Purchase this article with an account.
Jens Kowal, Pascal Dufour, Tobias Rudolph, Sebastian Wolf; Image Fusion System to combine Fluorescence Lifetime Imaging of the Retina with Optical Coherence Tomography Imaging. Invest. Ophthalmol. Vis. Sci. 2013;54(15):5537. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Fluorescence live time imaging (FLIM) in biology is a well-established modality to analyze in a controlled environment metabolic states of individual cells or groups of cells. The metabolic environment of the cells governs the existence of specific fluorophores, whereas each fluorophore features a characteristic autofluorescence live-time that can be assessed using FLIM. However, applying this modality in-vivo to diagnose individual metabolic pathologies of the retina or neurodegenerative diseases is a new field and will open new diagnostic opportunities. The purpose of this work was to develop a system that allows comparing structural changes in the retina imaged by OCT with changes seen in the corresponding autofluorescence live-time image.
Healthy volunteers as well as patients with demonstrated degenerative retinal diseases such as macular degeneration or vascular diseases such as diabetic retinopathy have been screened with both, a prototype fluorescence life-time imaging ophthalmoscope (FLIO) and a regular Heidelberg Engineering FD OCT. The FLIO prototype creates next to the actual autofluorescence decay measurement of two excitation wavelengths (498 - 560 nm, 560 - 720 nm) a scanning laser ophthalmoscope reference image. To automatically co-register the FLIO reference image with the OCT localizer image features have been detected using SURF in each image. Nearest neighbour feature matcher was deployed to establish correspondences between the two feature sets based on Euclidean Distances. Finally RANSAC in combination with a similarity transformation model has been used to find the best consensus set among all corresponding features.
Number of SURF features detected in normal group was found to be gradually higher then in the pathological group (1730/1400). The same applies for reduction rate of the used feature consensus sets (86%/89%) and for the resulting error (3.73/4.03) of both groups.
The developed automatic image fusion system performs equally well in both the healthy as well as the pathologic group. The presented system provides a reliable tool to compare structural changes in the OCT with changes seen in the FLIO images.
This PDF is available to Subscribers Only