Improved Eye-Movement Recordings for Medical Diagnostics

We believe that the video-based analysis of eye movements could drastically improve the correct diagnosis of frequent, debilitating disorders of the balance system, as well as motor disorders of the eyes. Specifically, the application of video-oculography (VOG) in these areas can provide objective, quantitative data for the description of the pathological symptoms, which are currently not available. While VOG gradually replaces older techniques for the recording of eye movements, existing systems still suffer from the fundamental difficulty of distinguishing between movements of the eyes on the one hand, and movements of the camera with respect to the head on the other hand. This limits the usability of VOG systems, especially for clinical applications where the camera can easily slip (for example during head movements, or with small children). With the proposed research, we want to reduce these artifacts, and make VOG reliable enough so that it can be used for the routine testing of patients suffering from dizziness or from oculomotor pathologies. Results of our preliminary investigations have shown that the carefully controlled utilization of light reflections on the surface of the eye can drastically reduce artifacts that are caused by movements of the camera with respect to the eye. Based on realistic simulations of these reflections we want to incorporate them into more accurate VOG based measurements of eye movements. To improve the accuracy of VOG recordings of eye rotations about the line of sight, we want to use novel analysis techniques to automatically find and track image features which have not been utilized so far. In collaboration with industrial partners and with highly established medical research institutions, we will use these advanced VOG systems to improve the diagnosis of frequently occurring medical disorders, specifically of patients with unilateral vestibular hypo-function or benign paroxysmal positional vertigo. In addition, we want to make video-based eye-movement recordings simple and robust enough that they can also be used for patients with oculomotor disorders.

The primary goal of the project `Improved Eye Movements Recordings for Medical Diagnostics` was to bring video-oculography, i.e. the recording of eye movements with video systems, from the laboratory into clinical practice. Thereby two principal problems had to be overcome: the movement of the video-camera with respect to the head; and the automated measurement of the rotation of the eye about the line-of-sight (this movement is small and involuntary, but provides a great deal of valuable information for the correct diagnosis of a range of diseases of eye and balance system). Our first efforts to eliminate measurement artifacts caused by the relative movement of video camera and head were based on realistic, 3-dimensional simulations of the effects of such movements. Our simulations showed that analytical compensations are feasible for up-down movements of the camera, which can be caused for example by wrinkling of the forehead. However, they cannot compensate arbitrary relative movements. As a result our partners at the LMU in Munich focused on changes in the system design to eliminate such movements mechanically as far as possible. These results were successful, and a new generation of head-mounted video-systems for the measurement of eye movements is currently introduced in clinical practice. The solution of the second problem, the automated measurement of torsional eye movement, required the application of new image processing techniques to the field of eye movement recordings. We showed that the so- called `Method of Maximally Stable Volumes` can solve this problem reliably, for non-uniform illumination as well as for large changes in eye and camera positions. Besides enthusiastic responses of medical doctors to the new light and robust systems for clinical diagnosis, we have also shown that these technologies allow new applications in the field of marketing and human machine interaction.