Cognitive Measures for Visual Concept Teaching with Intelligent Tutoring Systems

Abstract

This work presents a set of cognitive measures for describing radiological image databases. The measures allow an adequate choice of the next image to be presented to the learner when interacting with an ITS (Intelligent Tutoring System) shell that teaches about visual concepts. The goal is to minimise the learning time of radiological concepts and to permit the implementation of a wider variety of long-term, pedagogic strategies. The literature on concept teaching shows very few attempts to produce computer-based material, particularly on visual concept tutoring. Authoring environments have also been rare, showing that ITS for visual concepts are usually not easy to be re-programmed or expanded. We formalise general-purpose measures for dealing with ordering and complexity of image sequences based on the idea of cognitive load of images and teaching sessions. This load derives from the differences of capacities between novice and expert radiologists, which we summarise from the literature in three major dimensions: (1) visual recognition, (2) diagnostic and (3) verbal expressivity. A case study to exemplify the teaching of brain tumor diagnosis through CT-scans has been carried out with the local University Hospital. An implemented software tool, called SEQUENCE, supports part the authoring process for eliciting the cognitive measures and for setting ordering parameters. We are currently using the cognitive measures in ITS for a range of visual domains, as well as discussing their applicability in a significantly different class of domain, such as computer programming.