Clinical Careflows Aided by Uncertainty Representation Models

Choosing an appropriate support for Clinical Decision Support Systems is a complicated task, and dependent on the domain in which the system will intervene. The development of wide solutions, which are transversal to different clinical specialties, is impaired by the existence of complex decision moments that reflect the uncertainty and imprecision that are often present in these processes. The need for solutions that combine the relational nature of declarative knowledge with other models, capable of handling that uncertainty, is a necessity that current systems may be faced with. Following this line of thought, this work introduces an ontology for the representation of Clinical Practice Guidelines, with a case-study regarding colorectal cancer. It also presents two models, one based on Bayesian Networks, and another one on Artificial Neural Networks, for colorectal cancer prognosis. The objective is to observe how well these two ways of obtaining and representing knowledge are complementary, and how the machine learning models perform, attending to the available information.