Spectral Autofluorescence imaging of the retina for drusen detection

Abstract

Age Related Macular Degeneration (AMD) is a retinal disease that causes central vision loss and is one of the leading causes of blindness in the elderly population. If the disease is diagnosed in the early stages, treatment can help slow progression, delaying retinal degradation and preserving vision. While the symptoms of later stages can be diagnosed with retinal imaging techniques such as fundus cameras and optical coherence tomography, or with visual field tests, the early stages have no direct visible symptoms to be identified. Instead, the onset and progression of the disease can be defined through observation of an associated risk factor, small yellow lipid deposits that form under the retina, known as drusen. By observing the size, location, and distribution of these drusen, the different stages of the disease can be characterised. Specifically, early progression of AMD can be identified by observing hard drusen ( 125 µm). While drusen can be identified with current imaging techniques, proper definition of the type of drusen is challenging due to their small size, and low contrast. However, research by Tong et al., demonstrated a hyper-spectral autofluorescence technique that could isolate the spectral signature of drusen on ex-vivo retinal samples, allowing for accurate demarcation and classification between hard and soft. The aim of this project was to utilise these spectral retinal signatures, developing an in-vivo human multi-spectral imaging device, and investigating the hypothetical increase in specificity and sensitivity for drusen characterisation that it could provide. To test this hypothesis, a multi-spectral autofluorescence (AF) scanning laser ophthalmoscope (SLO) device has been developed, capable of observing the macular region of the retina at a lateral resolution of 20 µm. A 450 nm source was utilised to generate the retinal AF signal, and three spectral AF emission bands are interrogated: 488-532 nm, 532-595 nm, and 595-735 nm. Spectral deconvolution of these bands allows for the localisation of the drusen spectral signature on a retinal image. The viability of the device for drusen detection was assessed in an imaging investigation, demonstrating a documented in-vivo application of a multi-spectral AF SLO. The results show that the device was capable of producing high signal-to-noise ratio (SNR) broad spectrum AF images, but could be unreliable on an inter-patient basis due to an ineffective fixation target, causing system-patient misalignment. At the onset of the project, image quality metrics were defined that would in theory allow for the imaging of drusen on the retina. Demonstrating a 20 µm lateral resolution, a field-of-view encompassing the macular region, and limited image distortion, these metrics were achieved, proving that an accurate representation of the retina could be produced by the system. However, a myopic defocus was identified at increasing field angles due to imperfect objective lenses. The multi-spectral assessment demonstrated low image SNR, due to the splitting of the low intensity AF emission. As such, a conclusive answer on the drusen imaging viability of the technique could not be made. To conclude, this project has resulted in the construction and in-vivo assessment of a multi-spectral AF SLO device, demonstrating that low multi-spectral image SNR is the major challenge for drusen detection. Future development paths were identified to improve the SNR, namely the use of pupil dilation, which would allow for the capture of a larger quantity of retinal AF emission. Furthermore, implementation of a fixation target, additional avalanche photodiodes, and a re-assessment of laser safety, would help improve image quality and throughput. Should these improvements be implemented, and the feasibility of multi-spectral AF imaging be reassessed and confirmed, then the device should be placed in clinical trials to fully assess the specificity and sensitivity against traditional techniques for characterising drusen. This would allow for a final conclusion on the clinical diagnostic value of the technique.