Conventional imaging workstations employ standard user interface paradigms for interaction, including menu bars, tool palettes, and keyboard strokes. However, efficiency and functionality are sometimes contradictory objectives for a well-designed image viewing station. An alterative method of user interaction, gesture recognition, is demonstrated, with the goal of providing smoother, faster manipulation of images in a manner that is not as disruptive to the primary task of reading/interpreting a study. Users express their desired functionality with pre-defined mouse movements drawn directly onscreen. Thus, the user interface allows the user to remain focused on the study being viewed while providing access to a wide variety of tools. The gesture recognition system is integrated with a Java-based DICOM image viewer, and allows the user to define new gestures through a training interface. A basic linear classifier is used to partition the set of training gestures into groups that correspond to commands.

To explore a new gesture-based paradigm for expressing image workstation functions.