The Optical Imaging Laboratory has developed a hand-held optical imaging system that is capable of 3D tomographic imaging. However, the imaging system is limited by longer imaging times, and inaccuracy in the positional tracking of the hand-held probe. Hence, the objective is to improve the performance of the imaging system by improving imaging time and positional accuracy. This involves: (i) development of automated single Labview-based software towards near real-time imaging; and (ii) implementation of an alternative positional tracking device (optical) towards improved positional accuracy during imaging. Experimental studies were performed using cubical tissue phantoms (1% Liposyn solution) and 0.45-cc fluorescence target(s) placed under various conditions. The studies demonstrated a 90% reduction in the imaging time (now ~27 sec/image) and also an increase from 94% to 97% in the positional accuracy of the hand-held probe. Performance enhancements in the hand-held optical imaging system have improved its potential towards clinical breast imaging.

Most previous research has explored how words are processed in isolation. However, reading is a complex process where an interplay of various factors affects word identification. Moreover, previous research has mainly focused on alphabetical languages, so extension of the existent findings to non-alphabetical languages is crucial. The current dissertation uses natural reading paradigms to study eye-movements and neurophysiological correlates of the statistical properties of words that affect word recognition during natural reading in English and Chinese. Chapter 2 concerns the time-courses of word frequency and semantic similarity effects in the reading of English derived words. Previous research pointed to a paradox where behavioural experimental techniques showed earlier signatures of these properties than neuro-imaging techniques. By combining eye-tracking and EEG and applying analytical techniques that target the onset of these effects, this study aims at investigating this paradox. Results still show that neurophysiological responses are either largely absent or appear at the same time as shown in eye-movement data. Chapter 3 shows that the existence of spelling errors negatively impacts the recognition of correct spellings in Chinese. This is revealed by the “spelling entropy effect”, which measures the uncertainty about choosing between correct and alternative spelling variants. This is the first study that used co-registration of eye-tracking and EEG to explore the behavioral and neurophysiological signatures of this uncertainty. Chapter 4 studies how segmentation probabilities influence word segmentation and identification when reading Chinese. The results reveal that space becomes beneficial only when located at places where segmentation probability is considered high. This study is among the first to show beneficial effects of spacing at the sentence level and demonstrates how segmentation probabilities play a crucial role in Chinese word segmentation. Cumulatively, the results obtained point to the existence of numerous factors involved in word identification in both alphabetic and logographic languages, which should be explored using natural reading experimental paradigms, such as co-registration of EEG and eye-tracking, for obtaining a multifaceted view of word recognition processes. / Thesis / Doctor of Philosophy (PhD)

The use of complementary non-invasive imaging modalities has been proposed to track disease progression, particularly cancer, while simultaneously evaluating therapeutic efficacy. A major obstacle is a limited ability to compare parameters obtained from different modalities, especially those from exogenous contrast agents or tracers. We hypothesize that combining Magnetic Resonance Imaging (MRI) and Ultrasound (US) will improve characterization of the tumour microenvironment. In this study, we describe a co-registration apparatus that facilitates the acquisition of a priori co-registered MR and US images in vivo. This apparatus was validated using phantom data and it was found that the US slices can be selected to an accuracy of +/- 100µm translationally and +/- 2 degrees rotationally. Additionally, it was shown that MRI and US may provide complimentary information about the tumour microenvironment, but more work needs to be done to assess repeatability of dynamic contrast enhanced MRI and US.

Background: The objective of this study was to verify the effect that plot locational errors, termed plot co-registration errors, have on the strength of regression between LiDAR canopy metrics and the measured total standing volume (TSV) of plots in a Pinus radiata forest. Methods: A 737 hectare plantation of mature Pinus radiata located in Northern Hawkes Bay was selected for the study. This forest had been measured in a pre-harvest inventory and had aerial LiDAR assessment. The location of plots was verified using a survey-grade GPS. Least square linear regression models were developed to predict TSV from LiDAR canopy metrics for a sample of 204 plots. The regression strength, accuracy and bias was compared for models developed using either the actual (verified) or the incorrect (intended) locations for these plots. The change to the LiDAR canopy metrics after the plot co-registration errors was also established. Results: The plot co-registration error in the sample ranged from 0.7 m to 70.3 m, with an average linear spatial error of 10.6 m. The plot co-registration errors substantially reduced the strength of regression between LiDAR canopy metrics and TSV, as the model developed from the actual plot locations had an R2 of 44%, while the model developed from the incorrect plot locations had an R2 of 19%. The greatest reductions in model strength occurred when there was less than a 60% overlap between the plots defined by correct and incorrect locations. Higher plot co-registration errors also caused significant changes to the height and density LiDAR canopy metrics that were used in the regression models. The lower percentile elevation LiDAR metrics were more sensitive to plot co- registration errors, compared to higher percentile metrics. Conclusion: Plot co-registration errors have a significant effect on the strength of regressions formed between TSV and LiDAR canopy metrics. This indicates that accurate measurements of plot locations are necessary to fully utilise LiDAR for inventory purposes in forests of Pinus radiata.

Sub-pixel image alignment estimation is desirable for co-registration of objects in multiple images to a common spatial reference and as alignment input to multi-image processing. Applications include super-resolution, image fusion, change detection, object tracking, object recognition, video motion tracking, and forensics.Information theoretical measures are commonly used for co-registration in medical imaging. The published methods apply Shannon's Entropy to the Joint Measurement Space (JMS) of two images. This work introduces into the same context a new set of statistical divergence measures derived from Fisher Information. The new methods described in this work are applicable to uncorrelated imagery and imagery that becomes statistically least dependent upon co-alignment. Both characteristics occur with multi-modal imagery and cause cross-correlation methods, as well as maximum dependence indicators, to fail. Fisher Information-based estimators, together as a set with an Entropic estimator, provide substantially independent information about alignment. This increases the statistical degrees of freedom, allowing for precision improvement and for reduced estimator failure rates compared to Entropic estimator performance alone.The new Fisher Information methods are tested for performance on real remotely-sensed imagery that includes Landsat TM multispectral imagery and ESR SAR imagery, as well as randomly generated synthetic imagery. On real imagery, the co-registration cost function is qualitatively examined for features that reveal the correct point of alignment. The alignment estimates agree with manual alignment to within manual alignment precision. Alignment truth in synthetic imagery is used to quantitatively evaluate co-registration accuracy. The results from the new Fisher Information-based algorithms are compared to Entropy-based Mutual Information and correlation methods revealing equal or superior precision and lower failure rate at signal-to-noise ratios below one.

The use of complementary non-invasive imaging modalities has been proposed to track disease progression, particularly cancer, while simultaneously evaluating therapeutic efficacy. A major obstacle is a limited ability to compare parameters obtained from different modalities, especially those from exogenous contrast agents or tracers. We hypothesize that combining Magnetic Resonance Imaging (MRI) and Ultrasound (US) will improve characterization of the tumour microenvironment. In this study, we describe a co-registration apparatus that facilitates the acquisition of a priori co-registered MR and US images in vivo. This apparatus was validated using phantom data and it was found that the US slices can be selected to an accuracy of +/- 100µm translationally and +/- 2 degrees rotationally. Additionally, it was shown that MRI and US may provide complimentary information about the tumour microenvironment, but more work needs to be done to assess repeatability of dynamic contrast enhanced MRI and US.

For autonomous vehicles, 3D, rotating LiDAR sensors are critically important towards the vehicle's ability to sense its environment. Generally, these sensors scan their environment, using multiple laser beams to gather information about the range and the intensity of the reflection from an object. For multi--LiDAR systems, the placement of the sensors determines the density of the combined point cloud. I perform preliminary research on the optimal LiDAR placement strategy for an off--road, autonomous vehicle known as the Halo project. I use simulation to generate large amounts of labeled LiDAR data that can be used to train and evaluate a neural network used to process LiDAR data in the vehicle. The performance metrics of the network are then used to generalize the performance of the sensor pose. I also, describe and evaluate intrinsic and extrinsic calibration methods that are applied in the multi--LiDAR system.

No description available.

Die in den letzten Jahrzehnten aufgenommenen Satellitenbilder zur Erdbeobachtung bieten eine ideale Grundlage für eine genaue Langzeitüberwachung und Kartierung der Erdoberfläche und Atmosphäre. Unterschiedliche Sensoreigenschaften verhindern jedoch oft eine synergetische Nutzung. Daher besteht ein dringender Bedarf heterogene Multisensordaten zu kombinieren und als geometrisch und spektral harmonisierte Zeitreihen nutzbar zu machen. Diese Dissertation liefert einen vorwiegend methodischen Beitrag und stellt zwei neu entwickelte Open-Source-Algorithmen zur Sensorfusion vor, die gründlich evaluiert, getestet und validiert werden. AROSICS, ein neuer Algorithmus zur Co-Registrierung und geometrischen Harmonisierung von Multisensor-Daten, ermöglicht eine robuste und automatische Erkennung und Korrektur von Lageverschiebungen und richtet die Daten an einem gemeinsamen Koordinatengitter aus. Der zweite Algorithmus, SpecHomo, wurde entwickelt, um unterschiedliche spektrale Sensorcharakteristika zu vereinheitlichen. Auf Basis von materialspezifischen Regressoren für verschiedene Landbedeckungsklassen ermöglicht er nicht nur höhere Transformationsgenauigkeiten, sondern auch die Abschätzung einseitig fehlender Spektralbänder. Darauf aufbauend wurde in einer dritten Studie untersucht, inwieweit sich die Abschätzung von Brandschäden aus Landsat mittels synthetischer Red-Edge-Bänder und der Verwendung dichter Zeitreihen, ermöglicht durch Sensorfusion, verbessern lässt. Die Ergebnisse zeigen die Effektivität der entwickelten Algorithmen zur Verringerung von Inkonsistenzen bei Multisensor- und Multitemporaldaten sowie den Mehrwert einer geometrischen und spektralen Harmonisierung für nachfolgende Produkte. Synthetische Red-Edge-Bänder erwiesen sich als wertvoll bei der Abschätzung vegetationsbezogener Parameter wie z. B. Brandschweregraden. Zudem zeigt die Arbeit das große Potenzial zur genaueren Überwachung und Kartierung von sich schnell entwickelnden Umweltprozessen, das sich aus einer Sensorfusion ergibt. / Earth observation satellite data acquired in recent years and decades provide an ideal data basis for accurate long-term monitoring and mapping of the Earth's surface and atmosphere. However, the vast diversity of different sensor characteristics often prevents synergetic use. Hence, there is an urgent need to combine heterogeneous multi-sensor data to generate geometrically and spectrally harmonized time series of analysis-ready satellite data. This dissertation provides a mainly methodical contribution by presenting two newly developed, open-source algorithms for sensor fusion, which are both thoroughly evaluated as well as tested and validated in practical applications. AROSICS, a novel algorithm for multi-sensor image co-registration and geometric harmonization, provides a robust and automated detection and correction of positional shifts and aligns the data to a common coordinate grid. The second algorithm, SpecHomo, was developed to unify differing spectral sensor characteristics. It relies on separate material-specific regressors for different land cover classes enabling higher transformation accuracies and the estimation of unilaterally missing spectral bands. Based on these algorithms, a third study investigated the added value of synthesized red edge bands and the use of dense time series, enabled by sensor fusion, for the estimation of burn severity and mapping of fire damage from Landsat. The results illustrate the effectiveness of the developed algorithms to reduce multi-sensor, multi-temporal data inconsistencies and demonstrate the added value of geometric and spectral harmonization for subsequent products. Synthesized red edge information has proven valuable when retrieving vegetation-related parameters such as burn severity. Moreover, using sensor fusion for combining multi-sensor time series was shown to offer great potential for more accurate monitoring and mapping of quickly evolving environmental processes.

The Gamma Knife is a non-invasive stereotactic radiosurgery system used for treatments of deep targets in the brain. Accurate patient positioning is needed for precise radiation delivery to the target. The two latest versions of the Gamma Knife allow fractionated treatment by co-registering Cone-beam computed tomography (CBCT) images of the patient's position in the Gamma Knife with a diagnostic magnetic resonance (MR) image used for treatment planning. However, CBCT images often suffer from artifacts that degrade image quality, which may result in less accurate co-registration. This thesis project investigates the potential of a new reconstruction framework developed by Elekta, which incorporates scattering correction and noise filters, for the reconstruction of Gamma Knife CBCT images. The performance of the new reconstruction framework, along with its noise filter and scatter correction, is quantified using image quality metrics of phantoms, including contrast, uniformity, spatial resolution, and CT-number accuracy. Additionally, brain CBCT images of five patients are co-registered with their diagnostic MR images, and the mean target registration error is measured. The results indicate that the new reconstruction framework, without using scatter correction and noise filtering, performs equally well as the current framework in reconstructing Gamma Knife CBCT images, as it achieved similar image quality and co-registration accuracy. However, when the scatter correction was used, there were improvements in image uniformity and CT-number accuracy without compromising spatial resolution. Additionally, the introduction of a noise filter resulted in an improved contrast-to-noise ratio and low contrast visibility with minimal compromise of spatial resolution. Despite these image quality enhancements, there were no consistent improvements in co-registration accuracy, indicating that the co-registration is not sensitive to scatter or noise artefacts. / Gammakniven är en medicinteknisk apparat som används för icke-invasiv stereotaktisk strålkirurgi vid behandling av djupa mål i hjärnan. För att uppnå precision i strålbehandlingen krävs noggrann patientpositionering. De två senaste versionerna av Gammakniven tillåter fraktionerad behandling genom att co-registrera cone-beam computed tomography (CBCT)-bilder av patientens position i Gammakniven med en diagnostisk magnetresonans (MR)-bild som används för behandlingsplanering. Tyvärr lider CBCT-bilder ofta av artefakter som kan försämra bildkvaliteten och därmed minska precisionen i co-registreringen. Detta examensarbete undersöker ett nytt rekonstruktionsramverk som utvecklats av Elekta. Det nya rekonstruktionsramverket och dess tillhörande brusfilter och spridningskorrigering utvärderas för rekonstruktion av Gammaknivens CBCT bilder med hjälp av bildkvalitetsmått för fantomer, såsom kontrast, uniformitet, spatial upplösning och noggrannhet i CT-nummer. Dessutom co-registreras CBCT-bilder från fem patienter med deras diagnostiska MR-bilder, och det genomsnittliga registreringsfelet mäts. Resultaten visar att det nya rekonstruktionsramverket, utan användning av spridningskorrigering och brusfiltrering, presterar lika bra som det nuvarande ramverket för rekonstruktion av CBCT-bilder från Gammakniven. Båda ramverken ger liknande bildkvalitet och noggrannhet i co-registreringen av bilderna. Vid användning av spridningskorrigering observerades förbättringar i uniformiteten och noggrannheten i CT-nummer utan att den spatiala upplösningen försämrades. Införandet av brusfilter resulterade i ett förbättrat kontrast-brus-förhållande och synlighet av svaga kontrastskillnader med endast lite avkall på den spatiala upplösningen. Trots dessa förbättringar i bildkvaliteten observerades ingen konsekvent förbättring av noggrannheten i co-registreringen av bilderna, vilket tyder på att co-registreringen inte påverkas av spridnings- eller brusartefakter i stor utsträckning.