Aspects of Optical Broad Band Spectroscopy and Information Extraction - Applications in Medicine and Ecology

Abstract: The present thesis describes a number of aspects of modern electro-optical measurement technology also known as bio-photonics; this includes instrumentation, applications, sample interaction and data interpretation. The methods employed operate over several domains, and light measurements are discretized both in intensity, space, angle, time, polarization and energy. Mainly the spectral domain is investigated over two orders of magnitude from deep ultraviolet to thermal infrared, and mainly broad spectral features in solid and liquid samples are studied. The intensity employed ranges from microwatts to megawatts, time processes are studied between hundred picoseconds to weeks and measurement are carried out from the micrometer scale and up to hundreds of meters. An important aspect of this thesis is the development of realistic instrumentation with the intention that research should benefit the supporting society; this is a key point for the success of academic research in the developing world but also goes hand-in-hand with innovation, commercialization and entrepreneurship in Scandinavia. For this reason the thesis also encompasses a number of patent applications filed during the thesis work. Most of these realistic setups are based on spectroscopy using inexpensive light emitting diodes. Their application for medical diagnosis has been demonstrated with fiber sensors in the context of oncology, and microscopy in relation to parasitology. The thesis also covers optical diagnostics of animal populations of different species on the habitat scale; these studies are pursued by the use of laser radar (lidar) or telescopes. In these areas novel approaches for remotely classifying marked or unmarked flying animals open for the investigation of a new type of questions in field entomology and ornithology. In optical applications for medicine as well as ecology the understanding of the light interaction with complex biological tissue types is essential. Several aspects of such interaction are treated in the thesis. The complex optical interrogation together with the broad and overlapping spectral features in solid samples implies that an empirical approach of data evaluation and computer learning is often more valuable than forward modeling of expected signals. An ongoing theme throughout this thesis is data reduction and chemometrical evaluation. Here discrete light measurements and linear algebra form the basis for advanced statistical evaluation. This applies to the spectral domain where redundancy can be removed, but also topics such as dynamical processes and texture analysis are approached in the temporal and spatial domains, respectively.