Radiotherapy is key in treating head and neck cancer (HNC), but patient responses vary despite similar doses. Routine clinical imaging contains rich, patient-specific data that can be extracted and quantified as radiomics features or image biomarkers (IBMs). This thesis of Yan Li explores how IBMs can predict and quantify toxicities and outcomes, supporting more personalized and effective HNC radiotherapy.
Xerostomia is a common side effect in HNC patients after radiotherapy, mainly due to damage to the salivary glands. This thesis found that lower pre-treatment PET-derived metabolic activity in these glands predicts a higher risk of late xerostomia (dry mouth syndrome). More dedicated salivary-sparing strategies, such as proton therapy, can be prioritized for patients with higher xerostomia risks. Moreover, this thesis found patients with greater post-radiotherapy increases in water diffusion tend to have more severe xerostomia and larger reductions in salivary flow, though this was not statistically significant. Water diffusion, quantified by diffusion-weighted imaging, is more objective than xerostomia questionnaires and the less stable salivary flow measurement. These metrics together can provide a more comprehensive description of salivary gland damage, supporting better subsequent mitigation strategies.
This thesis found that combining clinical variables (e.g., HPV status) with IBMs from the gross tumor volume (GTV) effectively predicted recurrence-free survival and individual lymph node failure risks. These predictions can support optimal treatment strategies. For example, intensive treatment can be considered for patients with high risk of recurrence, and targeted nodal dissection may be applied to high failure risk lymph nodes to reduce the burden of nodal failure.
Yan Li is part of MoHAD.
