Modellbasierte dreidimensionale Fundusrekonstruktion: ein genereller Ansatz zur Integration multi-modaler Daten des menschlichen Augenhintergrundes

摘要

A multitude of imaging systems are used to assess different properties of the different tissues in the back of the eye to support the diagnostics of ocular diseases. The correlation of these data is currently only insufficiently supported by software tools. The physician therefore has to rely solely on his anatomical knowledge and his experience when analysing relations between morphological and functional measurements.In the present work, a novel approach to three-dimensional reconstruction of the fundus based on back-projection has been developed. A geometrical model of the eye is used to a priori introduce depth information, that is lost in the process of data generation depending on the modality utilized. This approach implicitly provides the integration of multi-modality data in the framework of a multi-dimensional parameter space.The first step in the reconstruction process is the analysis of the measurement modality. The properties are used for the back projection that utilizes a ray tracing approach to back project the planar image data onto the three-dimensional fundus plane of the eye model. Individual reconstructed parts on the spherical eye model are then aligned to each other by simple rotation and displacement.This novel approach has been used to construct a fundus panorama from a series of images acquired with the identical imaging modality. In addition to the challenge of data integration this poses the problem of dealing with images of peripheral areas of the retina. The spherical shape of the eye plays and increasingly important role for this reconstruction. It could be demonstrated that the developed approach results in greatly improved panoramic wide-angle images of the fundus.In parallel to the three-dimensional reconstruction novel approaches for the quantification of functional parameters have been developed. Especially functional measurements of the posterior part of the eye carry a high relevance since the describe better the consequences to patients sight. The measurement of retinal blood flow from angiographic image sequences has been improved. In addition, a system has been developed to automate the micro-perimetry examination with the scanning laser ophthalmoscope accounting for eye movements. Automatic micro perimetry with fundus control until now was not possible due to involuntary eye movements of the patient and only manual control. A novel fast algorithm has been developed to detect eye movements in real-time in the image data stream generated by the scanning laser ophthalmoscope. If a movement is detected the individual measurement is being discarded and a new testing is initiated. This new system results in a dramatic reduction of effort for the patient and the physician.These functional parameters can also be included into the eye model using the reconstruction approach. They can than be used to a comprehensive analysis. In the present work, the feasibility of the integration of multi-modality data of the human fundus has been demonstrated. Benefit for clinical patient management remains to be shown in subsequent studies in volunteers and patients.