In this study, we set the clinical prescription of clinicians and initial experience of physicists as optimization parameters to get group CP and IE respectively base on preoperative rectal cancer, exploring the feasibility of re-optimizing automatic plans in pinnacle9.1 AP.
Table 3 Table 4 shows that the fitness and uniformity of the T group decreased compared to the CP and IE groups. This difference was more obvious between the CP and T groups, mainly due to the loose conditions and OARs were not fully optimized, so the CP group had better homogeneity and moderation. While the initial conditions of the IE group were generated by the MP program and had undergone a round of manual optimization by physicists, so the dosiological difference in the target regions between them was not distinct. For overall quantification, the difference in HI,CI,GI in both the T and reference groups was within 1%. Figure 1 shows the comparison of the different dose distributions obtained by a patient with the three groups, which shows slight difference in fitness and uniformity among them. After the doctor's evaluation, it was considered difficult to distinguish the difference clinically, and the dose distribution of all the three groups was within a clinically acceptable range.
Table 4
Results of statistical tests for IE-T
Goals
|
IE
|
T
|
z
|
p
|
PGTV,V50,%*
|
96.12
|
96.31
|
-1
|
0.32
|
PCTV,V45,%*
|
96.07
|
95.93
|
2.6
|
0.01
|
PGTV,HI *
|
1.04
|
1.04
|
-2.67
|
0.01
|
PGTV,CI
|
0.80
|
0.81
|
1.02
|
0.31
|
PCTV,HI
|
1.15
|
1.15
|
-3.42
|
< 0.01
|
PCTV,CI *
|
0.86
|
0.86
|
-0.84
|
0.40
|
GI
|
3.56
|
3.53
|
1.67
|
0.10
|
Bowel_small,Dmax,cGy *
|
4802.5
|
4803.6
|
2
|
< 0.05
|
Bowel_small,V30,cGy *
|
20.73
|
19.98
|
4.57
|
< 0.01
|
Bowel_small,Dmean,cGy
|
2275.5
|
2283.7
|
2.34
|
0.02
|
Bladder,V40,%
|
24.08
|
21.52
|
5.11
|
< 0.01
|
Bladder,Dmean,cGy
|
3071.7
|
3076.1
|
3.06
|
< 0.01
|
Femur_L,Dmean,cGy
|
1516.3
|
1372.7
|
5.11
|
< 0.01
|
Femur_R,Dmean,cGy
|
1477.4
|
1357.5
|
4.18
|
< 0.01
|
The bands * in Goals are data that did not fit the normal distribution. |
Figure 2 shows the dosiological differences in OARs between the three groups in a boxtype manner. Compared to the reference groups, T group better protected the normal tissues in the pelvic region, including the bladder, small intestine, and bilateral femoral bones. Such an advantage is more obvious between the CP group, and the reason remains that the conditions of the CP group are based on clinical needs and have more space for further optimization. But even against the IE group based on the clinical experience of physicists, the T group in normal tissues protection was also evident, with the average dose of the femoral head even reduced by 10%. Considering that neoadjuvant chemotherapy is patients before and after radiotherapy, reducing the dose of OARs is important to reduce the occurrence of side reactions and improve long-term quality of survival[23–26]. Moreover, although this study is taking rectal cancer as an example, considering that targeted therapy or immunotherapy combined with radiotherapy will take the mainstream in future treatment of different tumors, how to reduce the side effect of OARs is always the goal of clinical adherence on the basis of rapid planning.
Overall, the T group ensures the coverage, fitness and uniformity of PGTV and PCTV, which effectively reduces the dose of OARs and has more dosisiological advantages, indicating that there is a re-optimization space for the experience-based AP plans.
The disadvantage is that if we try to improve the quality of AP plans through the method of this study, the MP plan must be firstly designed so as to set the reduced results as optimization parameters. So is there a simpler and more feasible way? The answer is sure, and both pinnacle's PLAN IQ and Varian's Rapid Plan(RP) can perform first-step machine learning based on the MP before predicting the results based on new patients. By using these tools, we obtained a simple and feasible way to reoptimize the preoperative rectal cancer AP program.
Despite the advantages of the AP plans, we found deviation values for OARs in this study (see bladder deviation in Fig. 2), so clinicians and physicists should strictly examine their results. In addition, only 35 patients were enrolled in this study, and some of the optimization results did not meet the normal distribution, resulting in the final adoption of the Wilcoxon rank sum tests, and the statistical conclusions were slightly less convincing. We believe that it is necessary to pre-screening patients before designing AP plans, and we will continue this study in later clinical works, increasing the number of cases while improving the methods.