ESTRO 2024 - Abstract Book

S4417

Physics - Machine learning models and clinical applications

ESTRO 2024

Figure 2

Conclusion:

We present a statistical model to estimate the risk of occult metastases in a LNL, which learns the probability of the tumor to spread to a LNL and the probability of clinically detecting metastases from a dataset of lymph node involvement where both pathological and clinical involvement is reported. The model may inform clinical trials on volume deescalated radiotherapy for HNSCC and eventually lead to further personalized guidelines of the elective nodal CTV. The results in Figure 2 suggest that level IV may be removed from the CTV for clinically N0 patients, and possibly for patient with clinical involvement of level II but not III.

Keywords: HNSCC, lymphatic spread, statistical modelling

References:

1. De Bondt (2007) European Journal of Radiology 64, 266- 272

2. Ludwig (2021) Scientific Reports 11, 12261