Page 61 - Computational Retinal Image Analysis
P. 61

3  Ophthalmic instruments     51


















                  FIG. 24
                  (A) In vivo B-scan image of the human optic nerve showing lamina cribrosa. Size of the
                  image: 4.2 × 3.7 mm. (B) 36 images of human lamina produced simultaneously and in
                  real-time using a MS OCT instrument. (C) Confocal en-face image produced at the same
                  time as the OCT images. The size of each en-face image is 2 × 2 mm. The en-face images
                  are separated by 15 μm. Here only 36 images are presented, however the technology
                  allows more images to be displayed in real-time. All distances are measured in air.


                  resolution than the near-infrared instruments, as well as different spectroscopic
                  contrast hence unique benefits to both fundamental research and clinical care of
                  several eye diseases [62].
                     Optical Coherence Elastography (OCE). Unlike OCT, which uses endogenous
                  contrast mechanisms within a sample, OCE adds an externally induced mechanical
                  stimulus alongside OCT to obtain structural information about the biomechanical
                  properties of the sample. Although not demonstrated yet for real-time in vivo in-
                  vestigations of the human eye, the technique has a huge clinical potential as OCE
                  relies on the fact that the elastic properties of tissues depend on whether they are
                  diseased or not.
                     Real-time OCT (RT-OCT). Speed on data acquisition and processing is cru-
                  cial when imaging moving organs and for instruments equipped with hand-held
                  probes. Although in the last years, in terms of data acquisition, a huge progress
                  has been made with the development of MHz line-rate swept lasers, the process-
                  ing of data is still not sufficiently fast to allow for true real-time display of images.
                  To alleviate the problem of data processing and display, parallel approaches in-
                  volving graphics processing unit (GPU) cards on CUDA parallel computing plat-
                  forms or field programmable gate arrays (FPGAs) are used but they add to the cost
                  of the instrument. Even if the processing time is drastically reduced, such methods
                  are not genuinely real-time, as the brightness of each pixel in the transversal sec-
                  tion is not established simultaneously with the beam being incident on that pixel.
                  To allow for real-time display, methods such as the Master/Slave, which do not
                  add complexity to the instrument is probably a trend that many researchers in the
                  field will follow.
   56   57   58   59   60   61   62   63   64   65   66