Our preliminary studies represent the first investigations towards the design and implementation of an ophthalmic augmented reality environment. As an analog to the operating microscope, the slit-lamp biomicroscope is perfectly suited to serve as an imaging and display conduit for AR applications. Further, since our application is quasi-two-dimensional, computational demands are far less than in three-dimensional image registration tasks, potentially allowing for efficient real-time performance.
Image registration has played a major role in facilitating augmented reality (AR) applications. As described by Feiner (1993), an AR system utilizes computer generated graphics, such that the virtual world is superimposed on, and enriches, the real world. Whereas virtual reality (VR) is the graphical construction of a synthetic environment, AR extracts information from the real world and augments it (Bowskill and Downie, 1995; Caudell, 1994).
Although there has been an explosive development of investigation into VR applications in medicine, AR applications might have a far greater utility, but have received much less attention (O'Toole et al. 1995). An early study described the superposition of ultrasound images on the abdomen, using a position-tracked, see-through head mounted display (Bajura et al. 1992). Recently, there has been great interest in neurosurgical applications of AR. Specifically, the intraoperative registration of CT, MR, and PET images on a living patient in the operating room may greatly facilitate surgical planning and execution (Gleason et al. 1994; Edwards et al. 1995; Grimson et al. 1996). Typically, registration of the patient with the radiographic data is accomplished by tracking fiducial markers placed on the skin surface and tracking the position of the operating microscope. Our application does not require fiducial markers or image tracking. The real-time image is registered with a previously stored, montaged data set.
We present design considerations for an ophthalmic augmented reality environment, and demonstrate the technical feasibility for each element of the proposed system. In subsequent investigations, angiographic images will be registered off-line to create a three-dimensional vector containing time-dependent angiographic data. Specifically, the individual frames from previously acquired angiographic studies will be registered with each other with high precision, and subsequently registered with the montage. Therefore, the angiographic data will be presented in cine form in the augmented reality environment, offering dynamic visualization of (for example) leakage from a CNVM, diabetic microaneurysm, or central serous retinopathy. Further, image registration and overlay will allow for real-time image comparison to judge disease progression and guide treatment. Finally, technologies permitting interactivity as described above will be incorporated to allow for telemedicine applications.