Introduction:
The increasing complexity of cancer diagnostics and more personalized treatment options,
also in head and neck oncology, require new techniques of patient information processing
and systems to support the decision-making process in the Head and neck tumor board
(HN-TB). For this purpose, a digital patient model of larynx cancer (LC) based on
Bayes ' networks (BN) was developed.
Methods:
The LC model was created according to accepted guidelines and analyses of HN-TBs.
The subnetwork "TNM-State" was successfully validated before. Now the subnetwork "therapy"
has been modeled and evaluated for primary treatment on the basis of 49 LC patient
cases and compared with the HN-TB decisions of the University Hospital Leipzig.
Results:
The LC model contains over 1000 information entities, resulting it the most comprehensive
human-readable BN model to represent a clinical decision. The subnet validation "therapy"
revealed an initial match of 76% of the model computation compared to the therapy
decision in the HN-TB. Further model optimization allowed for an improvement in correct
model prediction.
Conclusions:
These analyses show, as proof of concept, that it is possible to model the therapy
decision of LC on the basis of BN. Personalized medicine and targeted therapy are
of increasing importance in oncologic therapy and require structured and comprehensive
support of information management and decision-making. Further optimization and validation
may allow digital patient models to provide valuable contribution to the diagnostics
and therapy of head and neck carcinomas in the future.
