Figure 1a displays a correlation heat map that illustrates the relationship between the estimated NTCP and the skin surface area exposed to various radiation doses (Sx). Throughout the whole dose range, there was a strong correlation between the estimated NTCP and the skin's Sx. The S50 GyRBE and S40 GyRBE have the strongest correlations (0.95 and 0.94, respectively) with NTCP of skin. As shown in Fig. 1b, the NTCP increases nearly linearly with the percentage of skin surface area exposed to 50 Gy RBE and 40 Gy RBE, respectively. As anticipated, the increase in NTCP from 50 Gy RBE was greater than that from 40 Gy RBE for the same surface area of exposure. According to estimates, the rise in skin's NTCP per unit Gy RBE is 0.568 for skin exposed to 50 Gy RBE as opposed to 0.418 for skin exposed to 40 Gy RBE.
The pair plot representation of the three clusters, group-1 (G1), group-2 (G2), and group-3 (G3), formed by the elbo technique used in the K-means clustering algorithm with gEUD, NTCP (%), S40 GyRBE, and S50 GyRBE, respectively, is shown in Fig. 2. The findings show a distinct cluster of patients with different NTCPs, with G-1 showing a lower NTCP, G-2 exhibiting a medium NTCP, and G-3 exhibiting a greater NTCP of skin. There were 12 (27%) patients in G-3, 14 (32%) patients in G-2, and 18 (41%) patients in G-1 clusters.
Table 1 lists the mean±SD of the DSH indices of three clusters. For G1, G2, and G3 groups, the percentage of surface area (mean±SD) receiving 50 Gy RBE was 2.34±3.17, 19.88±5.77, and 36.56±9.43%, respectively. Figure 3 displays the box plot of the EUD (Figure. 3a) and the best fit DSH-based LKB model (Figure. 3b) for the G1, G2, and G3 clusters calculated from the initial 44 patients. The average (±SD) gEUD was 26.54±6.75 Gy RBE for G1, 38.73±1.80 Gy RBE for G2 and 45.67±2.20 Gy RBE for G3 clusters. The corresponding NTCP (%) for G1, G2 and G3 groups were 4.97±5.12, 48.12±12.72 and 87.28±7.73 respectively.
Table 1
The overall mean and standard deviation (SD) of the percentage of skin surface receiving X GyRBE (SXGyRBE) for three clusters (G1, G2 and G3) of patients.
|
Percentage of surface area receiving Sx Gy RBE for three grades of skin toxicity
|
Sx Gy RBE
|
Group-1
|
Group-2
|
Group-3
|
S10
|
67.9 ± 22.14
|
86.03 ± 3.55
|
93.44 ± 6.72
|
S20
|
47.59 ± 19.55
|
70.32 ± 5.06
|
86.3 ± 9.00
|
S30
|
29.29 ± 16.60
|
55.78 ± 6.17
|
75.24 ± 10.0
|
S40
|
13.24 ± 10.48
|
41.55 ± 6.60
|
59.77 ± 8.83
|
S50
|
2.34 ± 3.17
|
19.68 ± 5.77
|
36.56 ± 9.43
|
S60
|
0.10 ± 0.28
|
1.74 ± 2.33
|
8.85 ± 9.63
|
S70
|
0.0 ± 0.00
|
0.02 ± 0.06
|
0.61 ± 1.35
|
3.2 Analysis of skin sparing planning technique
The new planning strategy IMPT-SS, when applied prospectively to the 20 patients, met clinical objectives and produced a dose distribution that was nearly identical to that of the original plans, with the exception of skin dose. The comparison of the dose distribution and dosage difference of two planning strategies is shown in Fig. 4. Although both planning procedures cover the same target area, a considerable dosage differential was seen close to the skin. The comparative dosimetric parameters of targets and OARs were displayed in Tables 3 and 4. Table 3 demonstrates that there is no statistically significant difference between the two planning approaches in the target coverage for any of the CTVs. The D98% for CTV eval was above 98% of the dosage in both groups, and the D95% for CTVs was ≥ 95% of the prescribed dose. Similarly, both strategies were successful in achieving comparable OAR sparing, with the exception skin (P < 0.001), where the new strategy produced a much lower dose (Table 4).
Table 2
The overall mean and standard deviation (SD) of percentage of the surface receiving X Gy (SXGy), equivalent uniform dose (EUD) in Gy RBE and normal tissue complication (NTCP) in percentage in the two techniques, along with the P-values of the wilcoxon signed rank test. P < 0.05 is considered statistically significant.
|
Difference in percentage of skin surface exposed to various dose level (Sx GyRBE) from old and new planning techniques and corresponding gEUD and NTCP (%)
|
Sx(Gy RBE)
|
IMPT
|
IMPT-SS
|
P-value
|
S10
|
92.87 ± 5.8
|
83.05 ± 12.500
|
< 0.001
|
S20
|
82.25 ± 12.50
|
68.23 ± 17.26
|
< 0.001
|
S30
|
69.97 ± 17.26
|
54.44 ± 19.72
|
< 0.001
|
S40
|
56.59 ± 20.01
|
40.70 ± 17.45
|
< 0.001
|
S50
|
38.68 ± 18.01
|
18.22 ± 10.69
|
< 0.001
|
S60
|
5.18 ± 5.40
|
1.51 ± 2.2
|
< 0.001
|
S70
|
0.31 ± 0.65
|
0.02 ± 0.06
|
0.043
|
gEUD
|
44.00 ± 6.95
|
37.404 ± 6.49
|
< 0.001
|
NTCP%
|
76.97 ± 26.83
|
42.73 ± 24.55
|
< 0.001
|
IMPT-SS significantly (p < 0.01) reduced the SX GyRBE, gEUD, and associated NTCP, as indicated in Table 2. The mean NTCP values for IMPT-SS were 37% lower than from IMPT. In the IMPT and IMPT-SS plans, the mean ± SD of NTCP was 76.97 ± 26.83% and 42.73 ± 24.55%, respectively. The Wilcoxon signed-rank test's corresponding p values are shown in Table 2 together with the mean and standard deviation for the radiobiological and DSH indices. Both at higher and lower dosages, the IMPT-SS approach shows significantly less dose to the skin. In the IMPT technique, the median result for S20 GyRBE was 84.0% (range: 49–100%), but in the IMPT-SS approach, it was 68% (range: 22-98.4%). Similarly, the median value for IMPT and IMPT-SS plans was 38% (range: 0–87%) and 17% (range: 0–37%), respectively.
Table 3
The overall mean and standard deviation (SD) dosimetry parameters (D98%, D95%, and D1%) of the CTVs in two treatment planning approaches (IMPT and IMPT-SS), along with the P-values of the wilcoxon signed rank test. P < 0.05 is considered statistically significant.
|
D98%
|
|
P-value
|
D95%
|
|
P-value
|
D1%
|
|
P-value
|
Targets
|
IMPT
|
IMPT-SS
|
|
IMPT
|
IMPT-SS
|
|
IMPT
|
IMPT-SS
|
|
CTV1
|
98 ± 1.030
|
97.86 ± 0.91
|
0.144
|
98.85 ± 0.59
|
98.71 ± 0.72
|
0.366
|
105 ± 0.97
|
105.48 ± 1.54
|
0.456
|
CTV2
|
99.67 ± 0.98
|
98.73 ± 1.91
|
0.059
|
100.53 ± 0.99
|
100.13 ± 1.64
|
0.059
|
105 ± 0.97
|
105.48 ± 1.54
|
0.345
|
Table 4
The overall mean and standard deviation (SD) dosimetry parameters (D1%, Dmean) of the OARs in two treatment planning approaches (IMPT and IMPT-SS), along with the P-values of the wilcoxon signed rank test. P < 0.05 is considered statistically significant.
Parameters
|
D1% (Gy RBE)
|
Dmean (Gy RBE)
|
P-value
|
IMPT
|
IMPT-SS
|
IMPT
|
IMPT-SS
|
D1% (Gy RBE)
|
Dmean (Gy RBE)
|
lips
|
53.2 ± 18.4
|
52.5 ± 17.4
|
24.8 ± 14
|
21.43 ± 11.99
|
0.38
|
0.22
|
Oral cavity
|
64.2 ± 5.2
|
64.4 ± 5.2
|
39.8 ± 12
|
39.67 ± 12.14
|
0.33
|
0.50
|
larynx
|
61 ± 6.90
|
61.3 ± 7.0
|
27.9 ± 10.8
|
26.9 ± 11.67
|
0.38
|
0.35
|
Mucosa
|
57.8 ± 19.8
|
58.3 ± 19.9
|
32.1 ± 13.8
|
31.65 ± 13.63
|
0.31
|
0.46
|
Ipsilateral eye
|
26.3 ± 20.4
|
26 ± 20.20
|
7.7 ± 8.4
|
8.07 ± 8.43
|
0.49
|
0.48
|
contralateral eye
|
6 ± 11.50
|
5.6 ± 10.80
|
1.7 ± 3.2
|
1.59 ± 3.38
|
0.16
|
0.06
|
Ipsilateral Parotid
|
59 ± 15.0
|
59 ± 15.20
|
42.8 ± 17.8
|
41.94 ± 17.81
|
0.34
|
0.44
|
contralateral Parotid
|
39.5 ± 30.3
|
39.3 ± 30.2
|
18.3 ± 15.4
|
17.97 ± 15.11
|
0.45
|
0.36
|
Spinal cord
|
12.7 ± 6.5
|
12.2 ± 6.2
|
3.3 ± 2.3
|
3.51 ± 2.81
|
0.46
|
0.47
|
Brainstem
|
24.8 ± 16.1
|
24.2 ± 16.7
|
7.7 ± 7.5
|
7.86 ± 7.52
|
0.48
|
0.46
|
Ipsilateral optic nerve
|
21.1 ± 22.2
|
20.7 ± 21.8
|
13.7 ± 18.5
|
13.68 ± 18.2
|
0.45
|
0.49
|
Contralateral optic nerve
|
10.2 ± 15.5
|
10.1 ± 15.9
|
4.8 ± 9.7
|
4.48 ± 9.5
|
0.43
|
0.20
|
Chiasm
|
13.1 ± 18.2
|
12.3 ± 17.8
|
7.5 ± 11.8
|
7.01 ± 11.44
|
0.41
|
0.41
|