Purchase this article with an account.
A. Dietzel, J. Haueisen; Coregistration of Fluorescence Lifetime Imaging (FLIM) and Optical Coherence Tomography (OCT). Invest. Ophthalmol. Vis. Sci. 2010;51(13):2290.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
We aim to combine information between cross-sectional images (OCT) and metabolic data (FLIM).
The FLIM measurement at the human fundus was accomplished by a modified Heidelberg Retina Angiograph. Two spectral channels were used for auto-fluorescence detection with application of time-correlated single photon counting. A 30° region of the fundus was covered with a lateral resolution of 40 x 40 µm² resulting in one fluorescence intensity (FI) image per channel (150 x 150 pixel, up to several 1.000 gray scales). The cross-sectional images of the retina were acquired by a Heidelberg Spectralis OCT. For each scan an infrared image (1536 x 1536 pixel, 256 gray scales) was used to compensate eye-movements which covers a 30° area at the fundus. It was saved along with the cross-sectional scans and their positions on the infrared image. In the framework of a reproducibility study, which included six healthy male subjects (age 25 to 35), each subject was measured ten times at the fluorescence lifetime mapper and once at the Spectralis OCT. Since the nature of both image data is very different, a landmark based non-rigid registration process was designed. In each modality, bifurcations of vessels were used as landmarks. These bifurcations were extracted by using standard morphological operators. Two feature matrices were built based on the distances and angles between all bifurcations. With help of the two feature matrices a rigid global transformation was performed based on RMSD, and then a deformation field was obtained by interpolation.
Experimental results showed that an acceptable level of accuracy can be achieved to register OCT images and FI images. The mean error between all corresponding bifurcations was less than 2 pixels for more than 71% of all 120 (10 images per 6 subjects per 2 channels) registration examples. 15% of the FI images could not be registered due to insufficient image quality. The remaining 14% exhibited bad registration results for parts of the image which originated from doubled retinal vessels. Those effects were caused by eye- and head-movements.
The demonstrated combination of structural and functional information might improve future diagnostic approaches. We expect supplementary information.
This PDF is available to Subscribers Only