Room: Karl Dean Ballroom C
Purpose: Many tumors including head&neck squamous cell carcinoma (HNSCC) spread along the lymphatic network. Current imaging modalities can only detect sufficiently large metastases. Therefore, adjacent lymph node levels (LNL) are irradiated electively since they may harbor microscopic tumor. We apply Bayesian Networks (BN) to model lymphatic tumor progression. The model can subsequently be used to personalize the risk estimation of microscopic lymph node metastases in a newly diagnosed patient based on his/her distribution of macroscopic metastases.
Methods: A BN is a graphical representation of a joint probability distribution. We represent LNLs by binary random variables corresponding to the BN nodes. We connect two nodes by directed arcs if the tumor can spread from one LNL to the next. Arcs are associated with conditional probabilities, whose values are learnt from published clinical studies using maximum likelihood estimation. In addition, the primary tumor is represented by input nodes, and the observations (e.g. PET-CT imaging, FNA) are represented by output nodes. We demonstrate the concept for low-grade oropharyngeal carcinomas and their spread to ipsilateral lymph node levels Ib to IV.
Results: We demonstrate that the BN parameters can be efficiently learnt by merging pathology findings on microscopic tumor progression (which is limited to few published studies) and imaging data on macroscopic tumor progression (which is widely available in clinical practice). The trained network can be used to quantify how the distribution of macroscopic metastasis impacts the probability of microscopic involvement of the remaining LNLs. For example, the risk of level IV involvement increases from 2% for N0-patients to 7% when levels II/III harbor metastases, which would alter the inclusion of level IV in the CTV if a 5% threshold was used.
Conclusion: We present a statistical model of lymphatic tumor progression, which may be used to personalize elective CTV definition in head&neck cancer.
Decision Theory, Bayesian Statistics, Treatment Planning
IM/TH- Image Analysis (Single modality or Multi-modality): Computer-aided decision support systems (detection, diagnosis, risk prediction, staging, treatment response assessment/monitoring, prognosis prediction)