Main Session
Sep
29
PQA 07 - Head and Neck Cancer, Lung Cancer/Thoracic Malignancies, and Nursing and Supportive Care
Presenter(s)
Claire Fitzpatrick, BS - St.Lukes Radiation Oncology Network, Dublin, Dublin
C. Fitzpatrick1, C. Malone1,2, J. Nicholson1,3, C. Skourou1, R. Woods1, and S. M. Brennan III1,4; 1St. Luke's Radiation Oncology Network, Dublin, Ireland, 2Erasmus MC Cancer Institute, University Medical Center Rotterdam, Rotterdam, Netherlands, 3Applied Radiation Therapy Trinity, Discipline of Radiation Therapy & Trinity St James’s Cancer Institute, Trinity College Dublin, Dublin, Ireland, 4Trinity St James’s Cancer Institute, St.James's Hospital Dublin, Dublin, Ireland
Purpose/Objective(s):
To quantify the dosimetric benefit of a personalized margin strategy enabled by Bayesian Neural Network (BNN) predictive modelling. While our BNN predictive PTV model can identify patients with low intrafraction motion with high confidence, we sought to quantify the clinical benefit of margin reduction for these stable patients. We assessed OAR sparing achieved by reducing PTV margins from 3 mm to 2 mm as part of a risk/benefit analysis for using AI predictive models.Materials/Methods:
From 105 SGRT-monitored H&N patients, a BNN (trained to predict intrafraction motion) identified the 10 most stable patients (PTV margins <0.5mm). New plans were generated using 2 mm PTVs and compared against the standard clinical 3mm PTV plans. All planning structures were re-generated from the new 2mm PTVs. Original optimisation objective templates were used. OAR doses were further optimised while maintaining = 0.5% difference in target coverage. Plans were evaluated against institutional OAR constraints. The relevant surrounding OARs were compared with those from the original 3mm PTV plans, with differences >1Gy considered clinically relevant.Results:
Patients were grouped by primary tumour location to account for anatomical similarity: “Lower” (oral cavity, larynx, oro/hypopharynx) and “Upper” (sinonasal). The range in dose reduction across the main OARs per group can be seen in Table 1. Lower: Out of 106 relevant OAR constraints recorded for 7 patients, 74 were reduced by =1.5Gy (69%) and 66 were reduced by =2Gy (62%). Upper: Out of 48 relevant OAR constraints recorded for 3 patients, 29 were reduced by =1.5Gy (60%) and 20 were reduced by =2Gy (42%). Lower volume cases showed consistent OAR dose reduction to the pharyngeal constrictors and parotids, suggesting potential clinical benefit in prevention of dysphagia and xerostomia. A substantial reduction was also noted to the cord which may be of potential benefit in the re-irradiation setting. Upper volume cases were fewer and more complex. Two complex cases overlapped with numerous critical serial OARs showing limited reduction due to Dmax driven constraints. Planner dependent variation was observed in 2 cases. Workflow standardisation (e.g. Rapid plan) aims to reduce variability.Conclusion:
Reducing PTV margins by 1 mm resulted in =1.5 Gy OAR dose reduction in 66% of cases, supporting personalised PTV margins and guiding risk/benefit assessments when deploying AI based decision making tools. While the greatest benefit was observed in ‘lower’ volume H&N patients, further assessment is needed for complex sinonasal cases given the limited number of cases in our study.| Structure | Max Dose Reduction Range (Gy) | Structure | Max Dose Reduction Range (Gy) |
| Lower (n=7) | Upper (n=3) | ||
| Brainstem | 1.1-8.1 | Brainstem | 1.7-6.0 |
| Oral Cavity* | 1.2-4.2 | Chiasm | 1.6-1.7 |
| Oesophagus* | 1.4-8.6 | Optic Nerves | 1.0-1.2 |
| Parotids* | 1.5-4.4 | Eyes-Anterior Chamber | 1.1-4.3 |
| PharynxConst* | 1.0-9.6 | Eyes-Posterior Chamber | 1.2-9.5 |
| Spinal Cord | 1.0-6.7 | Pituitary | 3.7-6.5 |
| *Mean | |||