Vorschau

PDF, Englisch (PhD Thesis Nikolas Dimitriadis) - Hauptdokument Download (49MB) | Nutzungsbedingungen

Abstract

Fluorescence guided surgery supports doctors by making unrecognizable anatomical or pathological structures become recognizable. For instance, cancer cells can be targeted with one fluorescent dye whereas muscular tissue, nerves or blood vessels can be targeted by other dyes to allow distinction beyond conventional color vision. Consequently, intraoperative imaging devices should combine multispectral fluorescence with conventional reflectance color imaging over the entire visible and near-infrared spectral range at video rate, which remains a challenge. In this work, the requirements for such a fluorescence imaging device are analyzed in detail. A concept based on temporal and spectral multiplexing is developed, and a prototype system is build. Experiments and numerical simulations show that the prototype fulfills the design requirements and suggest future improvements. The multispectral fluorescence image stream is processed to present fluorescent dye images to the surgeon using linear unmixing. However, artifacts in the unmixed images may not be noticed by the surgeon. A tool is developed in this work to indicate unmixing inconsistencies on a per pixel and per frame basis. In-silico optimization and a critical review suggest future improvements and provide insight for clinical translation.