Page 262 - Computational Retinal Image Analysis

260 CHAPTER 13 Drusen and macular degeneration

of the gray-level image intensity values followed by a parametric transformation to
improve the homogeneity of the classification. A fuzzy logic approach was taken
to classifying the uncertain pixels based on local mean intensity values. Given the
coarse segmentation that this achieves, the authors aimed to quantify changes in dru-
sen over time. Patients were followed up over time across two visits 5 years apart, and
the study confirmed clinical qualitative observations that drusen change in number
and size.
3.4.2 Scanning laser ophthalmoscopy
It has been observed in several studies that scanning laser ophthalmoscopy (SLO) is
also an effective imaging modality for viewing drusen. As early as 1995, Kirkpatrick
et al. [74] noted that drusen also appear as characteristic structures in indirect mode
infrared scanning laser ophthalmoscope (SLO) images, with manual counting of
drusen in SLO showing no significant difference from that in standard fundus pho-
tographs for 6 out of 5 patients tested. In 2011, Acton et al. [75] observed similar
findings with a commercial SLO device. They aimed to assess drusen quantification
using retro-mode imaging with the Nidek F-10 scanning laser ophthalmoscope. The
authors captured both stereoscopic fundus photographs and retro-mode images in
31 eyes of 20 patients at varying stages of AMD. The images were independently
assessed by two experienced graders for the number and size of drusen. Drusen were
further assessed in eight of these patients using OCT. Significantly fewer drusen
were observed in fundus photography than in retro-mode, and the predominant de-
posit diameter was on average smaller in retro-mode than in fundus. Agreement be-
tween graders for both types of imaging was substantial for number of deposits and
moderate for size while retro-mode drusen corresponded to OCT in all cases.
A further comparison was made by Diniz et al. [76] aimed to evaluate the ef-
ficiency of drusen detection by scanning laser ophthalmoscopy (SLO) using several
infrared confocal apertures and differential contrast strategies. The authors imaged
11 eyes with non-neovascular AMD using infrared imaging with a Nidek F-10 with
the central, ring, right side (AR) and left side (AL) apertures, both with and without
differential contrast. They also obtained a conventional color fundus photograph for
comparison. In each image, the drusen were manually outlined by two graders. The
number of drusen and total drusen area were calculated and the measures averaged.
Inter-grader reliability was evaluated, with agreement between graders being high.
In terms of manual grading, the addition of differential contrast did not seem to im-
prove drusen detection. The authors found that drusen number and area grades were
significantly higher using right and left apertures. Use of the lateral confocal aperture
may highlight subclinical drusen.

3.4.3 Drusen detection in OCT
Drusen have been found to be visible in OCT as small elevations in the RPE, suggest-
ing potential for this modality in drusen detection and diagnosis. The performance of
the integrated automated analysis algorithms in three SD-OCT devices was evaluated
by Schlanitz et al. [77] for their potential to identify drusen in eyes with early-stage