Plan quality ‐ RayStation

Transcription

Plan quality ‐ RayStation
2013‐11‐28
A deliverability comparison of equivalent dual‐
arc VMAT plans generated in two different TPSs
KRISTOFFER PETERSSON, MEDICINSK STRÅLNINGSFYSIK, LUND Plan quality ‐ RayStation
• Pareto fronts
1
2013‐11‐28
Pareto front evaluation‐ Method
Better
Objective OAR
TPS 1
TPS 2
Worse
Better
Objective Target
Worse
Dmean OAR
Area under the plan (AUP)
VPTV, <95%
2
2013‐11‐28
Area under the plan (AUP)
Dmean OAR
Δiy
• AUPi =∆ ∙ ∆
Δix
VPTV, <95%
Area under the plan (AUP)
Δiy
Dmean OAR
• AUPi =∆ ∙ ∆
Δix
VPTV, <95%
3
2013‐11‐28
Area under the plans (AUPs)
∆
∙∆
∆
∙∆
.. ∆
∙∆
Δny
Δ2y
Δ1y
Dmean OAR
• AUPs = ∑ AUPi = ∑
Δ1x
Δ2x
VPTV, <95%
Δnx
35
Eclipse
RayStation
25
15
0.0
25
Eclipse
RayStation
20
15
10
70
1.0
2.0
VPTV, D<95% (%)
3.0
4.0
60
Eclipse
55
RayStation
50
45
40
0.0
1.0
2.0
VPTV, D<95% (%)
3.0
4.0
35
25
0.0
1.0
2.0
VPTV, D<95% (%)
Eclipse
RayStation Wilcoxon
Eclipse
RayStation
Wilcoxon
Eclipse
H&N 1
76.1
89.2
p=0.36
144
110
p=0.04
RayStation
H&N 2
69.0
128
p<0.01
199
191
p=0.84
H&N 3
169
177
p=0.88
226
213
p=0.88
15
3.0
H&N 2 in RayStation
30
25
Eclipse
20
RayStation
15
10
0.0
70
H&N 3 in Eclipse
65
AUPs
H&N 1 in RayStation
45
35
H&N 2 in Eclipse
30
0.0
Dmean Parotis dx (Gy)
3.0
Dmean Parotis dx (Gy)
Dmean Parotis dx (Gy)
35
1.0
2.0
VPTV, D<95% (%)
Dmean Parotis dx (Gy)
H&N 1 in Eclipse
45
Dmean Parotis dx (Gy)
Dmean Parotis dx (Gy)
Pareto fronts
1.0
2.0
VPTV, D<95% (%)
3.0
4.0
H&N 3 in RayStation
65
60
55
Eclipse
50
RayStation
45
40
0.0
1.0
2.0
VPTV, D<95% (%)
3.0
4.0
4
2013‐11‐28
Pareto fronts
Eclipse
RayStation
25
15
0.0
30
25
Eclipse
RayStation
20
15
10
70
1.0
2.0
VPTV, D<95% (%)
3.0
4.0
60
Eclipse
55
RayStation
50
45
40
0.0
1.0
2.0
VPTV, D<95% (%)
3.0
35
Eclipse
RayStation
Gold
25
15
0.0
1.0
2.0
VPTV, D<95% (%)
3.0
H&N 2 in RayStation
30
25
Eclipse
20
RayStation
Gold
15
10
0.0
70
H&N 3 in Eclipse
65
H&N 1 in RayStation
45
35
H&N 2 in Eclipse
0.0
Dmean Parotis dx (Gy)
3.0
Dmean Parotis dx (Gy)
Dmean Parotis dx (Gy)
35
1.0
2.0
VPTV, D<95% (%)
Dmean Parotis dx (Gy)
Dmean Parotis dx (Gy)
35
Dmean Parotis dx (Gy)
H&N 1 in Eclipse
45
1.0
2.0
VPTV, D<95% (%)
3.0
4.0
H&N 3 in RayStation
65
60
Eclipse
55
RayStation
50
Gold
45
40
0.0
4.0
1.0
2.0
VPTV, D<95% (%)
3.0
4.0
Pareto fronts
Eclipse
10
RayStation
5
0
1.0
2.0
3.0
VPTV, D<95% (%)
4.0
5.0
Dmean Ren sin (Gy)
15
Eclipse
30
RayStation
25
20
0.0
2.0
4.0
VPTV, D<95% (%)
35
30
Eclipse
RayStation
25
20
15
0.0
2.0
4.0
VPTV, D<95% (%)
RayStation
5
0
Eclipse
6.0
1.0
2.0
3.0
VPTV, D<95% (%)
4.0
RayStation Wilcoxon
Abdomen 1
77.1
45.6
Abdomen 2
46.0
80.9
Abdomen 3
256
261
Eclipse
RayStation
Wilcoxon
p=0.44
118
71.8
p<0.01
p=0.05
83.7
87.4
p=0.84
p=0.38
254
260
p=0.21
5.0
Abdomen 2 in RayStation
35
Eclipse
30
RayStation
25
20
0.0
6.0
Abdomen 3 in Eclipse
40
Eclipse
10
40
Dmean Ren sin (Gy)
35
15
0.0
Abdomen 2 in Eclipse
40
Dmean Ren sin (Gy)
Abdomen 1 in RayStation
20
0.0
Dmean Ren sin (Gy)
25
Abdomen 1 in Eclipse
20
2.0
4.0
VPTV, D<95% (%)
6.0
Abdomen 3 in RayStation
40
Dmean Ren sin (Gy)
Dmean Ren sin (Gy)
25
AUPs
35
30
Eclipse
RayStation
25
20
15
0.0
2.0
4.0
VPTV, D<95% (%)
6.0
5
2013‐11‐28
Pareto fronts
Eclipse
10
RayStation
5
0
1.0
2.0
3.0
VPTV, D<95% (%)
4.0
5.0
Dmean Ren sin (Gy)
15
Eclipse
30
RayStation
25
20
0.0
2.0
4.0
VPTV, D<95% (%)
35
30
Eclipse
RayStation
25
20
15
0.0
5
0
2.0
4.0
VPTV, D<95% (%)
1.0
2.0
3.0
VPTV, D<95% (%)
4.0
5.0
Abdomen 2 in RayStation
35
Eclipse
RayStation
Gold
30
25
20
0.0
6.0
Abdomen 3 in Eclipse
40
Eclipse
RayStation
Gold
10
0.0
Dmean Ren sin (Gy)
35
15
40
Abdomen 2 in Eclipse
40
Dmean Ren sin (Gy)
Abdomen 1 in RayStation
20
0.0
Dmean Ren sin (Gy)
25
Abdomen 1 in Eclipse
20
2.0
4.0
VPTV, D<95% (%)
6.0
Abdomen 3 in RayStation
40
Dmean Ren sin (Gy)
Dmean Ren sin (Gy)
25
35
Eclipse
RayStation
Gold
30
25
20
15
6.0
0.0
2.0
4.0
VPTV, D<95% (%)
6.0
Pareto fronts
25
Intracranial 1 in Eclipse
20
Eclipse
RayStation
15
10
0.0
0.5
1.0
1.5
Dmean Hemisphere dx(Gy)
Dmean Hemisphere dx(Gy)
25
Intracranial 1 in RayStation
20
RayStation
10
0.0
0.5
5
0
Dmean Hippocampus sin(Gy)
1.0
2.0
VPTV, D<95% (%)
3.0
DmeanCochlea dx (Gy)
Eclipse
RayStation
10
25
Eclipse
RayStation
20
15
10
1.0
2.0
VPTV, D<95% (%)
RayStation
p=0.21
25.3
18.8
p=0.26
p<0.01
164
114
p<0.01
123
142
p=0.71
10.7
14.3
Intracranial 2
16.4
104
Intracranial 3
15.0
78.1
p<0.01
Wilcoxon
1.5
3.0
4.0
15
Eclipse
10
RayStation
5
0
0.0
30
Intracranial 3 in Eclipse
0.0
Eclipse
Intracranial 2 in RayStation
20
Dmean Hippocampus sin(Gy)
Dmean Cochlea dx (Gy)
Intracranial 2 in Eclipse
30
1.0
RayStation Wilcoxon
VPTV, D<95% (%)
15
0.0
Eclipse
Intracranial 1
Eclipse
15
VPTV, D<95% (%)
20
AUPs
1.0
2.0
VPTV, D<95% (%)
3.0
Intracranial 3 in RayStation
25
Eclipse
20
RayStation
15
10
0.0
1.0
2.0
VPTV, D<95% (%)
3.0
4.0
6
2013‐11‐28
Pareto fronts
25
Intracranial 1 in Eclipse
20
Eclipse
RayStation
15
10
0.0
0.5
1.0
1.5
Dmean Hemisphere dx(Gy)
Dmean Hemisphere dx(Gy)
25
Intracranial 1 in RayStation
20
Eclipse
RayStation
Gold
15
10
0.0
0.5
1.0
VPTV, D<95% (%)
VPTV, D<95% (%)
Intracranial 2 in Eclipse
Eclipse
RayStation
10
5
0
0.0
Dmean Hippocampus sin(Gy)
30
1.0
2.0
VPTV, D<95% (%)
3.0
25
Eclipse
RayStation
20
15
10
0.0
1.0
2.0
VPTV, D<95% (%)
3.0
4.0
15
Eclipse
RayStation
Gold
10
5
0
0.0
30
Intracranial 3 in Eclipse
1.5
Intracranial 2 in RayStation
20
DmeanCochlea dx (Gy)
15
Dmean Hippocampus sin(Gy)
Dmean Cochlea dx (Gy)
20
1.0
2.0
VPTV, D<95% (%)
3.0
Intracranial 3 in RayStation
25
Eclipse
RayStation
Gold
20
15
10
0.0
1.0
2.0
VPTV, D<95% (%)
3.0
4.0
Pareto fronts
Eclipse
RayStation
34
32
0.0
Dmean Small Intestine(Gy)
1.0
2.0
VPTV, D<95% (%)
3.0
4.0
43
42
Eclipse
RayStation
41
40
39
0.0
42
1.0
2.0
VPTV, D<95% (%)
3.0
4.0
Pelvis 3 in Eclipse
40
38
Eclipse
RayStation
36
34
32
0.0
1.0
2.0
VPTV, D<95% (%)
3.0
Pelvis 1 in RayStation
4.0
Eclipse
RayStation Wilcoxon
Eclipse
RayStation
Wilcoxon
67.8
63.2
p=0.47
38
Eclipse
36
RayStation
Pelvis 1
78.3
Pelvis 2
21.3
37.9
p=0.02
26.6
35.2
p=0.04
Pelvis 3
37.0
68.2
p=0.03
85.4
65.2
P<0.01
34
32
0.0
44
Pelvis 2 in Eclipse
Dmean Small Intestine(Gy)
36
Dmean Small Intestine(Gy)
38
44
Dmean Small Intestine(Gy)
40
Pelvis 1 in Eclipse
1.0
2.0
VPTV, D<95% (%)
3.0
4.0
70.8
p=0.41
Pelvis 2 in RayStation
43
42
Eclipse
RayStation
41
40
39
0.0
42
Dmean Small Intestine(Gy)
Dmean Small Intestine(Gy)
40
AUPs
1.0
2.0
VPTV, D<95% (%)
3.0
4.0
Pelvis 3 in RayStation
40
38
Eclipse
RayStation
36
34
32
0.0
1.0
2.0
VPTV, D<95% (%)
3.0
4.0
7
2013‐11‐28
Pareto fronts
RayStation
34
32
Dmean Small Intestine(Gy)
1.0
2.0
VPTV, D<95% (%)
3.0
4.0
Pelvis 2 in Eclipse
43
42
Eclipse
RayStation
41
40
39
0.0
42
1.0
2.0
VPTV, D<95% (%)
3.0
4.0
Eclipse
RayStation
34
32
0.0
1.0
2.0
VPTV, D<95% (%)
3.0
32
30
0.0
1.0 2.0 3.0 4.0
VPTV, D<95% (%)
4.0
5.0
6.0
Pelvis 2 in RayStation
43
42
Eclipse
RayStation
Gold
41
40
39
0.0
Pelvis 3 in Eclipse
36
Eclipse
RayStation
Gold
34
1.0
2.0
VPTV, D<95% (%)
42
40
38
36
44
Dmean Small Intestine(Gy)
0.0
Pelvis 1 in RayStation
38
Dmean Small Intestine(Gy)
Eclipse
36
44
Dmean Small Intestine(Gy)
40
Pelvis 1 in Eclipse
38
Dmean Small Intestine(Gy)
Dmean Small Intestine(Gy)
40
3.0
4.0
Pelvis 3 in RayStation
40
38
Eclipse
RayStation
Gold
36
34
32
30
0.0
1.0
2.0
VPTV, D<95% (%)
3.0
4.0
Plan quality ‐ RayStation
• Conclusions: Significant difference in plan quality
• In Eclipse
• RayStation fronts superior for:
• 1 Intracranial
• Eclipse fronts superior for:
•
•
•
•
1 H&N, 1 Abdomen, 1 Intracranial, 2 Pelvis
• In RayStation
• RayStation fronts superior for:
•
•
•
•
1 H&N 1 Abdomen, 1 Intracranial, 1 Pelvis
• Eclipse fronts superior for:
• 1 Pelvis
8
2013‐11‐28
Inledning ‐ Bakgrund
RaySearch Laboratories
• SharePlanTM
• Automatisk konvertering av tomoterapi‐dosplaner 
step‐and‐shoot IMRT planer (levereras på vanlig linac).
• Välfungerande men begränsat användningsområde.
Inledning ‐ Bakgrund
RaySearch Laboratories
• RayStation
Modul för “Fallback planning” • Vidareutveckling av SharePlan funktionen
• Automatisk konvertering av alla typer av dosplaner 
alla modaliteter som finns tillgängliga på en vanlig linac.
•
•
•
•
3D conformal radiation therapy (CRT) Step‐and‐shoot IMRT Sliding window IMRT VMAT
9
2013‐11‐28
Syfte
Kontrollera levererbarheten
• automatiskt genererade backup‐
planer skapade i RayStation.
• Kontrollera # MU i planerna
• Jämfört med motsvarande planer som skapats i Eclipse.
• Dual‐arc VMAT planer
• Kontrollera mot levererbarhet
• Jämfört med motsvarande planer som skapats i Eclipse.
• Korrelation?
10
2013‐11‐28
Metodik
152 st planer skapas i Eclipse
Automatiskt genererade backup‐planer skapas i RayStation
• 6MV
• Baseras på Eclipseplaner
• Dual‐arc VMAT
• 3 backup planer per Eclipseplan
• Pareto‐optimala
•
1 Pareto‐front per patientfall
• 12 Patient fall
•
•
•
•
3 H&N
3 hjärntumörer
3 buktumörer
3 tumörer i bäckenet
• Olika Target vs. OAR viktfaktorer
• 1000:1, 100:1, 10:1
• Samma inställningar/restriktioner
• Ex: •
•
•
•
Kollimatorvinkel: 30°
gantry kontrollpunktseparation: 2°
Beräkningsgrid: 2.5 x 2.5 x 2.5 mm3
Etc.
• 1 backup‐plan per Eclipseplan sparas
• Bäst plankvalité
Metodik
• 304 planer (152 Eclipse +152 RayStation)
• Levererade med en TrueBeam linac (Varian)
• Mätta med Delta4 system (ScandiDos AB)
11
2013‐11‐28
ΔDm
δ
y
Γ(rm,rc)
Dc,rc
δ(rm,rc)
rc
Δdm
rc-rm
Dm,rm
x
Metodik
Gamma analys
Krav:
• Dosdifferens 3%
• DTA 2 mm
Tröskelvärde 15%
Kliniskt acceptabel plan:
≥ 95 % godkända mätpunkter
Statistisk metod:
Wilcoxon signed‐rank test
12
2013‐11‐28
Resultat
Gamma‐analys
Andel godkända mätpunkter • Signifikant högre för RayStation planer jmf. Eclipse
• Wilcoxon, (p<0.001)
• 93.0%≤ Eclipse ≤ 100%, 98.6%≤ RayStation ≤ 100%
• 66/152 RayStation > Eclipse
7/152 RayStation < Eclipse
• Enbart 3 planer < 95%, ej kliniskt acceptabla.
# MU i planerna
• Signifikant fler MU i Eclipse planer jmf. motsvarande RayStation plan
• Wilcoxon, (p<0.001)
• I medel: 29%
# MU korrelerar med grad av levererbarhet (andel godkända mätpunkter)
Pearson product‐moment correlation method r = ‐0.77
p<0.001
Korrelation mellan # MU och grad av levererbarhet
Eclipse
Andel godkända mätpunkter[%]
100
RayStation
99
98
97
Pearson korrelation
r = ‐0.77
p < 0.001
96
95
94
93
92
0
500
1000
MU
1500
2000
13
2013‐11‐28
Två patientfall med lägre värden
Andel godkända mätpunkter[%]
99
98
97
96
95
94
93
r = ‐0.79
p < 0.001
92
Eclipse H&N 2
100
Andel godkända mätpunkter [%]
Eclipse
Abdomen 2
RayStation
Abdomen 2
100
RayStation H&N 2
99
98
97
96
95
94
r = ‐0.89
p < 0.001
93
92
500
600
700
800
MU
Eclipse Arc 1 av 2
900
1000
0
500
1000
MU
1500
2000
RayStation Arc 1 av 2
14
2013‐11‐28
Eclipse Arc 1 av 2
RayStation Arc 1 av 2
Diskussion – Plankvalité
Plankvalité har kontrollerats
• Pareto‐front utvärdering
• Jämförbar mellan systemen
• Signifikant skillnad för 5/12 fall
• RayStation‐front bättre för:
•
•
•
•
1 H&N 1 Buktumör 1 hjärntumör 1 tumör i bäckenet
• Eclipse‐front bättre för :
• 1 tumör i bäckenet
15
2013‐11‐28
Slutsats
• Signifikant högre levererbarhet för RayStation planer vs. Eclipse planer.
• Antal MU signifikant högre för Eclipse planer vs. RayStation planer (29% i medel).
• OBS! • Antalet MU i en plan korrelerade signifikant med dess grad av levererbarhet.
•
Enbart 3/304 planer var ej kliniskt acceptabla(< 95%) RayStation Manus
• Differences in plan quality for available treatment techniques:
• VMAT, • IMRT (SW and SS), • 3DCRT (non‐coplanar)
• Evaluation methods
• Objective value, • CGA
• Approach
• “Gold standard” plans created in TomoTherapy planning system the 12 cases
• Plans with the highest plan quality clinically available:
• Field width: 1.05 cm
• Pitch: 0.172
• Modulation factor: 5 • Backup plans created and evaluated in RayStation
16
2013‐11‐28
Objective values
After Backup generation
Optimized # iterations
1.00E+01
1.00E+01
1.00E+00
1.00E+00
1.00E‐01
1.00E‐01
1.00E‐02
1.00E‐02
1.00E‐03
1.00E‐03
1.00E‐04
1.00E‐04
1
2
3
4
3DCRT
5
SS‐IMRT
6
7
SW‐IMRT
8
9
10
11
1
12
2
3
4
3DCRT
Dual‐Arc
5
6
SS‐IMRT
7
8
SW‐IMRT
9
10
11
12
Dual‐Arc
After Backup generation
Optimized # iterations
1.00E+01
1.00E+01
1.00E+00
1.00E+00
1.00E‐01
1.00E‐01
1.00E‐02
1.00E‐02
1.00E‐03
1.00E‐03
1.00E‐04
1.00E‐04
1
2
3
3DCRT
4
5
SS‐IMRT
6
7
SW‐IMRT
8
9
10
11
1
12
2
3
4
3DCRT
Dual‐Arc
5
6
SS‐IMRT
7
8
SW‐IMRT
9
10
11
12
Dual‐Arc
• After optimization (approximate
dose):
• No significant difference
between techniques
directly after Backup gen.
• If # iterations is optimized:
• VMAT quality
improves significantly
• VMAT significantly
superior to other
techniques
• SS‐IMRT significantly
superior to 3DCRT
• After final dose:
• SS‐IMRT significantly
superior to SW‐IMRT
• If # iterations is optimized:
• VMAT quality
improves significantly
• VMAT significantly
superior to other
techniques
Objective values
• After optimization (approximate
dose):
• If ”Reduce OAR” is used:
• Plan quality improves
significantly for all available techniques
• SS‐IMRT significantly
superior to VMAT
Direkt efter Backup‐skapandet
1.00E+01
1.00E+00
1.00E+00
1.00E‐01
1.00E‐01
1.00E‐02
1.00E‐02
1.00E‐03
1.00E‐03
1.00E‐04
1.00E‐04
1
2
3
3DCRT
• After final dose:
• If ”Reduce OAR” is used:
• No significant
improvement
• No significant
difference between
techniques
Reduce OAR
1.00E+01
4
5
SS‐IMRT
6
7
8
SW‐IMRT
9
10
11
12
1
2
Dual‐Arc
3
Direkt efter Backup‐skapandet
1.00E+01
1.00E+00
1.00E+00
1.00E‐01
1.00E‐01
1.00E‐02
1.00E‐02
1.00E‐03
1.00E‐03
1.00E‐04
1.00E‐04
2
3
3DCRT
4
5
SS‐IMRT
6
7
SW‐IMRT
8
9
5
6
7
SS‐IMRT
8
9
10
11
12
9
10
11
12
Dual‐Arc
Reduce OAR
1.00E+01
1
4
3DCRT
10
Dual‐Arc
11
12
1
2
3
4
3DCRT
5
6
SS‐IMRT
7
8
Dual‐Arc
17
2013‐11‐28
CGA‐ Method
• Inspired by VGA‐ Visual Grading Analysis
• Determine image quality (Radiology)
• Dual‐arc VMAT plans, 7‐beam SS‐IMRT, 7‐beam 3DCRT (non‐coplanar) plans
• Demonstrated for ROs (individually)
• Plans side‐by‐side
• Mimic ordinary clinical rounds
• Dose distributions
• DVHs
• ROI data
• Clinical assessment
• Identify clinical relevant differences between plans
CGA‐ Method
• ROs grade plans and motivate their opinion
•A
SS‐IMRT/3DCRT plan considerably better than the VMAT plan
•B
SS‐IMRT/3DCRT plan somewhat better than the VMAT plan
•C
SS‐IMRT/3DCRT plan as good as the VMAT plan
•D
VMAT plan somewhat better than the SS‐IMRT/3DCRT plan
•E
VMAT plan considerably better than the SS‐IMRT/3DCRT plan
18
2013‐11‐28
CGA‐ Results (so far!)
Optimized # iterations
1.00E+01
1.00E+00
1.00E‐01
1.00E‐02
1.00E‐03
1.00E‐04
1
2
3
4
5
3DCRT
6
SS‐IMRT
7
8
9
10
11
12
Dual‐Arc
CGA‐ Results (so far!)
• Hot‐spots are not penalized enough in the optimization
• Problem for complex cases
• Problem for IMRT techniques 19
2013‐11‐28
20