetry Toolkit” application. Imaging data were reconstructed and an automatic rigid registration between CT scans was performed. The results were loaded on the “Dosimetry Toolkit” application. The OARs (liver, kidneys and spleen) were manually segmented using the CT images collected at 4h post-injection, and then rigidly propagated to the 24h, 72h, and 192h images. For each time point, the segmented volume was maintained, but sometimes it was adjusted by translation or rotation. For the OARs considered in our study the partial volume effect correction was considered negligible. Information about the administered activity, the date and time of administration, the radionuclide and the CF (in counts.s-1.MBq-1) were entered. To obtain the TIAC, the time-activity curves were fitted using a mono-exponential function, the only fitting model available in the “Dosimetry Toolkit” application. Then, the TIAC values were exported to OLINDA/EXM® V1.0 to calculate the OAR absorbed doses. The organ masses included in this software were determined from the organ volume defined on the CT images using “Dosimetry Toolkit” and the biological tissue density (1.06 g.cm-3 for liver and spleen; 1.05 g.cm-3 for kidney).
PLANET®Dose (Fig. 1b). The transversal slices reconstructed using the “Preparation for Dosimetry Toolkit” application and the corresponding CT images were uploaded on PLANET®Dose. A study was created for each organ to simulate the Dosimetry Toolkit registration method. The rigid and automatic registration was centred on the OARs and then saved. The first CT scan was manually segmented and then propagated rigidly to the others. The volume of each OAR remained constant at all time points. As mentioned above, the partial volume effect correction was considered negligible for the OARs studied.
Information about the administered activity, the date and time of administration, the radionuclide and the CF (in Bq.counts-1) was entered. The time-activity curve was fitted using a similar approach as the one used in DTK (i.e. mono-exponential function) to provide the TIAC (h) and the TIA (MBq.s). The mean absorbed doses were calculated using the LDM and DK methods, with correction of density.
Additionally to the relative difference (in %), the root mean-square deviation (RMSD) of organ masses, TIACs and absorbed doses per cycle obtained with PLANET®Dose and DTK/OLINDA (taken as reference) was calculated as follows:
Where X(i) was organ masses, TIACs or ADs obtained for the dosimetry analysis i
For the AD, the results obtained using the LDM and DK methods from PLANET®Dose, with density correction, were compared with those obtained using DTK+OLINDA/EXM V1.0.
To evaluate independently the absorbed dose calculation methods, the TIACs from PLANET®Dose were exported to OLINDA/EXM® V1.0. The organ masses included in the software were those determined from the organ volume defined on the CT images using PLANET®Dose and the biological tissue density. The mean AD obtained with PLANET®Dose (density correction) and with “PLANET®Dose + OLINDA/EXM® V1.0” (taken as reference) were compared (Fig. 1c).
The Lin’s concordance correlation coefficient (36) was used to evaluate the agreement between PLANET®Dose LDM and the reference method (DTK+OLINDA) and then, between PLANET®Dose LDM and PLANET®Dose + OLINDA. These analyses were performed using the values for all patients and organs after the two infusions. In addition, absolute differences between the ADs obtained with the two approaches with regard to the average value of the two were assessed for each organ using the Bland-Altman plot analysis (37). The 95% limits of agreement, from -1.96 to +1.96 SD, were calculated for each organ.
The paired Student’s t-test was used to compare the OAR absorbed doses calculated with DTK+OLINDA and PLANET®Dose LDM (n=40). This analyse was performed using the mean values for all patients and organs after the two infusions.