Feasibility and Reproducibility of ConeBeam CT Guided Lung Radiotherapy using Registrations to Bone,

1
Feasibility and Reproducibility of Cone-Beam CT
Guided Lung Radiotherapy using Registrations to
Bone, Carina and Tumor J. Higgins1,2, A.
Bezjak1,2, K. Franks1, D. Payne1,2, J. Cho1,2, L.
Le2,3, J. Bissonnette1,2
1 Radiation Medicine Program, Princess Margaret
Hospital, Toronto, ON, Canada 2 University of
Toronto, Toronto, ON, Canada 3 Departments of
Biostatistics, Princess Margaret Hospital,
Toronto, ON, Canada
CONCLUSIONS
RESULTS
PURPOSE/OBJECTIVES
Traditionally in lung portal imaging, stable
bony landmarks have been used to correct set-up
error and increase target precision. Advances in
3-dimensional imaging though, have led to
improved visualization of anatomical structures
and soft tissue, thereby offering alternative
landmarks to consider for image-guidance. Soft
tissue (tumor) matching has already been adopted
in stereotactic lung Cone-Beam Computed
Tomography (CBCT) imaging 1. However, in
locally advanced lung cancer with both a moving
peripheral tumour and fixed central disease, the
optimal CBCT matching method to ensure target
coverage is still unknown. Improved
reproducibility and accuracy of image
registrations will contribute to initiatives such
as margin reduction, dose escalation and
decreased toxicity. Furthermore, in a patient
population where survival rates remain
disappointing 2, the quest to implement such
strategies is strongly mandated in order for
radiotherapy to impact treatment outcomes. The
goal of this study was to examine three different
thoracic landmarks and establish the optimal
matching method for conventional lung
image-guidance. Feasibility, reproducibility
and accuracy of automatic and manual image
registration was assessed using bone (spine),
carina and tumor as registration landmarks.
Table 1. Intra-Class Correlation Values gt0.75
indicate good reproducibility
A
B

• Image matching took an average of 1 minute per
  method/per patient except for manual tumor, which
  took 4 minutes.
• Inter-observer variation for each matching method
  is shown in Table 1
• Automated registrations using bone and carina
  resulted in similar degrees of tumor coverage
  (Table 2)

Figure 3. Carina position before (A) and after
(B) image registration
Carina is a promising and novel target
surrogate for conventional lung image-guided
radiotherapy (Figure 3). Although similar
results were obtained with bone matching,
automatic carina matching proved to be a less
ambiguous and more reproducible method.
Current practice in our institution is
Automatic Bone Matching with Visual Carina
Assessment to verify the match. Correlations
between Gross Tumor Volume (GTV) and carina
during respiration need to be investigated next
to release the full potential of this non-rigid
landmark.
 
METHODS
Table 2. of observations grading tumor coverage
vs. matching method

• Bland Altman Plots (Figure 2) showed no
  significant trends between automatic and manual
  registration methods for bone and carina

Day-1 verification CBCT data sets were analyzed
independently by 4 observers for 30
randomly-selected lung patients using 6
registration methods (720 observations).
Clip-boxes were created around each landmark
(Figure 1) for automated registrations, whilst
visual overlay of the planning and CBCT images
defined the manual process. All registrations
were timed and intra-class correlation (ICC)
calculated to assess reproducibility.
Subsequently, 4 radiation oncologists graded
tumor coverage, on CBCT images that displayed
results from each matching method, to determine
the accuracy of each method and landmark. The
grades were as follows Grade 1 Tumor within
internal target volume (ITV) Grade 2 Tumor
inside planning target volume (PTV) partly
outside ITV Grade 3 Tumor partly outside
PTV The Bland-Altman error analysis was used to
measure the level of agreement between manual
and automated registrations.
BONE

Figure 4. Tumor borders as seen on CBCT images
with PTV margin (Blue)

• Tumor border delineation on CBCT was ambiguous
  and challenging (Figure 4). Contrast enhancement
  at soft tissue interfaces is necessary before
  this method can be considered for conventional
  lung practice.

-0.4
-0.4
-0.4
-0.4
REFERENCES
CARINA
1 Purdie, T.G., et al, Int J Radiat Oncol Biol
Phys, 2007. 68(1) p. 243-52. 2 Rengan, R., et
al, Int J Radiat Oncol Biol Phys, 2004. 60(3) p.
741-7.
ACKNOWLEDGEMENTS
The authors acknowledge the following Radiation
Therapists for their contribution to data
collection Derek Chan, Heidi Chan, Jasmine Chen,
Catherine Dupuis, Brenda Fung and Chrison Lee.
Thank you also to Graham Wilson and Doug Moseley
for contributing to the facilitation of data
collection for this project. This project was
supported in part by Elekta Inc., B. Research
Grant.
-0.4
-0.4
-0.4
ITV
PTV
Figure 2. Bland-Altman Plots Upper Panel
Automatic Bone (AB) vs. Manual Bone (MB) Lower
Panel Automatic Carina (AC) vs. Manual Carina
(MC) Mean Difference, --- Upper Lower
Agreement Limit
A
B
C
Figure 1. Clip boxes for automatic matching (A)
bone (B) carina (C) tumor