An intelligent user interface system for diagnosis of epilepsy

As signal processing algorithms in EEG/MEG (Electroencephalography/Magnetoencephalography) become more advanced, understanding these algorithms and selecting the parameters involved becomes increasingly difficult for users without expertise in signal processing. In this paper, an intelligent system designed to bridge this knowledge gap is proposed, with the goal of helping neurologists select the proper tools for the diagnosis of epilepsy. The interaction with the user occurs in a human-like conversional fashion, and the questions are all constructed with medical terminology familiar to clinician users. The system's knowledge is represented in a pair of description concept matrices (CDMs) and an information-based algorithm is used to select questions by optimizing information-theoretic criteria.