Imaging Foveolar cones
Foveal cones are difficult to image due obviously to their small size (down to 2 microns in diameter) and possibly other factors as well. Today, a brief survey of adaptive optics (AO) retinal imaging literature will result in many fantastic images of the human cone mosaic. BUT DON'T BE FOOLED! a more careful inspection would reveal that almost none of the images were acquired at the center of the fovea (foveola), the most important part of the retina responsible for fine spatial vision. In fact, the only images of the smallest foveal cones in the literature were acquired using the flood illuminated AO system at the University of Rochester which uses a completely incoherent light source and an expensive 97 actuator piezo-electric deformable mirror (DM). The image below is a 1 deg by 1 deg patch cropped out of a larger montage centered approximately at the foveal center. This particular image was acquired from a young (24 yrs) emmetropic eye with an axial length of 24.08 mm (6 mm pupil). All the cones are resolved with their positions identified (not shown) for one of our studies. This, at least in my biased opinion, represents an important advancement for AO scanning laser ophthalmoscopy (AOSLO) and the use of MEMS deformable mirrors for retinal imaging. As more sophisticated AO systems for vision research become realized in the future, who knows what other features will be revealed in the retina!
Adaptive optics control
I developed AO control software using a task/state formality as in any finite state machine. TranRunC is a C library developed by Professor David Auslander for facilitating the task/state design when implementing controllers for mechanical system. He describes the library in detail in his book Control Software for Mechanical Systems: Object-Oriented Design in a Real-Time World. Unfortunately when I took Auslander's course in Fall 2007, he switched to C (hence the name TranRunC). His book was published in 2002 and only presents the material using mainly matlab, C++ and Java (TranRunJ). Nevertheless, the principles and some of the code are identical!
American
control conference talk ppt
on/off demo (51 mb)
Retinal layers step through (128 mb)
Undilated
retinal videos for a normal eye (Justin Crepp)
Photoreceptor layer (15 mb)
Blood vessel layer (59 mb)
Cone counting
The image below is processed by our cone counting algorithm. GUIs are still being developed so we are currently sharing our code in Matlab function format only. Brief documentations are given in each function and can be accessed by entering "help function-name" at the Matlab command line if the function is downloaded to your current directory.
AOim2xy - Takes an image in matrix form (uses function imread) and computes the cone location and outputs the X and Y coordinates of each cone as two column vectors of integers. Needs Matlab Image Processing Toolbox. Optimal filtering is not used in this function because it cannot be implemented easily without the Signal Processing Toolbox.
Dynamics of the human tear film (Rochester)
The Hartmann-Shack wavefront sensor has become ubiquitous recently because it has been demonstrated to make reliable wavefront measurements for vision correction. However, it has come into question multiple times in recent papers that the Hartmann-Shack wavefront sensor fails to measure aberrations due to the tear film and should not be used for certain experiments. Our work demonstrated that this is not exactly the case. The progression of tear film degradation can be seen clearly in the raw Shack-Hartmann image data. If such changes in aberrations cannot be seen in the reconstructed wavefront, the method for wavefront reconstruction must not be appropriate for this application. The corresponding video of the wavefront computed using Roddier's iterative Fourier transform reconstructor can be downloaded here. The reconstructed wavefront clearly reflects what can be seen from the raw data alone. Although this reconstructor gets the job done, a more accurate version is now available (see paper by Bahk in Optics Letters here).