Patient positional correction strategies during radiotherapy delivery

1. Biplab Sarkar;Ph.D.
Fortis Memorial Research Institute,
Gurugram
Correction Strategies

2. What is to be corrected: Setup Error
Setup Error: Spatial (Translational & Rotational) error
between desired and actual therapy delivery position.
 Gross Error: Unacceptably large Positional Error 
can potentially underdoes PTV or Overdose OAR.
Part -I
Non-Image Based Gross positional
Error Correction Strategies.


3. How to identify & Correct Grass Positional Error.
 What is Gross Error: Anything beyond applied CTV to
PTV margin.
Gross error attributed to
(1) Incorrect patient, anatomical site or patient orientation
(2) Incorrect Field size, Shape and Orientation
(3) Incorrect iso-centre shift of Unacceptable magnitude
Gross Change in PTV Direction Although Rare
possible for: Bilaterally symmetric organs like breast
or partial head and neck and brain cases.

4. How to identify and hence correct Grass error
Place the patient on couch do a LASER alignment with
Fiducial/BB; Check the SSD’s
Check table to Isocentre (TI) height : Should Match with CT TI

5. After matching the CT TI with the Setup TI
Apply the shifts (BB to ISO) obtained from TPS and
check the SSD values (all three or which ever visible)
and TI
Score SSD values and TI with the TPS obtained
values
Any gross error recheck the shifts in TPS and
performed shift.

6. Part-I: Non Image Based Correction Strategies
Under all circumstances
TI should exactly match
with CT Simulation TI
(for Bulky pelvis
patients with in ±3 mm)
-Otherwise reset
the patient
Where it is not Possible
to match TI /SSD
Patient having
(1) severe pain
(2) can not laid down
comfortably
(3) Old and unstable patient.

7. Part-I: Non Image Based Correction Strategies
 SSD and TI Values will give the envelope where patient
lying in three dimensional Euclidian space.
Iso-centre
Lt Lateral
SSD
Rt Lateral
SSD
Table
Table to Isocentre Height
Anterior SSD
Envelope

8. SSD and TI
Intended and measured TI values should be noted for
future references; before start of the therapy delivery .
Caution: (1) Do not do a correction on basis of SSD
unless otherwise a gross shift is observed.
(2) If you are quipped even with most primitive
imaging techniques; take a image and see.

9. Who should practice this method :
 TI Verification
 All Centres ; for all patient before 1St Fraction.
 SSD verification
 Philosophically
 Every body: Including most advance radiotherapy centres.
 Technically
If the Machine have a SSD Scale.
Where TI (or) /SSD Verification should NOT be
practiced.
 Any kind of Stereotaxy involving a external reference frame
specially in frameless stereotaxy.
Require a good work practice in
Baseline setup (TI & SSD matching)
Otherwise whatever Imaging and
protocol will not help to achieve a good
result.

10. Systematic and random Error
Systematic error: Is a deviation that occur in the same direction and
is of a similar magnitude for each fraction throughout treatment
course
Random : Is a deviation that very in direction and magnitude for each
fraction
Random Component
Systematic Component
Courtesy: Prof Sonke JJ

11. Variation Management vs Target margin
Manage variation Effectively : Reduce margin
How to reduce margin: Off line :
Focus on Systematic error Collect as much
data as possible take necessary action to
reduce systematic error
Courtesy: Prof Sonke JJ

12. Part -II
Image Based Positional Error Correction.
References
De Boer HC, Heijmen BJ. A protocol for the reduction of systematic patient
setup errors with minimal portal imaging workload. International Journal of
Radiation Oncology* Biology* Physics. 2001 Aug 1;50(5):1350-65.
(2) Bel A, Van Herk M, Bartelink H, Lebesque JV. A verification procedure to
improve patient set-up accuracy using portal images. Radiotherapy and
Oncology. 1993 Nov 1;29(2):253-60.
(3) van So¨rnsen de Koste JR, de Boer JC, Senan S, et al. Reduction
of PTVs in lung cancer patients by CT-simulation and use
of DRRs in setup verification. Proceedings of the 6th International
Workshop on Electronic Portal Imaging; 2000. p. 7.

13. How should a set-up correction be made?
A deviation  is observed
Correction c = - is applied [online correction strategy]
Question: Is this feasible for all patients? Definitely No
Incorporate the determined setup margins while contouring
Decisions during imaging sessions for subsequent patients
should take into consideration:
Setup margin used at the time of contouring
Various action levels (thresholds) set by department
protocol
 Offline correction strategy
Online correction strategy
15

14. Online Correction strategies
Positional correction should be made before therapy delivery
using 2D, 3D (/6D), 4D imaging.
No Correction Required (For PTV) Positional error well
within setup margin.
Although inside the setup margin Positional Correction
should be done when OAR’s are at potential overdosing risk.
Online Correction: Can correct for random error and reduce the
systematic error.
Question: Is this feasible for all patients?
Answer: No

15. Pre Correction Imaging
Post correction Imaging

16. 18 Courtesy: Dr. T Ganesh

17. 19
Where margins are tight  high possibility of geometrical
miss
Most Interesting example is Breast radiotherapy using VMAT
/IMRT or any conformal techniques
Where should a Online set-up correction be made?

18. Is there a benefit of daily online correction strategy?
It will be of benefit for patients
With very large random variations and/or
Having target volumes in close proximity to critical
structures
The majority of benefit in margin reduction comes not from
reduction of random error, but in fact from minimization of the
systematic error
Choose patients for online correction strategy judiciously
20

19. Offline correction strategy
Images before treatment are acquired
Match to a reference image is made offline (i.e.
without the patient on the couch) – after treatment
Reduces both the
Magnitude of the individual patient systematic set-
up error
….and the population systematic error
22
Only for systematic errors; Random error cannot be corrected off-line)

20. 23

21. Offline correction strategy
No Action Level (NAL): It Involves the systematic error being
calculated after 3-4 fractions and a correction performed which
is the total magnitude of systematic error; regardless of the
tolerance of the treatment site.
eNAL: First 3-4 # imaging+ once weekly Imaging < Tolerance:
No action
Shrinking Action Level (SAL): uses a action level that
reduces according to number of fractions.
The running mean error over all acquired images is compared
with current action level and Tx Setup adjusted by this amount
if the discrepancy exceed the action level
24

22. NAL: The way to do it…
25
Ref
-T +TAL-G AL+G
 Day 1 - Verification image taken - Gross error found
 Reason identified – Setup adjusted – Image repeated
 No gross error present – Treated
 Note: Only gross error checked. No matching done; shifts not measured.

Courtesy: Dr. T Ganesh

23. NAL: The way to do it
26
Ref
-T +TAL-G AL+G
AL-NAL AL+NAL
  
 Imaging repeated for fraction #2 to #5 – systematic component of the setup error
is calculated as the mean of 5 shifts [Blue vertical line]
 Question: Should the isocenter be shifted or not?
 It depends on AL-NAL or AL+NAL – Action level for the NAL protocol
 Mean is less than AL-NAL or AL+NAL (<±2mm)  No need to shift isocenter
 In this case, mean (blue) is greater than AL-NAL or AL+NAL  Shift isocenter
Mean of 5 shifts
Courtesy: Dr. T Ganesh

24. NAL: The way to do it…
27
New
Ref
-T +TAL-G AL+G
 Reference is adjusted – We have accounted for the systematic component – We
now have a NEW REFERENCE
 Images taken on #6, #11, #16
 Found within tolerance (week 2, 3, 4)
 However, for image taken on #21 shift is found out of tolerance
 What to do now?
Courtesy: Dr. T Ganesh

25. NAL: The way to do it…
28
Ref
-T +TAL-G AL+G
 Weekly imaging (#21, week 5) is found to be out of tolerance (offline). What
to do now?
 Action taken
 Check setup instructions and annotations
 Repeat the verification image at the NEXT fraction (i.e. #22) – find out the
reason if it is out of tolerance again
Courtesy: Dr. T Ganesh

26. To implement NAL offline correction strategy
Action levels for Gross Error (ALG)
Tolerance values (T)
Shifts within these tolerance values need not require
correction
ALNAL for determining whether to shift the isocenter or not
29
Ref
-T +TAL-G AL+G
AL-NAL AL+NAL
Courtesy: Dr. T Ganesh

27. NAL Protocol
Courtesy: Prof Sonke JJ

28. Shrinking Action Level (SAL)
Uses an action level that reduces according to the number
of fractions imaged
Calculate running mean error over all acquired images
Compare with current action level
If it (mean error) exceeds, adjust setup
Adv: Avoids setup being corrected prematurely
Disadvantage: Following any correction, process is restarted
and information obtained prior to the restart is lost
31
Courtesy: Dr. T Ganesh

29. Shrinking Action Level (SAL)
32
First treatment session
Measurement set-up deviation
N = N+1
N > Nmax ?
Correct next
set-up with:
c = - 
N = 1
STOP
measure-
ment
 > 
N
YES
NO
YES
NO
Underlying concept:
 As the number of treated
fractions increase, SHRINK
the action level
 Once you make corrections,
subsequent fractions should
NOT show significant
deviations
∑- Systematics Error
α- Initial action level
Nmax - Number of
fractions without
Correction
Courtesy: Dr. T Ganesh

30. SAL Protocol
Courtesy: Prof Sonke JJ

31. The real benefit: Speedy treatment executions
Set your tolerance values (T) equal to the margins
you have incorporated
Shifts will rarely be required34
Determine
,  and
margins
Incorporate
the data in
contouring
Daily
treatment
executions
Courtesy: Dr. T Ganesh

32. The common mistake: Mixing up of offline & online :
ad hock correction protocol
If a patient is designated for offline correction
strategy, ONLY that strategy should be followed
If you measure and correct the shifts with the
patient lying on the table, then you are mixing two
incompatible strategies
End result:
You don’t gain anything
Waste of time
Probability of wrong shifts
35
Courtesy: Dr. T Ganesh

33. Workflow
Doctors decide whether offline OR online
Offline
Cooperative; less anxiety; stable region (skull based); no
proximity to critical OARs; large number of fractions (20 or
more)
Online
Non-cooperative; high anxiety levels; proximity to critical
OARs; less number of fractions – either daily or alternate
days
36
Courtesy: Dr. T Ganesh

34. Workflow
If offline, will also decide
NAL or SAL (one can start with NAL)
NAL
− Whether first 3 or 4 or 5 fractions
− Tolerance (will be site specific)
37
Courtesy: Dr. T Ganesh

35. Summary
Margins remain a problem in radiotherapy
Assuming CTV is delineated accurately, treatment
planning step should ensure its adequate coverage
by
Accurate patient modeling
Treatment verification strategies that understand
the patient specific nature of setup variation
38

36. Part –III
Rotational error and its correction
strategies
(Rigid and Deformed rotation)

37. Incorporating Rotational Error In Routine Clinical Practice
 All modern Imaging system CBCT/Exactrack offers a 6D shift
which can be executed by a robotic couch
 However no contouring station/Planning system accept a CTV to
PTV in terms of rotation
 No margin formula account for Rotational error(s).

38. Where Rotational setup errors are most important:
Frameless stereotaxy
 Studied in frameless stereotaxy about rotational errors
Whenever loaded Couch will act like a LEVER of Class-1
LASER Shift
Due to Fulcrum
effect ≈ 1 cm

39. Consequence
 If not corrected it offers a complete geometrical miss of the tumour.
Calculated
Pre Correction
Post Correction

40. How to Tackle Rotational Error: What other people said
Handel Rotational error
independently
IJROBP
JACMP
BJR

41. ONIMARU et.al tried to combine the rotational and
translational error : Using matrix method
 However it have a incorrect mathematics : Matrix method indicate a
sequential rotational movement
 In an Euclidian Space translational movements are independent but
rotational motions are dependent in translational motion.
 All motions (translational and rotational) should be simultaneous which
Onimaru does not proposed

42. What we required to effectively tackle the
rotational Errors ?
A formulation which can convert rotational shifts in terms of
translational shift Or 6D to resultant 3D translational shift
If a, b, c, α, β, γ are translational and rotational shift
respectively then resultant translational shift is

43. 0
2
4
6
8
10
12
14
16
18
20
0.00 0.50 1.00 1.50 2.00 2.50
Volumeinccandpercentage(inLOG10
scale)
GTV radius in cm
Variation of absolute volume between PTV_R and PTV_NR (cc)
Variation of % volume between PTV_R and PTV_NR
PTV With rotational correction (PTV_R)
PTV without any rotational correction (PTV_NR).
Beyond 4 cm GTV/CTV
diameter mere influence of
rotational correction (<3%).
3 %

44. Clinical significance
6D to resultant 3D will allow to incorporate the rotational
margin (in-terms of translation) While calculating CTV to PTV
Pre
Correction
Post
Correction
Conclusion:
(1) Stertiotactic GTV to CTV margin is ≈5 mm without
appropriate table position correction Unacceptable high.
(2) Established 1 mm margin for Stereotaxy
(3) Incorporating Rotation reduces PTV margin

45. Advantage of converting 6D shift to 3D
formulation
In general if Robotic couch is not working  Not possible to
treat Frameless stereotactic patients.
Validation result suggest 6D reduced Resultant 3D shift
offers enough spatial dosimetric accuracy in dose coverage.
Hence one can treat a frameless stereotaxy patient Even
with a Regular 3D Couch or Robotic Couch
Not Functioning.
However we strongly recommend a repeat CBCT after
Resultant 3D shift.

46. In TPS Only three dimensional shifts are applicable
3D translation shifts were calculated (for pre and post table
correction condition) from 6D CBCT data
Formula Validation :TPS

47. Matching techniques: 2D /3D/4D/Adaptive/Deformed
 At least 3 Structure should be visible in the outlined field.
 Anatomically atoned structures are most stable structure

48. Matching techniques: 2D /3D/4D/Adaptive/Deformed
One need to standardise the matching
protocol for the institutions
FMRI CBCT matching Protocol
T+R corrections Online and T for Offline
Site Matching technique
Brain Bone
Pelvis Soft tissue
Thorax Soft tissue + Spine (if near)
Abdomen (liver/ pancreas) Soft tissue
Head and neck Bone+ Soft tissue
Standardisation and Judicious choice of clip box is important to get a good
matching
Note : For bone matching take a rigid correlated structures

49. Matching techniques: 2D /3D/4D/Adaptive/Deformed
 Brain: Most Standard and easiest Matching technique

50. Matching techniques: 2D /3D/4D/Adaptive/Deformed
 Head Neck: difference in the matching Clip-box changes
the rotational error ; However not significant

51. Matching tec: 2D /3D/4D/Adaptive/Deformed H&N
Human Anatomy behaves as a semi-rigid body hence shows
a deformation; sometime deformation attributes to anatomical
changes as well

52. Matching tec: 2D /3D/4D/Adaptive/Deformed Pelvis
Improper preparation
Bladder feeling was wrong during simulation : Very common in
Summer in Delhi due to heavy dehydration
 Do not try to match in such situation : Have an adaptive CT

53. Matching tec: 2D /3D/4D/Adaptive head neck/ De
Adaptive CT is most common in H/N cancer in our Centre

54. Matching techniques: 2D /3D/4D/Adaptive/Deformed
 Breast: Judicious choice of Clip box is essential

55. 4D Imaging / Lung SBRT / Adaptive radiotherapy
Matching techniques: 2D /3D/4D/Adaptive/Deformed

56. 4D Imaging / Lung SBRT / Adaptive radiotherapy
Matching techniques: 2D /3D/4D/Adaptive/Deformed
1. ITV Based
technique
2. Gated technique
ITV based technique
Using 4D CT room to
shrink the PTV margin
depending upon the
Tumour motion.
For Gated Technique
only 3D-3D matching
Do NOT Mix 3D with
4D imaging

57. 4D Imaging / Lung SBRT / Adaptive radiotherapy
Matching techniques: 2D /3D/4D/Adaptive/Deformed
1. ITV Based
technique
2. Gated technique
ITV based technique
Using 4D CT room to
shrink the PTV margin
depending upon the
Tumour motion.
For Gated Technique
only 3D-3D matching
Do NOT Mix 3D with
4D imaging

58. Where Online set-up correction is difficult to
make Even with 4D CT
63
Most prominent Example is Liver SBRT Imaging in 4D

59. Results: CTV to PTV Margin EPID and Exactrac
Site
Commo
n
session
s
iView GT
sessions
Exactra
c
session
s
Directio
n
iView GT(cm) Exactrac (cm)
Systema
tic ()
Random
()
Margin
Systema
tic ()
Random
()
Margin
Cranium 2708 2654 1537
Sup/Inf 0.18 0.21 0.60 0.21 0.21 0.67
Right/Left 0.09 0.13 0.32 0.18 0.20 0.59
Ant/Post 0.12 0.34 0.54 0.20 0.15 0.61
Breast 2004 1053 1358
Sup/Inf 0.10 0.24 0.43 0.25 0.38 0.89
Right/Left 0.31 0.38 1.04 0.37 0.32 1.15
Ant/Post 0.15 0.28 0.58 0.41 0.43 1.33
Head &
Neck
2320 1735
1356
Sup/Inf 0.14 0.15 0.44 0.16 0.14 0.49
Right/Left 0.09 0.10 0.30 0.21 0.16 0.63
Ant/Post 0.13 0.43 0.64 0.14 1.39 1.31
Thorax 2330 1435 1236
Sup/Inf 0.32 0.48 1.13 1.58 5.00 7.46
Right/Left 0.27 0.36 0.92 0.35 0.80 1.43
Ant/Post 0.11 0.25 0.44 0.36 0.77 1.43
Pelvis 2230 1375
1260
Sup/Inf 0.39 0.46 1.30 0.48 0.40 1.48
Right/Left 0.29 0.61 1.15 0.29 0.72 1.23
Ant/Post 0.12 0.26 0.48 0.18 0.22 0.61

60. Results: What is the correlation between them
0.511
0.315
0.698
0.315
-0.053
0.263
0.344
0.284
0.339
0.044
0.892
0.897
-0.0840.16
0.209
0.467
0.566 0.208
0.215
-0.2
0
0.2
0.4
0.6
0.8
1
MV_Z_Pelvis_KV
MV_X_Braest_KV
MV_Y_Braest_KV
MV_Z_Braest_KV
MV_X_brain_KV
MV_Y_brain_KV
MV_Z_brain_KV
MV_X_H&N_KV
MV_Y_H&N_KV
MV_Z_H&N_KVMV_X_Brain mets_KV
MV_Y_Brain mets_KV
MV_Z_Brain mets_KV
MV_X_TA_KV
MV_Y_TA__KV
MV_Z_TA_KV
MV_X_LUNG_KV
MV_Y_LUNG_KV
MV_Z_LUNG_KV
Pearson Corelation cofficient Between Kv
and Mv imaging Pearson Corelation

61. Summary
Feedback loop between treatment verification and
contouring should be correctly established & strong
Develop meaningful imaging protocols
Determine clinical set-up accuracy
For each site (for a patient population)
Apply treatment margins accordingly
What we contour is what we treat!
66

62. I cannot tell how the truth may be;
I say the tale as it was said to me."
— Walter Scott
"
Thank You!