Innovative 3-D computer vision algorithms that can enable a new generation of """"""""spatially aware"""""""" instruments by providing real-time absolute positioning capability non- invasively, similar to the Global Positioning Satellite System for terrestrial applications, are proposed. They rely on a spatial reference map that is constructed during the diagnostic exploration. The position information can be used for numerous purposes, including surgical navigation, guidance, planning, on-line treatment monitoring, error detection, alarms and safety shutoffs is when the tool strays from the target, change analysis, and even surgical simulation. This is a much better paradigm for instrument design than (the largely unsuccessful) tracking algorithms that measure relative displacements, and are thus prone to drift and tracking loss. The proposed algorithms will operate robustly and accurately at frame rates for extended periods in poor and variable imaging conditions. They could be incorporated into existing clinical instruments without the need for precise calibration, which is often not possible anyway, because the patient's anatomy (e.g., the eye) is part of the imaging system. The algorithms will be validated in the context of laser retinal surgery - compelling as the only long-term proven treatment for the leading blindness- causing conditions affecting over 20 million people in the US. Yet, the current success rate of this procedure is less than 50%, largely due to the lack of spatial mapping and navigation aids in current clinical instruments. Beyond laser retinal surgery, the algorithms may be applied whenever: (1) precise locations on the retina are important (e.g., perimetry); (2) the retinal periphery is of interest (AIDS/CMV, diabetes); (3) motion compensation is needed; (4) a tool such as a laser or endoscope is to be monitored or guided precisely at a chosen location; or 5) even when stable measurements of the vasculature (retinopathy of prematurity) and retinal changes are of interest (e.g., angiogenesis research). Overall, the core methods are expected to be broadly useful in a number of minimally-invasive surgical techniques, including emerging alternatives to laser.
