Benign Paroxysmal Positional Vertigo (BPPV) is one of the most common peripheral causes of vertigo. The diagnosis and treatment of BPPV are simple to perform for experienced specialists. Unfortunately, BPPV is often misdiagnosed by physicians, meaning that many patients are not treated appropriately. One method to reduce the rate of misdiagnosis is to have an automated diagnostic tool that can suggest whether a dizzy patient might be suffering from BPPV. One way of automating the diagnostic part of this procedure is to record videos of the eyes just after the head repositioning movement and to use image processing to detect the presence and direction of an elicited ocular nystagmus. Such a method has become increasingly feasible in the last few years because of the decreasing cost of video-oculography (VOG) systems and the increase in computer processing power. To characterise the nystagmus, it is necessary to extract the horizontal, vertical and torsional components of the eye movements from the recorded videos. The horizontal and vertical components can be easily determined by tracking the position of the pupil; this is performed by most VOG systems. The measurement of the torsional component via VOG, though, has required significant user input: either the semi-invasive addition of coloured markings to the white of the eye, which can then be easily found and tracked; or manual selection of iris features, which are then tracked using cross-correlation. We are investigating the feasibility of measuring the torsional component in a completely non-invasive way. In contrast to currently applied procedures, we are tracking the position of visible iris features that are guaranteed to be stable over time. The initial step is to find the pupil and iris and compensate for any horizontal and vertical eye movements, as well as possible slippage of the system. The iris features are then automatically determined by an algorithm that finds Maximally Stable Volumes (MSVs) in the recorded movie. MSVs are image regions that are darker than points on their border that are also stable in time, meaning that they are guaranteed to exist in consecutive image frames; the concept has been shown to be well suited for robust feature tracking. Preliminary results show that it is feasible to determine the torsional component of eye movements from iris feature tracking. Moreover, the method performs quickly enough that it may be applicable in clinical practice.
|Publication status||Published - 2009|
|Event||Neuroscience 2009 - Chicago, United States|
Duration: 17 Oct 2009 → 21 Oct 2009
|Period||17.10.2009 → 21.10.2009|