Inter-observer variability in
hippocampus delineation on MRI scans
for Hippocampal Avoidance - PCI trial
C.Chen, M.de Ruiter, F.Bartel, M.Kwint, J.Belderbos F.Vandaele, S.Sunaert, K.De Jaeger, N.Dollekamp,
E.Dieleman, R.Achten, D.de Ruysscher, S.Schagen,
N
SCLC: M
09
PCI-PCI
SCLC:
M
12
PHA-PCI
• N
SCLC: PCI versus geen PCI
• Indien PCI: HA-PCI = 2vs + leaves
• SCLC: Standaard PCI (=2vs) en HA-PCI
• Indien standaard PCI: 2vs standaard
ZONDER leaves
• Indien HA-PCI: VMAT met planning
N
SCLC- tumoren:
M09PCI
HA-PCI + Leaves 30Gy/12 fracties/4x pw
• Standaard PCI: 2vs
• Hippocampus Avoiding Prophylactic Cranial
Irradiation(HA-PCI): VMAT
Extensive disease: 20 Gy / 5 fracties / 5x pw Limited disease: 25 Gy / 10 fracties / 4x pw
The localization of the hippocampus (braun) in the brain
http://www.visiblebody.com.
Phase III trial randomizing PCI
with or without hippocampal
avoidance NCT01780675
Overview
• Motivation of the study
• Study preparation
• Results
• Discussion
• Future work
Motivation
• Sources of variations in the
Hippocampus Avoiding Prophylactic Cranial Irradiation
(HA-PCI) trial:
– Image registration (CT and MRI)
– Inter- and intra- observer delineation
– Dose planning – During treatment
• Motivation:
– The inter-observer variability in the HA-PCI trial
– The influence of delineation variability on dose planning
NCT01780675 objectives and endpoints
Primary objective:
• to reduce memory function loss with hippocampal
avoidance PCI (HA-PCI) compared to standard whole brain PCI in SCLC patients
Secondary objectives:
• to assess early and late neurotoxicity and quality of life • to assess structural and functional brain abnormalities on
MRI
• to assess the incidence and location of brain metastases PFS and OS
PCI
H
ippocampus
A
voidance
SCLC (LD and ED)
PCI
full PCI HA-PCI
•Monitoring cognitive sequelae :
•Neuro-cognitive tests 6 times:
•Baseline and 4/8/12/18/24
• Repeat MRI
•Does HA-PCI increase QoL? R
Interobserver variability study preparation
• 5 randomly selected trial patients (10 structures)
– A high resolution T1 weighted MRI scan was registered to the planning CT scan (1mm)
• 7 participating institutions
• According to the RTOG atlas
on axial slices of the MRI
Results - volume
Volume per structure
Delineation structure per observer mean volume (cm3) SD volume (cm3) Mean difficulty score 1-L* 2.53 0.30 1.7 2-L* 2.35 0.33 1.8 3-L* 2.19 0.31 2.0 4-L* 2.25 0.48 2.3 5-L* 2.61 0.41 2.5 1-R* 2.72 0.29 1.7 2-R* 2.35 0.14 1.8 3-R* 2.35 0.22 2.0 4-R* 2.53 0.41 2.3 5-R* 2.75 0.47 2.5
Results - volume
Volume per structure
Delineation structure per observer mean volume (cm3) SD volume (cm3) Mean difficulty score 1-L* 2.53 0.30 1.7 2-L* 2.35 0.33 1.8 3-L* 2.19 0.31 2.0 4-L* 2.25 0.48 2.3 5-L* 2.61 0.41 2.5 1-R* 2.72 0.29 1.7 2-R* 2.35 0.14 1.8 3-R* 2.35 0.22 2.0 4-R* 2.53 0.41 2.3 5-R* 2.75 0.47 2.5
Results - volume
Volume per structure
Delineation structure per observer mean volume (cm3) SD volume (cm3) Mean difficulty score 1-L* 2.53 0.30 1.7 2-L* 2.35 0.33 1.8 3-L* 2.19 0.31 2.0 4-L* 2.25 0.48 2.3 5-L* 2.61 0.41 2.5 1-R* 2.72 0.29 1.7 2-R* 2.35 0.14 1.8 3-R* 2.35 0.22 2.0 4-R* 2.53 0.41 2.3 5-R* 2.75 0.47 2.5
Results - volume
Volume per structure
Delineation structure mean volume (cm3) SD volume (cm3) Mean difficulty score 1-L 2.53 0.30 1.7 2-L 2.35 0.33 1.8 3-L 2.19 0.31 2.0 4-L 2.25 0.48 2.3 5-L 2.61 0.41 2.5 1-R 2.72 0.29 1.7 2-R 2.35 0.14 1.8 3-R 2.35 0.22 2.0 4-R 2.53 0.41 2.3 5-R 2.75 0.47 2.5
Patient 4 and 5 are more difficult to delineate larger variations in
Results – hippocampus volume
Volume per structure
• SD delineated volume varied from 0.14-0.48 cm3
• The overall inter-observer variability of the volume:
• Intraclass correlation coefficient: ICC=0.19
Results –overall delineation variability
Delineation structure per observer Overall delineation variability (mm) 1-L* 0.7 2-L* 1.0 3-L* 0.7 4-L* 0.9 5-L* 0.8 1-R* 0.8 2-R* 0.9 3-R* 0.6 4-R* 0.9 5-R* 1.0 Ranges from 0.6-1.0 mmroot-mean-square of the local SD per sampled points on the median surface
Results - volume
Volume per structure
Volume per observer lower volume
than other observers
Problematic regions
Problematic regions
Uncus area
Problematic regions
Problematic regions
Future work
• Comparison to the golden standard
• Improving delineations
– Adherence to RTOG guidelines
Web-based workshop on delineation – Case studies of difficult patients
• The influence of delineation variability on dose planning • Is the 5 mm margin in dose planning sufficient or too
Discussion – axial sequences
• Why do we choose to delineate on axial sequences (more difficult)?
Because it’s easy for dose planning.
• Do all institutions follow the RTOG atlas? No
• When the hippocampus is rotated (or asymmetrical), it becomes more difficult to delineate in axial sequences. • Always validate the delineations in sagittal view!
Thanks for the teamwork
Sanne Schagen Michiel de Ruiter Marianne Kuenen Willem Boogerd Pietje Muller Chun Chen Eugene Damen Harm van Tinteren Anne Lisa Wolf Margriet Kwint Emmy Lamers Casper Carbaat Katrien de Jaeger Friederike Koppe Marjan van de Pol Joachim Widder Edith DielemansDirk de Ruysscher Sabine Deprez Stefan Sunaert
Frank van den Heuvel Johan Vansteenkiste Yolande Lievens Katrien Vandecasteele Rik Achten Paul Meijders Chirs Goor Paul Parizel Matthew Holt
• NSCLC hippocampal sparing study is closed
– Results not yet available
• This also used a simpler sparing technique, with 2
lateral fields and 2 blocks.
• HVLT Hopkins Verbal Learning Test is used for
cognitive changes. End point is if patient loses 4
points on a test following HPPCI then the HP
sparing will have failed.
Pros and Cons
VMAT Practical (Pacman) technique
96.4% Coverage 85.7% Coverage
7.4Gy mean dose (α/β=2) Hippocampus 3.9Gy mean dose (α/β=2) Hippocampus
BeamModulator, 4mm leaf MLC 1cm leaf MLC
Complex plan / dosimetry Simple plan
Fairly long irradiation time Short irradiation time
Techniek
• Beams
– Kies A2 en 10 MV.
– 2 Dual Arcs loodrecht op elkaar.
• Gantry 340º -180º CW; tafel 90º; collimator 110º • Gantry 178º -182º collimator 20º
– Beam on Time 4x100sec ~ 7 minuten
Objectives
Start ( = Stap 1)
BRAIN4-HC10 Min Dose 2375 100
BRAIN4-HC10 Uniform dose 2500 70
HIPPPOCAMPUS_R/L Max Dose 800 1
LENS_R/L Max Dose 800 10
Stap 2
BRAIN4-HC10 Min Dose 2375 100
BRAIN4-HC10 Uniform dose 2500 70
HIPPPOCAMPUS_R/L Max Dose 600 1
Objectives
Stap 3
BRAIN4-HC10 Min Dose 2375 100
BRAIN4-HC10 Uniform dose 2500 70
HIPPPOCAMPUS_R/L Max Dose 600 10
LENS_R/L Max Dose 800 10
Stap 4 BRAIN4-HC10 Min Dose 2375 100
BRAIN4-HC10 Uniform dose 2500 70
HIPPPOCAMPUS_R/L Max Dose 600 10
LENS_R/L Max Dose 800 10
Objectives
Evt Stap 5 zelf toevoegen
BRAIN4-HC10 Min Dose 2375 100
BRAIN4-HC10 Uniform dose 2500 70
HIPPPOCAMPUS_R/L Max Dose 600 10
LENS_R/L Max Dose 800 10
EYE_R/L Max DVH 1500 30 20
Criteria
• PTV:
- V23.75Gy > 95%
- V28.75Gy < 10% ± 8%
(bij voorkeur D1% <115%)
• Hippocampi:
- Dmean < 8.5 Gy ± 0.5Gy
- D1 < 10 Gy ± 1 Gy
• Lenzen:
- Dmax < 10 Gy