Machine Learning Methods with Applications to Diagnosis

This chapter extends probabilistic analytical methods introduced in Chap. 2. The principle of transformation of medical information into data visualization is introduced in this chapter. This enables transforming a diverse array of clinical problems into simple, standard forms which lend themselves more easily to diagnostic solutions. Data distillation followed by data visualization allows clinical problems to be formulated into decision making “nodes” which lend themselves to graphical methods of problem solving. This method is exploited widely in computer science and forms the basis for many optimization algorithms used ubiquitously. Candidate diagnosis, differential diagnosis can be selected and evaluated probabilistically. Like Fault Tree Analysis, these methods can be qualitative and quantitative. Illustrative case examples are provided following introduction of the theoretical fundamentals.