Precision of pre-SIRT predictive dosimetry International Course on THERANOSTICS AND MOLECULAR RADIOTHERAPY Department of Nuclear Medicine – Medical Physics Jules Bordet Institute, Université Libre de Bruxelles Brussels, Belgium Hugo Levillain
Precision of pre-SIRT predictive dosimetry
International Course on THERANOSTICS AND MOLECULAR RADIOTHERAPY
Department of Nuclear Medicine – Medical Physics
Jules Bordet Institute, Université Libre de Bruxelles
Brussels, Belgium
Hugo Levillain
Imaging workflow pre-SIRT
Treatment Planning
3D CT Angiogram
Cone Beam CT (CBCT) 99mTc-MAA injection
99mTc-MAA SPECT/CT
Prevent post-therapy complication and select patient with good potential
outcome
3
1
Patient’s Selection
Diagnostic MRI, CT, PET/CT, SPECT/CT
1
Hepatobiliary scintigraphy Dynamic Planar + SPECT/CT
Patient stratification
2
Imaging workflow post-SIRT
CBCT
90Y SIR-Spheres injection
Treatment 7 to 15 days after
simulation
90Y SIR-Spheres PET/CT
Patients follow up 6-8 weeks after treatment
Follow-up MRI, CT, PET/CT, SPECT/CT
3
4
We see what We treat
2
Response
assessment
Personalized pre-SIRT predictive dosimetry
& SIRT therapeutic window
Optimize the therapeutic efficacy
Minimize the dose to organs at risk: Safety
Maximize the dose delivered to the tumor
3
Problems
Therapeutic window
To be defined according to cutoffs definitions
Which dosimetry methodology?
Personalized pre-SIRT predictive dosimetry
Which dosimetry methodology?
Predictive power?
Precision and accuracy of pre-SIRT predictive
dosimetry
4
Problems
Therapeutic window
To be defined according to cutoffs definitions
Which dosimetry methodology?
Personalized pre-SIRT predictive dosimetry
Which dosimetry methodology?
Predictive power?
Precision and accuracy of pre-SIRT predictive
dosimetry
5
SIRT therapeutic window: therapeutic
effect
6 Cremonesi et al., 2014, frontiers in Oncology
Progressive disease
Threshold for response
Response
Partial Response / Stable disease
SIRsphere
Therasphere
SIRT therapeutic window: therapeutic effect
Dose-response relationship based on 99mTc-MAA
SPECT/CT for mCRC patients
Flamen et al. 2008:
Significant response (ΔTLG>50%) for Dmean > 66 Gy
No standardization of image acquisition and reconstruction
Evolution of image modalities
7
R² = 0.26
-200-180-160-140-120-100
-80-60-40-20
020406080
100
0 20 40 60 80 100 120 140 160 180
TL
G c
han
ge
%
Simulated Absorbed Dose (Gy)
Flamen P. et al. 2008, Physics in Medicine and Biology
Problems
Therapeutic window
To be defined according to cutoffs definitions
Which dosimetry methodology?
Personalized pre-SIRT predictive dosimetry
Which dosimetry methodology?
Predictive power?
Precision and accuracy of pre-SIRT predictive
dosimetry
8
Dosimetry methodology
Medical Internal Radiation Dose (MIRD) formalism
Procedure guidelines based on MIRD model
Cumulated Activity (biological data): 3D quantification
S Factor (Physical data): Energy deposition into target volume
Organ/sub-organ level
Hypothesis: uniform distribution
9
Bolch.W. et al. 2009, Journal of nuclear medicine : official publication, Society of Nuclear Medicine
Personalized 3D dosimetry (Planet Onco 3.0, Dosisoft®)
MIRD at voxel scale
Time Integrated Activity (TIA) map at voxel scale
Voxel S values (VSV) (Dieudonne et al., 2011)
Dieudonné, A., 2011 , Journal of Nuclear Medicine
10
Dosimetry methodology
Personalized 3D dosimetry (Planet Onco 3.0, Dosisoft®)
Dose Volume Histogram (DVH)
Dieudonné, A., 2011 , Journal of Nuclear Medicine
11
90Y SIR-Spheres PET/CT 99m Tc-MAA SPECT/CT
VSV ⊗
Convert to 90Y-MS TIA map
90Y SIR-Spheres isodoses
Dosimetry methodology
Personalized 3D dosimetry: Dose Volume Histograms
Dosimetry methodology
40 Gy – 100% uniform
40 Gy – 60% Heterogeneous
40 Gy – 60% uniform
20
40
60
80
100
0 40 60 80 100
V (%)
0 40 60 80 100 0 40 60 80 100
D (Gy)
Personalized 3D dosimetry: Dose Volume Histograms
Dosimetry methodology
40 Gy – 100% uniform
40 Gy – 60% Heterogeneous
40 Gy – 60% uniform
20
40
60
80
100
0 40 60 80 100
V (%)
0 40 60 80 0 40 60 80 100
D (Gy)
100
Personalized 3D dosimetry: Dose Volume Histograms
Dosimetry methodology
40 Gy – 100% uniform
40 Gy – 60% Heterogeneous
40 Gy – 60% uniform
20
40
60
80
100
0 40 60 80 100
V (%)
0 40 60 80 100 0 40 60 80 100
D (Gy)
Jules Bordet Institute’s pre-SIRT dosimetry workflow
I) Images selection
99mTc-MAA SPECT/ CT (reference images)
Diagnostic images (FDG PET/CT, SPECT CT, IRM, CT)
Additional images: CBCT, hepatobiliary SPECT/CT
13
Dosimetry methodology
Jules Bordet Institute’s pre-SIRT dosimetry workflow
II) 99mTc-MAA SPECT/ CT Liver delineation
Manually
14
Dosimetry methodology
Errors
Jules Bordet Institute’s pre-SIRT dosimetry workflow
III) Images rigid co-registration
Automatic + visual inspection and manual corrections
15
Dosimetry methodology
Errors
Jules Bordet Institute’s pre-SIRT dosimetry workflow
IV) 99mTc-MAA treated liver delineation (e.g. right liver)
Manually
Validation with CBCT
16
Dosimetry methodology
Errors
Wahl.R. et al., 2009 , Journal of Nuclear Medicine
Jules Bordet Institute’s pre-SIRT dosimetry workflow
V)Lesions’ delineation
Fixed threshold based on PERCIST
17
Dosimetry methodology
Errors
Jules Bordet Institute’s pre-SIRT dosimetry workflow
VI) Dosimetry using partition model
1) Safety
homogenous MAA distribution
Activity giving 40 Gy to the treated liver
18
Dosimetry methodology
Prescribed activity = 987 MBq
Lesion mean absorbed dose = 288 Gy 20
40
60
80
100
0 50 100 150 200 250 300
V (%)
D (Gy)
20
40
60
80
100
Jules Bordet Institute’s pre-SIRT dosimetry workflow
VI) Dosimetry: Personalized 3D dosimetry (Planet Onco 3.0,
Dosisoft®)
2) Optimization
Dose volume histogram based
Activity giving 40% of the treated liver receive less than 40 Gy
EBRT safety criteria transformed using BED
19
Dosimetry methodology
Before optimization 35 Gy-40%
0 20 40 60 80 100 120 140 160
100
After optimization 40 Gy-40%
20
40
60
80
0 50 100 150 200
Jules Bordet Institute’s pre-SIRT dosimetry workflow
VI) Dosimetry: Personalized 3D dosimetry (Planet Onco 3.0,
Dosisoft®)
2) Validation
Lesion mean absorbed dose = 339 Gy (+50 Gy)
Prescribed activity
20
Dosimetry methodology
Simulation 99mTc-MAA isodoses on baseline FDG PET/CT
Jules Bordet Institute’s pre-SIRT dosimetry workflow
Pre-SIRT predictive dosimetry
Minimize the dose to the liver
Maximize the dose delivered to the tumor
Personalized activity prescription
21
Dosimetry methodology
Problems
Therapeutic window
To be defined according to cutoffs definitions
Which dosimetry methodology
Personalized pre-SIRT predictive dosimetry
Which dosimetry methodology
Predictive power?
Precision and accuracy of pre-SIRT predictive
dosimetry
22
Biology Vascularization
Hemodynamics
Dosimetry MAA as a surrogate of microspheres distribution?
Same catheter position?
Dosimetry methodology
Prepared vs prescribed activities?
Administered vs prescribed activities?
SIRT needs precise/accurate determination of
administered 90Y SIR-Spheres activity
Pre-SIRT dosimetry errors sources
23
Precision and accuracy of pre-SIRT
dosimetry
SIRT needs accurate/precise determination of
administered 90Y SIR-Spheres activity
Affected by Stochastic and Systematic errors
Clinical dosimetry need first Precision
24
Re
spo
nse
Si
de
eff
ect
s
Re
spo
nse
Si
de
eff
ect
s
Stochastic Precision Systematic Accuracy
Precision and accuracy of pre-SIRT
dosimetry
Dealing with accuracy
No gold-standard applicable in clinical routine
Expensive
Radionuclide calibrator and/or sources transport for calibration
4
Re
spo
nse
Si
de
eff
ect
s
Constant
Cross-validation of radionuclide
calibrator with 90Y SIR-Spheres
PET/CT’s calibration
25
Quality assurance of activity administration
in SIRT
Evaluate the difference between administered
90Y SIR-Spheres activities:
Computed 𝐴𝑝𝑟𝑒𝑝𝑎𝑟𝑒𝑑 − 𝐴𝑟𝑒𝑠𝑖𝑑𝑢𝑎𝑙
Quantified on 90Y PET/CT images
26
Retrospective study based on 73 patients
Single catheter position
No quantifiable shunt to non-target organs
Patients underwent
99mTc-MAA SPECT/CT
MAA perfusion volume (1-5% max)
Activity prescription
90Y SIR-Spheres PET/CT
27
Quality assurance of activity administration
in SIRT
Administered activity
Computed (State of the art): 𝐴𝑝𝑟𝑒𝑝𝑎𝑟𝑒𝑑 − 𝐴𝑟𝑒𝑠𝑖𝑑𝑢𝑎𝑙
Quantified on 90Y PET/CT
Corrected from nuclear decay
28
Quality assurance of activity administration
in SIRT
1
2
Computed administered activity 𝐴𝑝𝑟𝑒𝑝𝑎𝑟𝑒𝑑 − 𝐴𝑟𝑒𝑠𝑖𝑑𝑢𝑎𝑙
90Y SIR-Spheres activity to inject
Prepared in accordance to the prescription
Measured with radionuclide calibrator CRC-15R Capintec
Residual activity
Pre/Post-injection box doses’ rates measurements
29
Quality assurance of activity administration
in SIRT
1
Quantified administered activity
90Y SIR-Spheres PET/CT images
GE-Healthcare Discovery 690 PET/CT
2 bed positions of 30min, voxel size: 2.7×2.7×3.3mm3
Images’ reconstruction (QUEST phantom study)
3D OSEM, 18 iterations, 3 subsets, 13.7 Gaussian post-filtering
Corrections: Attenuation, Diffusion, TOF, PSF inhomogeneity
30
Quality assurance of activity administration
in SIRT
2
Willowson.K. et al., 2015 , European Journal of Nuclear Medicine and Molecular Imaging
Willowson.K. et al., 2015 , European Journal of Nuclear Medicine and Molecular Imaging
Carlier.T. et al., 2016 , Medical Physics
Quantified administered activity
Requires a volume
90Y SIR-Spheres PET/CT volume delineation?
No methods
Noise
Partial volume effect
Utilize the MAA volume
90Y SIR-Spheres PET/CT quantification
Predictive power of MAA for the volume
31
Quality assurance of activity administration
in SIRT
2
Quantified administered activity
Registration corrected by visual inspection
32
Quality assurance of activity administration
in SIRT
2
Quantified vs Computed administered activities
Univariate linear regression
Agreement between computed and quantified activities
Relative difference between the 2 measurements
Mean ± SD
Plot of errors as a function of treated volume size
Plot of errors as a function of the amount of injected activity
33
Quality assurance of activity administration
in SIRT
Quantified vs Computed administered activities
34
R2=0.96
Qu
anti
fied
90Y-
MS
acti
vity
(M
Bq
)
Computed 90Y-MS activity (MBq)
Quality assurance of activity administration
in SIRT
Quantified vs Computed administered activities
35
Mean relative difference=3.2±12.3%
Freq
ue
ncy
Relative difference (%)
Quality assurance of activity administration
in SIRT
Quantified vs Computed administered activities
36
Quality assurance of activity administration
in SIRT
Rel
ativ
e d
iffe
ren
ce (
%)
Rel
ativ
e d
iffe
ren
ce (
%)
Treated Volume (cm3) Administered activity (MBq)
Problems
Therapeutic window
To be defined according to cutoffs definitions
Which dosimetry methodology
Personalized pre-SIRT predictive dosimetry
Which dosimetry methodology
Predictive power
Precision and accuracy of pre-SIRT predictive
dosimetry
37
Conclusion
Relative difference between computed and
quantified administered activities is relatively low
Good correlation of the 2 measurements: R2 = 0.96,
mean relative difference = 3.2 ± 12.3%
Knowledge of the precision of the correlation
Importance of images’ registration and segmentation
No correlation with the amount of injected activity or
with treated volume size
38
Conclusion
Feasibility of a quality assurance process
90Y SIR-Spheres PET/CT can be used directly to
measure the administered activity
Demonstration of the predictive power of pre-SIRT
predictive dosimetry at liver scale
39
Perspectives
90Y-MS PET/CT and 99mTc-MAA distributions’
correlation
Treatment 90Y SIR-Spheres isodoses
On baseline FDG PET/CT
Simulation 99mTc-MAA isodoses
On baseline FDG PET/CT
40
Take home message
Therapeutic Window
Predictive dosimetry : treatment outcome
optimization
Personalized 3D Dosimetry workflow
Volume delineation and image registration
Safety: 40 Gy to the treated volume
Optimization: 40 Gy 40% based on DVH
Validation
Personalized activity prescription
41
Thank you!
International Course on THERANOSTICS AND MOLECULAR RADIOTHERAPY