Session Item

During the course of RT, patients with head and neck cancer (HNC) may develop significant anatomical changes. Re-planning with adaptive radiotherapy (ART) may ensure adequate dose coverage and sparing of organs at risk. Advanced imaging modalities play an essential role in the customization of the radiation treatment as shown through the use of ART and radiomic. We report the early results of our study which aims to evaluate a machine learning-based radiomic approach to predict outcome and toxicity of HNC patients treated with ART by means of CT, MRI and PET-scan.

Patients with locally advanced HNC treated with radical RT from October 2021 to October 2022 were analyzed. Enrollment is ongoing and it will end after inclusion of 50 patients. The target volumes were delineated on CT scan and adjusted on MRI and FDG-PET. The total dose was 66/60/54 Gy delivered in 30 fractions with VMAT-simultaneous integrated boost technique. For the whole sample, the 95% of the prescribed dose covered at least 95% of the PTV. At week 3 from RT start, CT simulation, MRI and FDG-PET were repeated for re-planning. The new plan started in week 4 (figure 1). We performed the comparison of patients and disease characteristics between original plan and re-simulation (adaptive plan) on CT, MRI and FDG-PET. Univariate analysis was conducted for main patients and disease factors.

At the time of analysis, 30 patients were enrolled. The median age was 69 years. Most patients were males (n=22) and former smokers (n=15). The common site of primary tumor was oropharynx (n=19) and HPV was found in 50% of patients. Stage disease was T4N1 in 47% of cases. All patients completed the radiation treatment as planned. Median GTV-T and GTV-N at baseline and at re-planning were 22cc and 20cc, and 5.7cc and 5.5cc, respectively. Between original plan and adaptive plan, the median difference (Delta) in volume of the parotid gland ipsilateral to the tumor was 3cc and the median Delta of its Dmean was 3.8Gy (figure 2). No relevant changes were described for contralateral parotid. At MRI analysis, the median absolute deviation of Delta-radiomics features was significantly associated with smoke (p<0.05), HPV presence (p<0.005) and primary tumor (p<0.005). At FDG-PET analysis, a significant difference was found for both SUVmax and SUVmean between baseline and interim scan for both GTV-T (p<0.0001) and GTV-N (p=0.0005). The decrease in GTV-T and GTV-N was also statistically significant (p<0.0001 and 0.04). Delta GTV-T showed association with gender (p=0.006), while Delta SUVmean of GTV-T was related to HPV presence  (p=0.035).

The preliminary results of our study highlighted important differences between original RT plan and adaptive plan, combining both ART and radiomic analysis. However, it is necessary to wait for the enrollment closing and an adequate follow-up to evaluate the benefit for survival and toxicity.