Background
Methods and materials
Patient selection
P1
|
P2
|
P3
|
P4
|
P5
|
P6
|
P7
|
P8
| |
---|---|---|---|---|---|---|---|---|
Tumour Site
| L Tonsil | L Tonsil | R Tonsil | Nasopharynx | L Tongue | R Tonsil | Para-nasal sinus | Tongue |
No. fractions
| 30 | 30 | 30 | 30 | 30 | 30 | 30 | 30 |
Surgery
| Yes | No | No | No | Yes | No | Yes | No |
Staging
| T2N1M0 | T3N2bM0 | T1N0M0 | T1N0M0 | T2N2bM0 | T2N2M0 | T4N2M0 | T4N3aM0 |
GTV [
cm
3
]
| 14.1a
| 88.8 | 26.2 | 82.8 | - | 47.5 | - | 201.1 |
CTV
50–54
[cm
3
]
| 201.8 | 446.1 | 144.6 | 567.5 | 299.4 | 314.1 | - | 179.1b
|
CTV
60
[cm
3
]
| - | - | - | - | 214.3 | - | 583.4 | 331.7 |
CTV
66
[cm
3
]
| 68.5 | 186.8 | 38.8 | 153.2 | 57.7 | 108.1 | - | - |
Additional imaging
| PET | PET | - | MRI | MRI | - | MRI | - |
Number of CTs
| 2 | 3 | 2 | 2 | 2 | 2 | 3 | 2 |
Days after CT
1
| 42 | 26/42 | 36 | 30 | 25 | 35 | 42/69 | 31 |
Days after RT start
| 25 | 8/24 | 21 | 16 | 7 | -c
| -c/20 | 20 |
Weight change between CTs
[kg]
| −7.1 | −10.7/+0.8 | −1.3 | −3.8 | −2.6 | 0.0 | +1.6/- 0.8 | −6.8 |
Reason for re-
CT
| Tumour shrinkage | Patient position | Swelling | Patient position | Swelling | Patient position | Patient position | Tissue Loss |
New Mask
| Yes | Yes/No | Yes | No | No | No | No/No | No |
New plan
| Yes | Yes/Yes | Yes | No | No | No | NA/No | No |
Structure selection
Workflow
Quantitative analysis
Qualitative analysis
Time saving
Deformable image registration software and algorithm
Results
Accuracy of DIR generated structures
P1
|
P2
|
P3
|
P4
|
P5
|
P6
|
P7
|
P8
|
Mean
|
S.D
| |
---|---|---|---|---|---|---|---|---|---|---|
GTV
| 0.80 | 0.83 | 0.84 | 0.89 | 0.93 | 0.77 |
0.84
|
0.06
| ||
CTV
54
| 0.88 | 0.89 | 0.84 | 0.87 | 0.91 | 0.92 | 0.84 |
0.88
|
0.03
| |
CTV
60
| 0.91 | 0.88 | 0.86 |
0.89
|
0.02
| |||||
CTV
66
| 0.85 | 0.92 | 0.82 | 0.89 | 0.87 | 0.94 |
0.88
|
0.04
| ||
Brain Stem
| 0.83 | 0.96 | 0.67 | 0.92 | 0.88 | 0.91 | 0.80 | 0.68 |
0.83
|
0.11
|
Spinal Cord
| 0.78 | 0.84 | 0.73 | 0.75 | 0.74 | 0.78 | 0.77 | 0.63 |
0.75
|
0.06
|
R Parotid
| 0.73 | 0.76 | 0.79 | 0.74 | 0.90 | 0.90 | 0.77 | 0.54 |
0.77
|
0.11
|
L Parotid
| 0.82 | 0.77 | 0.84 | 0.86 | 0.82 | 0.91 | 0.82 | 0.71 |
0.82
|
0.06
|
Mean
|
0.81
|
0.85
|
0.79
|
0.84
|
0.86
|
0.90
|
0.81
|
0.72
| ||
S.D. |
0.05
|
0.07
|
0.07
|
0.07
|
0.06
|
0.05
|
0.05
|
0.12
|
Analysis of a potential bias
P1
|
P2
|
P3
|
P4
|
P5
|
P6
|
P7
|
P8
|
Mean
|
S.D
| |
---|---|---|---|---|---|---|---|---|---|---|
GTV | 0.90 | 0.83 | 0.99 | 0.97 | 0.98 | 0.97 |
0.94
|
0.06
| ||
CTV54 | 0.95 | 0.97 | 0.99 | 0.98 | 0.99 | 0.99 | 0.96 |
0.98
|
0.01
| |
CTV60 | 0.99 | 0.94 | 0.97 |
0.96
|
0.03
| |||||
CTV66 | 0.97 | 0.87 | 0.95 | 0.98 | 0.98 | 0.98 |
0.95
|
0.04
| ||
Brain Stem | 0.86 | 1.00 | 0.68 | 0.98 | 0.98 | 0.95 | 0.87 | 0.65 |
0.87
|
0.14
|
Spinal Cord | 1.00 | 0.95 | 0.82 | 0.89 | 0.89 | 0.91 | 0.77 | 0.91 |
0.89
|
0.07
|
R Parotid | 0.89 | 1.00 | 1.00 | 0.98 | 0.98 | 0.95 | 0.90 | 0.95 |
0.95
|
0.04
|
L Parotid | 0.93 | 1.00 | 0.88 | 0.98 | 0.97 | 0.94 | 0.85 | 0.74 |
0.91
|
0.09
|
Mean
|
0.93
|
0.94
|
0.90
|
0.97
|
0.97
|
0.96
|
0.86
|
0.88
| ||
S.D. |
0.05
|
0.07
|
0.12
|
0.04
|
0.04
|
0.03
|
0.06
|
0.13
|
Impact of registering contrast-enhanced and non-enhanced CT scans
P2
|
P7
| |||
---|---|---|---|---|
CT2
|
CT3
|
CT2
|
CT3
| |
GTV
| 0.83 | 0.66 | ||
CTV
54
| 0.89 | 0.87 | ||
CTV
60
| 0.88 | 0.89 | ||
CTV
66
| 0.92 | 0.79 | ||
Brain Stem
| 0.96 | 0.84 | 0.80 | 0.82 |
Spinal Cord
| 0.84 | 0.75 | 0.77 | 0.79 |
R Parotid
| 0.76 | 0.82 | 0.77 | 0.74 |
L Parotid
| 0.77 | 0.83 | 0.82 | 0.81 |
Mean
|
0.85
|
0.79
|
0.81
|
0.81
|
S.D. |
0.07
|
0.07
|
0.05
|
0.05
|
Qualitative analysis
Time saving
Patient
|
CT set
|
Adapt
|
Re-contour
|
Gain
|
---|---|---|---|---|
P1 | CT2 | 30 (±10) | 54(±1) | 24 |
P2 | CT2 | 30(±10) | 86(±1) | 56 |
P2 | CT3 | 30(±10) | 90(±1) | 60 |
P3 | CT2 | 30(±10) | 89(±1) | 59 |
P4 | CT2 | 30(±10) | 72(±1) | 42 |
P5 | CT2 | 30(±10) | 50(±1) | 20 |
P6 | CT2 | 27(±1) | 20(±1) | −7 |
P7 | CT2 | 21(±1) | 45(±1) | 24 |
P7 | CT3 | 15(±1) | 44(±1) | 29 |
P8 | CT2 | 45(±1) | 120(±1) | 75 |
Mean
|
29
|
67
|
38
| |
S.D. |
8
|
30
|
25
|
Discussion
Differences between DSI scores of structures or patients
Differences between quantitative and qualitative results
Review of discrepancies
-
Limitations in soft-tissue contrast in CT images Limitations in soft-tissue contrast in CT imaging hamper accurate delineation of specific organs such as parotids, spinal cord and brain stem. In addition, the distinction between individual vertebrae that are compressed, or between the external patient contour and build-up material, can be difficult. As the DIR algorithm uses the differences in Hounsfield units (HUs) to track patient deformation, these limitations may result in a lower DSI score. Figure 3a,b displays an example of this limitation resulting in a low DSI score for the spinal cord.
-
Intra-observer variation For a number of delineated structures, considerable differences were observed between the delineation of the same structure on the two consecutive CTs. Inspection of the scan acquisition settings revealed a variation in mAs settings but this did not noticeably affect the soft-tissue contrast of the CT images in this study. Although limitations in soft-tissue contrast may be an important cause of intra-observer variation, some of the observed differences clearly reflected a difference in interpretation of the two CT images that could not be explained by limitations in soft-tissue contrast only. Furthermore, some discrepancies were caused by the availability of new and/or additional information in clinical notes. Figure 3c shows an example of intra-observer variation in delineation of the brainstem.
-
Differences in CT-contrast enhancement Although no systematic difference between the DSI scores for data sets with or without contrast enhanced CT images could be demonstrated, it was estimated that contrast enhancement did impact the accuracy of the DIR result in a number of cases where structures were proximal to relatively large blood vessels.
-
Errors due to a limited CT slice resolution The CT slice resolution applied for treatment preparation in this study was 3 mm which is the default CT slice resolution for the majority of patients at the WBCC. Any uncertainty in the actual deformation at the superior and inferior ends of delineated structures can therefore potentially result in a 3 mm error at these locations and a decrease in overlap of DIR-generated and RO re-contoured structures. Figure 3d displays an example of this limitation resulting in a low DSI score for C4.
-
Clinical target volume not adapted for tumour shrinkage during treatment Tumour shrinkage in between the two consecutive CT images was observed for a number of patients. In those instances the DIR algorithm seemed to be able to closely track this phenomenon. However, as no proof exists that all microscopic lesions have been sterilised, changes to the CTV during the course of treatment are usually only applied very reluctantly. This was observed for a number of cases and contributed to a lower DSI score (Figure 3e-f).