Uncertainty and Real-Time Therapy Planning: Incremental Markov-Model Approaches

From its inception, medical artificial intelligence has grappled with the problem of uncertainty. Early systems such as MYCIN [4] and INTERNIST [13] used an informal description of probabilistic knowledge, while more recent efforts [9,17] have taken a more formal approach to capturing the uncertainty. Formal methods provide precise and justifiable reasoning methods to support the conclusions they reach. However, they trade computational efficiency for this gain in precision, and for large problems optimal solutions prove intractable. The problem becomes even more difficult when a system is required to perform under constrained or unpredictable deadlines.