Patients
This retrospective study was approved by the institutional research ethics board of our hospital. All patients were enrolled with prior written informed consent to treatment and the use of image data. We used 4D-CT images of seven patients with stage IE gastric MALT lymphoma who received RT at our institution between February 2019 and June 2020.
CT simulation procedures
To minimize variations in stomach volume, patients were instructed to fast for at least eight hours before planning simulation. They underwent CT simulation by using a Discovery RT CT scanner (GE Healthcare, Chicago, IL) in the supine position with their arms raised. 4D-CT scan was performed using a Smart Deviceless 4D application with the parameters of 120 kV, 70 mA, a gantry rotation time of 0.5 s, a slice thickness of 2.5 mm, and cine mode [15]. The cine durations were set to the respiratory cycles plus the gantry rotation time, and the cine images were sorted into 10 respiratory phases by a phase-binning algorithm. The average intensity projection (AIP) of CT images was generated from projection data of all respiratory phases. All CT images were exported to the treatment planning system (Eclipse, Version 15.6; Varian Medical Systems, Palo Alto, CA) and registered by the hardware arrangement.
Definition of target volumes and OARs
Two radiation oncologists with an experience of 5 and 17 years defined the target volumes and OARs, after reaching a consensus according to the Radiation Therapy Oncology Group contouring atlases [16]. In all patients, GTV was identified based on the endoscopic examination findings, and we confirmed that GTV was contained within the whole stomach. The CTV was defined as the whole stomach, which was delineated on the AIP CT image data sets using fused 4D-CT images to cover stomach of all 10 respiratory phases (CTV-4D) [15]. A margin of 10 mm, which included intra- and inter-fractional variations in stomach volume, respiratory movement, and patient set-up, was added to CTV-4D for generating the planning target volume (PTV) [3, 17]. OARs defined as the kidneys, liver, small bowel, and spinal cord were also delineated on the AIP CT images [15].
Treatment planning procedures
3D-CRT, IMRT, and VMAT plans were generated based on AIP CT images. The 3D-CRT plan consisted of four beams (gantry angles of 0°, 90°, 180°, and 270°) with 15-MV X-ray. Multileaf collimator apertures were created, encompassing the PTV with a 5 mm margin in all directions. The IMRT plan using sliding window technique consisted of seven beams (gantry angles of 24°, 75°, 126°, 177°, 231°, 282°, and 333°) with 6-MV X-ray. The VMAT plan consisted of a double arc of 360° with 6-MV X-ray [13]. The prescribed dose was 30 Gy in 20 fractions. The goal of each plan was as follows: the minimum coverage dose for 95% of the PTV (D95) > 95% of the prescribed dose, mean dose (Dmean) of the liver < 12.5 Gy, Dmean of each kidney < 10 Gy, maximum dose (Dmax) of the spinal cord < 30 Gy, and Dmax of the small bowel < 31.5 Gy [13]. The dose calculation was performed with the anisotropic analytical algorithm and a gird size of 2.5 mm.
Dosimetric parameters for plan evaluation
Each plan was analyzed using a dose-volume histogram (DVH). The PTV coverage was evaluated base on D95. The homogeneity index (HI) of the PTV was calculated as follows [18]:
where D1 and D99 are the minimum dose covering 1 and 99% of the PTV, respectively; and Dp is the prescription dose. The lower HI suggests a better homogeneity.
The conformity index (CI) of the PTV was calculated as follows [13]:
where BV95 is the volume of the body receiving 95% of the prescribed dose. The closer CI is to 1, the better conformity.
The OAR dose was evaluated by the Dmean for parallel organs (kidneys and liver), and the Dmax for serial organs (small bowel and spinal cord).
We used an RT plan analysis software; Plan IQ (Version 2.3.2; Sun Nuclear, Melbourne, FL), which obtains a comprehensive and objective assessment of treatment plans by calculating the plan quality metric (PQM). The PQM score settings were based on previous reports and clinical importance (Table 1) [13, 18]. The PQM score (%) was calculated as follows [13]:
where PQMraw means the total score of each dosimetric parameter, and PQMmax means the sum of the perfect score of each dosimetric parameter, which was set to 120.
Table 1
Plan quality metric scores for each dosimetric parameter
|
Target
|
|
|
OAR
|
|
|
|
|
|
PTV
|
|
|
Liver
|
Right kidney
|
Left kidney
|
Spinal cord
|
Small bowel
|
Score
|
D95 (cGy)
|
HI
|
CI
|
Dmean (cGy)
|
Dmean (cGy)
|
Dmean (cGy)
|
Dmax (cGy)
|
Dmax (cGy)
|
0
|
< 2550
|
0.2≦
|
ー
|
2000≦
|
1800<
|
1800<
|
3000<
|
|
1
|
2550
|
0.18
|
< 0.2,1.8<
|
1950
|
1800
|
1800
|
3000
|
|
2
|
2600
|
0.16
|
0.2,1.8
|
1900
|
1700
|
1700
|
2900
|
|
3
|
2650
|
0.14
|
0.3,1.7
|
1850
|
1600
|
1600
|
2800
|
|
4
|
2700
|
0.12
|
0.4,1.6
|
1800
|
1500
|
1500
|
2700
|
|
5
|
2750
|
0.1
|
0.5,1.5
|
1750
|
1400
|
1400
|
2600
|
|
6
|
2800
|
0.08
|
0.6,1.4
|
1700
|
1300
|
1300
|
2500
|
2700
|
7
|
2850
|
0.06
|
0.7,1.3
|
1650
|
1200
|
1200
|
2400
|
2750
|
8
|
2900
|
0.04
|
0.8,1.2
|
1600
|
1100
|
1100
|
2300
|
2800
|
9
|
2950
|
0.02
|
0.9,1.1
|
1550
|
1000
|
1000
|
2200
|
2850–2950,3050–3150
|
10
|
3000
|
0
|
1
|
1500
|
950
|
950
|
2100
|
3000
|
11
|
|
|
|
1450
|
900
|
900
|
2000
|
|
12
|
|
|
|
1400
|
850
|
850
|
1900
|
|
13
|
|
|
|
1350
|
800
|
800
|
1800
|
|
14
|
|
|
|
1300
|
750
|
750
|
1700
|
|
15
|
|
|
|
1250
|
700
|
700
|
1600
|
|
16
|
|
|
|
1200
|
650
|
650
|
1500
|
|
17
|
|
|
|
1150
|
600
|
600
|
1400
|
|
18
|
|
|
|
1100
|
550
|
550
|
1300
|
|
19
|
|
|
|
1050
|
500
|
500
|
1200
|
|
20
|
|
|
|
1000
|
400
|
400
|
1100
|
|
OAR: organ at risk, PTV: planning target volume, D95: the minimum coverage dose of 95% of the PTV, HI: homogeneity index, CI: conformity index, Dmean: mean dose, Dmax: maximum dose. |
Statistical analysis of dosimetric parameters
For the comparison of treatment plans, the Friedman test was used for analysis of variance (ANOVA), and the Wilcoxon signed-rank test was used for post-hoc analysis with a Bonferroni correction. Differences with p-values of < 0.05 were considered to be statistically significant. Statistical analyses were performed with the SPSS software (Version 26.0; IBM, Armonk, NY).