Abstract
Abstract: :
Purpose: To characterize high performance cameras for hyperspectral fundus imaging. Methods: Swine eyes are illuminated using a scanning monochromator coupled into a modified fundus camera. Hyperspectral images are acquired using five different digital cameras each having useful, but generally conflicting characteristics. Cameras are compared based on speed, dynamic range, noise and spatial resolution. The five cameras are chosen such that one of these characteristics is optimized (generally at the expense of the other properties). Performance is quantified by imaging retinal blood vessels filled with both oxygenated blood, and comparing the change in absorption with the signal to noise ratio of the camera, particularly in the wavelength range between 540 and 600nm. Results: Typical results, shown below, compare a Roper VersArray 512B back illuminated CCD with a Basler A601 CMOS camera. Plotted is the image irradiance as a function of position through a cross section of a blood vessel filled with oxygenated blood. Data are shown for 27 distinct wavelengths. Notice that the lower noise Roper camera clearly resolves the slight absorption differences which occur in the green region of the spectrum whereas the SNR of the Basler camera is too low to resolve the difference Conclusions:High performance cameras offer improved retinal images due to reduced noise. This allows image differencing algorithms to operate more aggressively in seeking multispectral retinal features.
Keywords: imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • retina