T Rachi*,

• 
  T Rachi

• 
  
• 
  
• 
  
• 
  
• 
  
• 
  
• 
  

PO-GePV-M-104 (Sunday, 7/10/2022)   [Eastern Time (GMT-4)]

ePoster Forums

Purpose: Head and neck volumetric modulated radiotherapy (VMAT) with chemotherapy tends to increase the symptoms of mucositis. In particular, mucositis around the soft palate is often included in the irradiated field and can affect the patient's quality of life (QOL). Replanning termed as adaptive radiotherapy (ART) is performed when QOL is affected. There is no fixed trigger for ART, and in most cases, ART is performed according to the subjective judgment of the physicians. The purpose of this study was to identify quantitative features in the RT Plan in DICOM for the development of mucositis using machine learning. The characteristics in the RT plan make it possible to predict the occurrence of mucositis and support the judgment regarding ART.

Methods: Existing treatment plan data was divided into two groups: 42 patients without ART and 18 patients with ART due to mucositis. Python 3.7.6. was used as the program environment. The irradiation field size and irradiation dose for each VMAT segment were extracted from the RT plan. Subsequently, 660 3D lists were created were created by adding the clinical target volume (CTV) data [field size (cm2), CTV volume (cm3), irradiation dose (Gy)]. These 3D lists were classified by the support vector machine (SVM) with linear regression kernel and the K-nearest neighbor (KNN), and decision boundaries were calculated. The accuracy score was determined using 20% of all data as the test data.

Results: The accuracy scores of SVM and KNN were 0.75 and 0.76, respectively. In other words, the occurrence of mucositis could be predicted with approximately 75% accuracy.

Conclusion: This result suggests that the three elements included in the 3D list can inform the decision of replanning due to mucositis. In the future, it can be considered that creating multidimensional lists with further conditions would improve the accuracy of the prediction.

Funding Support, Disclosures, and Conflict of Interest: This study was suppoted by KAKENHI.

Keywords

Not Applicable / None Entered.

Taxonomy

TH- External Beam- Photons: adaptive therapy

Contact Email