Inductive Learning of Simple Diagnostic Scores

Abstract

Abstract. Knowledge acquisition and maintenance in medical domains with a large application domain ontology is a difficult task. To reduce knowledge elicitation costs, semi-automatic learning methods can be used to support the expert. We propose diagnostic scores as a promising approach and present a method for inductive learning of diagnostic scores. It can be be refined incrementally by applying different types of background knowledge. We give an evaluation of the presented approach with a real-world case base.

A knowledge-based clinical toxicology consultant for diagnosing single exposures

“…If physicians understand the method by which the system obtains its answers, they are more likely to trust the system and use it within the spectrum of its intended purpose. According to Atzmueller et al [8], "understandability and interpretability of…learned models is of prime importance" and "ideally, the learning method constructs knowledge in the same representation the human expert favors." For this reason, our system makes use of pre-test probabilities (prior probabilities) and LRs, which are commonly used throughout the medical field, and presents its results to the user as a differential diagnosis.…”

“…Automatic system generation was another primary objective. Although Atzmueller et al [8] state that "pure automatic learning methods are usually not good enough to reach a quality comparable to manually built knowledge bases," automatic methods offer certain advantages that should not be overlooked. Automatically trained systems fully bypass the need to interview experts, which is the knowledge acquisition bottleneck of traditional KBS development [11], as well as relieving the knowledge engineer of the burden of acquiring extensive knowledge about toxicology to implement the system.…”