Clinical decision support models for oropharyngeal cancer treatment: design and evaluation of a multi-stage knowledge abstraction and formalization process

• DOI: 10.1007/s11548-022-02675-3
• Publication Date: 2022-06-03T05:02:36Z
• Authors: Jan Gaebel; Stefanie Mehlhorn; Alexander Oeser; Andreas Dietz; Thomas Neumuth; Matthaeus Stoehr
• Keywords: Abstraction

Treatment decisions in oncology are demanding and affect survival, general health, quality of life. Expert systems can handle the complexity oncological field. We propose application a hybrid modeling approach for decision support models consisting expert-based implementation model structure machine-learning (ML) based parameter generation. demonstrate our treatment oropharyngeal cancer.We created clinical on Bayesian Networks iteratively optimized its characteristics using structured knowledge engineering approaches. combined manual adaptation individual concepts with automatic learning parameters causalities. Using data from 94 patient records, we targeted needed objectivity significance.In three iteration steps, assessed cross-validations. The initial aggregated accuracy 0.529 could be increased to 0.883 final version. predictive rates target nodes range 0.557 0.960.Combining different methodological approaches requires balancing subject matter amount information available dataset ML application. Our method showed promising results because flaws one overcome by other approach. However, technical integrability as well acceptance must always ensured.