*5.2.1 Application of radiomics to selection of NPC cases for dose escalation*

Radiomics refers to the extraction of features in the regions of interest (ROI) from medical images [86]. The extracted features can be the image voxel intensity, ROI texture and shape features, etc. [87]. These extracted radiomics features can be used to correlate with clinical data such as recurrence and metastasis status of patients, so as to develop tools for predicting treatment outcome in future patients based on individual patients' image radiomics features. Research articles have been published to evaluate the chance of local recurrence in NPC patients, and it was reported that local recurrence can be predicted using pre-treatment imaging with a concordance index of over 0.8 [88, 89]. The future direction could be to incorporate radiomics study for more accurate and individualized patient selection instead of based on their staging. With the attempt to generate own local recurrence prediction model based on radiomics features, NPC patients indicated for GTV dose escalation could be more accurately identified.

### **6. Summary**

Radiotherapy is necessary for the treatment of various head and neck cancers either as a primary treatment or adjuvant treatment after surgery to cure the disease. To achieve optimal radiotherapy treatment, we need to understand the rationale of IMRT and the procedure of treatment planning. With the help of treatment planning computer, inverse planning procedure can accomplish treatment plans with highly conformal radiation dose to PTV and dose avoidance from OARs. Because of the conflicting nature of the 2 major dosimetric goals: high PTV dose and low nearby OARs dose, the optimal radiotherapy treatment is usually achieved by experienced planners who are able to carefully balance the trade-off between the conflicting goals. Nevertheless, the present development of knowledge-based planning could provide a guidance for planners to decide the trade-off in a more objective manner. In addition, the development of 4-pi VMAT and research of radiomics may strengthen the advantage of IMRT in terms of OARs sparing and tumor dose escalation.

*Dosimetry*
