Background
Methods
Plan reconstruction from dynalog files
Dose-volume histogram percentage of agreement
PA
) as an indicator of the similarity of two DVHs. Given DVH1 and DVH2, the PA
is defined as
Validation of the plan reconstruction
GPR
, i.e., the percentage of voxels that pass gamma analysis [12] with criteria 1%,1 mm, better than 95%. The size of the PSF is 23 Gigabytes. For the patient-dependent part of the linac and the voxelized geometries, DPM was selected as the Monte Carlo radiation transport engine. Simulations were run for 1×108 histories in a dual Xeon E5-2670V3 CPU with 12 cores each, and hyper-threading. The simple splitting variance-reduction technique was applied in the patient geometry with a splitting factor of 300. The obtained dose distributions had an average standard statistical uncertainty less than 1% in all cases.GPR
2,1) and by evaluating the DVHs percentage of agreement. All the analysis was done with the functions available in the PRIMO system.Sensitivity analysis
GPR
) and the PA
. For this purpose, the position errors captured in the dynalog files of the two clinical cases were magnified. Magnification was made by rescaling the errors up to a maximum error Σ. Only errors larger than 0.01 mm were magnified. For scaling, the altered “actual" position of a leaf, \(P^{\prime }_{a,}\) was calculated as,
PA
, (GPR
2,2) and (GPR
3,3) were calculated for the body region, PTVs and OARs defined for the clinical cases. The root-mean-square error (RMS
) of all leaf positions in the dynalog files was evaluated in each case as,
Results
Verification of the plan reconstruction
Region | Low resolution | High resolution | |||
---|---|---|---|---|---|
PA [%] | GPR 2,1 [%] | PA [%] | GPR 2,1 [%] | ||
Prostate | |||||
PTV
1
| 99.5 | 100 | 100 | 100 | |
PTV
2
| 99.5 | 100 | 100 | 100 | |
PTV
3
| 99.6 | 100 | 100 | 100 | |
PTV
4
| 99.7 | 100 | 99.9 | 100 | |
PTV
total
| 99.6 | 100 | 99.9 | 100 | |
Rectum | 99.7 | 99.8 | 99.9 | 99.9 | |
Bladder | 99.8 | 100 | 99.7 | 100 | |
Body | 99.9 | 99.4 | 99.9 | 99.4 | |
H&N | |||||
PTV
1
| 99.5 | 99.9 | 99.7 | 100 | |
PTV
2
| 99.7 | 99.4 | 99.9 | 99.9 | |
Spinal canal | 99.7 | 99.9 | 99.7 | 99.8 | |
Parotid left | 99.7 | 99.8 | 99.6 | 99.9 | |
Parotid right | 99.4 | 99.8 | 100 | 99.8 | |
Body | 99.8 | 99.4 | 99.8 | 99.4 |
Clinical case | Plan | Control points | Time [min] |
---|---|---|---|
Prostate | Original low resolution | 177 | 25 |
Original high resolution | 1594 | 35 | |
Reconstructed from expected positions | 1536 | 29 | |
Head&Neck | Original low resolution | 194 | 41 |
Original high resolution | 1561 | 50 | |
Reconstructed from expected positions | 1584 | 49 |
Sensitivity analysis
RMS
of 0.68 and 0.47 mm for the prostate and head&neck cases, respectively, the values obtained for PA
and GPR
2,2 are similar to those obtained for the comparison of the original doses with the expected doses. The impact on the dose is however noticeable for Σ=30 mm with RMS
of 2.03 and 1.41 mm for the prostate and head&neck cases, respectively.
Region | Σ=10 mm | Σ=30 mm | |||||
---|---|---|---|---|---|---|---|
PA [%] | GPR 2,2 [%] | GPR 3,3 [%] | PA [%] | GPR 2,2 [%] | GPR 3,3 [%] | ||
Prostate | |||||||
PTV
1
| 99.6 | 100 | 100 | 98.3 | 90.7 | 99.9 | |
PTV
2
| 99.5 | 100 | 100 | 98.1 | 92.2 | 99.9 | |
PTV
3
| 99.6 | 100 | 100 | 98.4 | 97.7 | 100 | |
PTV
4
| 99.5 | 100 | 100 | 98.2 | 94.5 | 99.5 | |
PTV
total
| 99.5 | 100 | 100 | 98.3 | 94.9 | 99.7 | |
Rectum | 99.8 | 100 | 100 | 99.2 | 100 | 100 | |
Bladder | 99.0 | 100 | 100 | 97.2 | 99.3 | 99.9 | |
Body | 99.5 | 100 | 100 | 98.6 | 99.8 | 100 | |
H&N | |||||||
PTV
1
| 99.7 | 100 | 100 | 98.6 | 93.3 | 100 | |
PTV
2
| 99.7 | 100 | 100 | 98.9 | 98.2 | 100 | |
Spinal canal | 99.6 | 100 | 100 | 99.0 | 100 | 100 | |
Parotid left | 99.8 | 100 | 100 | 99.5 | 100 | 100 | |
Parotid right | 99.3 | 100 | 100 | 98.1 | 100 | 100 | |
Body | 99.7 | 100 | 100 | 99.2 | 99.9 | 100 |
RMS
as small as 0.14 mm, as it is shown in Tables 4 and 5. Columns marked with an asterisk (∗) correspond to the (unmodified) dynalog file as it was generated by the MLC controller during the treatment. Tables 4 and 5 also show that the PA
is more sensitive than the GPR
. GPR
2,2 is insensitive to a RMS
<0.24 mm for the head&neck case and to a RMS
<0.28 mm for the prostate case. Values of GPR
3,3 lower than 99% were obtained only for Σ=10.0 mm (not shown) for both clinical cases. It was observed that, in general, sensitivity of the GPR
is dependable on the size of the region in which it is calculated. Notice e.g., that for Σ=10.0 mm, GPR
2,2 drops to 0 for the small volume (13.5 cm3) PTV
1 of the prostate case; however, it is 98.4% for the body region with volume 28554 cm3.
PA
resulting from the comparison of the dose obtained from the plan reconstructed from expected positions with the dose obtained from a plan in which the absolute value of position errors |ε| were scaled up to a maximum ΣPA [%] | ||||||
---|---|---|---|---|---|---|
Region (prostate) | Σ[mm] | |||||
0.33 (∗) | 2.0 | 3.0 | 4.0 | 5.0 | 10.0 | |
RMS[mm] | ||||||
0.02 | 0.14 | 0.21 | 0.28 | 0.34 | 0.68 | |
PTV
1
| 100 | 99.1 | 98.6 | 98.2 | 97.7 | 95.5 |
PTV
2
| 100 | 99.0 | 98.6 | 98.1 | 97.7 | 95.4 |
PTV
3
| 100 | 99.1 | 98.6 | 98.2 | 97.8 | 95.6 |
PTV
4
| 100 | 99.0 | 98.5 | 98.0 | 97.6 | 95.3 |
PTV
total
| 100 | 99.0 | 98.6 | 98.1 | 97.7 | 95.4 |
Rectum | 100 | 98.8 | 98.1 | 97.6 | 97.0 | 94.3 |
Bladder | 100 | 98.4 | 97.6 | 98.9 | 96.2 | 92.7 |
Body | 100 | 98.8 | 98.3 | 97.7 | 97.2 | 94.7 |
Region (H&N) | Σ[mm] | |||||
0.55 (∗) | 2.0 | 3.0 | 4.0 | 5.0 | 10.0 | |
RMS[mm] | ||||||
0.03 | 0.10 | 0.14 | 0.19 | 0.24 | 0.47 | |
PTV
1
| 100 | 99.5 | 99.1 | 98.8 | 98.5 | 97.0 |
PTV
2
| 100 | 99.5 | 99.3 | 99.1 | 98.8 | 97.6 |
Spinal canal | 99.9 | 99.4 | 99.1 | 98.8 | 98.6 | 97.3 |
Parotid left | 99.9 | 99.8 | 99.6 | 99.3 | 99.2 | 98.5 |
Parotid right | 99.9 | 99.1 | 98.3 | 98.1 | 97.2 | 95.2 |
Body | 100 | 99.5 | 99.2 | 99.0 | 98.7 | 97.4 |
GPR
2,2 in percentage resulting from the comparison of the dose obtained from the plan reconstructed from expected positions with the dose obtained from a plan in which the absolute value of position errors |ε| were scaled up to a maximum ΣGPR 2,2[%] | ||||||
---|---|---|---|---|---|---|
Region (prostate) | Σ[mm] | |||||
0.33 (∗) | 2.0 | 3.0 | 4.0 | 5.0 | 10.0 | |
RMS[mm] | ||||||
0.02 | 0.14 | 0.21 | 0.28 | 0.34 | 0.68 | |
PTV
1
| 100 | 100 | 99.9 | 95.0 | 67.5 | 0.0 |
PTV
2
| 100 | 100 | 99.7 | 98.4 | 92.2 | 5.5 |
PTV
3
| 100 | 100 | 100 | 99.3 | 96.7 | 21.6 |
PTV
4
| 100 | 100 | 99.8 | 98.2 | 92.9 | 29.9 |
PTV
total
| 100 | 100 | 99.8 | 98.4 | 92.5 | 22.4 |
Rectum | 100 | 100 | 100 | 99.9 | 99.7 | 90.8 |
Bladder | 100 | 100 | 100 | 99.9 | 99.8 | 91.7 |
Body | 100 | 100 | 100 | 100 | 99.9 | 98.4 |
Region (H&N) | Σ[mm] | |||||
0.55 (∗) | 2.0 | 3.0 | 4.0 | 5.0 | 10.0 | |
RMS[mm] | ||||||
0.03 | 0.10 | 0.14 | 0.19 | 0.24 | 0.47 | |
PTV
1
| 100 | 100 | 100 | 99.3 | 93.9 | 27.8 |
PTV
2
| 100 | 100 | 100 | 99.8 | 98.6 | 74.9 |
Spinal canal | 100 | 100 | 100 | 100 | 98.6 | 100 |
Parotid left | 100 | 100 | 100 | 100 | 98.2 | 100 |
Parotid right | 100 | 100 | 100 | 100 | 97.2 | 100 |
Body | 100 | 100 | 100 | 100 | 100 | 99.1 |
PA
of PTV 1 versus RMS
obtained for the sensitivity tests that conserve or not the sign of the leaf position error ε. Both clinical cases are included, but not differentiated, in the table. The table shows that, as expected, systematic differences between the dose distributions are directly proportional to the RMS
. It also shows that, with independence on the sign of the leaf position error, when roughly 50% or more voxels have systematic deviations larger than 1.2%, the value of PA is less than 99%. This suggests that PA ≤99% could be established as a threshold for treatment verification failure.
RMS
and the PA
of PTV 1RMS [mm] | α[%]/Δ[%] of ref. max. dose | PA[%] of PTV
1
| |
---|---|---|---|
ε
| 0.47 | 6/0.5 | 99.7 |
0.68 | 32/0.7 | 99.6 | |
0.95 | 31/0.9 | 99.2 | |
1.35 | 53/1.2 | 99.1 | |
1.41 | 46/1.1 | 98.6 | |
2.03 | 63/1.7 | 98.3 | |
|ε| | 0.10 | 18/0.6 | 99.5 |
0.14 | 36/0.7 | 99.1 | |
0.19 | 48/0.9 | 98.8 | |
0.21 | 77/1.2 | 98.6 | |
0.24 | 61/1.1 | 98.5 | |
0.28 | 87/1.5 | 98.2 | |
0.34 | 93/1.8 | 97.7 | |
0.47 | 85/2.0 | 97.0 | |
0.68 | 96/3.4 | 95.5 |
Discussion and conclusions
RMS
<1.2 mm would pass the verification if made by the methods described here. When errors are predominantly in one direction as e.g., in the failure of a MLC carriage, they can be detected in the dose for RMS
as low as 0.2 mm. These two findings put together indicate that the impact on the dose cannot just be inferred from the RMS
. Instead, the PA
evaluated in the PTVs and the percentage of voxels with a given systematic dose deviation are quite sensitive measures of that impact.PA
, hereby introduced, showed to be more sensitive than GPR
2,2. Also that GPR
3,3, in general, and GPR
2,2 evaluated for the patient body region, are not per se good evaluators of deviations introduced in the dose by errors captured in the dynalog files.