Introduction: Breast density is an important independent risk factor for breast cancer and is negatively associated with diagnostic sensitivity of mammography. Measurement of breast density can be used to identify women at increased risk of developing breast cancer and those who would benefit from additional imaging. However, measurement techniques are generally subjective and do not reflect the true three-dimensional nature of the breast and its component tissues.Method: A semi-automated method for determining the volume of glandular tissue from digitised mammograms has been developed in Manchester. It requires a calibration device (stepwedge) to be imaged alongside the breast during mammography, with magnification markers on the compression paddle to accurately determine breast thickness. Improvements to the design of the stepwedge and markers have enabled the method to be applied to the screening population for the first time. 1,289 women had their volumetric breast density measured using this method and additionally completed a questionnaire on breast cancer risk factors.Results: The method has demonstrated excellent intra- and inter-observer agreement. The median percentage breast density in the study cohort was 8.4% (interquartile range 4.9 – 14.2%). There was no significant difference between left and right breasts; the difference between MLO and CC views was significant (CC view was denser), but values were closely correlated (r = 0.92, p < 0.001). The median glandular volume was 60.1cm3 and exhibited no significant variation between left/right breasts or CC/MLO views. A number of breast cancer risk factors were found to be significantly correlated with glandular volume and percentage breast density, including age, weight, BMI, parity, current HRT use and current smoking. The strength of correlation was equal to or greater than that of visually assessed mammographic density. Glandular volume and percentage breast density measurements demonstrated strong relationships with visually assessed mammographic density, which has been shown to be highly correlated with risk.Conclusions: These findings are promising and suggest that volumetric breast density measured using this method should be associated with breast cancer risk. However, further work is required to establish this relationship directly. The method will be used in a large study, known as PROCAS (Predicting Risk Of Cancer At Screening) which aims to develop individualised breast cancer risk prediction models; these have the potential to form the basis of tailored screening intervals. Preliminary work has been undertaken to adapt the method for full field digital mammography, which suggests that it is possible to use the integrated digital detector as the calibration device.

Scaling of CMOS to nanometer dimensions has enabled dramatic improvement in digital power efficiency, with lower VDD supply voltage and decreased power consumption for logic functions. However, most traditionally prevalent ADC architectures are not well suited to the lower VDD environment. The improvement in time resolution enabled by increased digital speeds naturally drives design toward time-domain architectures such as voltage-controlled-oscillator (VCO) based ADCs. The major obstacle in the VCO-based technique is linearizing the VCO voltage-to-frequency characteristic. Achieving signal-to-noise (SNR) performance better than -40dB requires some form of calibration, which can be realized by analog or digital techniques, or some combination. A further challenge is implementing calibration without degrading energy efficiency performance. This thesis project discusses a complete design of a 10 bit three stage ring VCO-based ADC. A lookup-table (LUT) digital correction technique enabled by the "Split ADC" calibration approach is presented suitable for linearization of the ADC. An improvement in the calibration algorithm is introduced to ensure LUT continuity. Measured results for a 10 bit 48.8-kSps ADC show INL improvement of 10X after calibration convergence.

Scene understanding is a central field in computer vision that attempts to detect objects in a scene and reason about their spatial, functional and semantic relations. While many works focus on things (objects with a well-defined shape), less attention has been given to stuff classes (amorphous background regions). However, stuff classes are important as they allow to explain many aspects of an image, including the scene type, thing classes likely to be present and physical attributes of all objects in the scene. The goal of this thesis is to restore the balance between stuff and things in scene understanding. In particular, we investigate how the recognition of stuff differs from things and develop methods that are suitable to deal with both. We use stuff to find things and annotate a large-scale dataset to study stuff and things in context. First, we present two methods for semantic segmentation of stuff and things. Most methods require manual class weighting to counter imbalanced class frequency distributions, particularly on datasets with stuff and thing classes. We develop a novel joint calibration technique that takes into account class imbalance, class competition and overlapping regions by calibrating for the pixel-level evaluation criterion. The second method shows how to unify the advantages of region-based approaches (accurately delineated object boundaries) and fully convolutional approaches (end-to-end training). Both are combined in a universal framework that is equally suitable to deal with stuff and things. Second, we propose to help weakly supervised object localization for classes where location annotations are not available, by transferring things and stuff knowledge from a source set with available annotations. This is particularly important if we want to scale scene understanding to real-world applications with thousands of classes, without having to exhaustively annotate millions of images. Finally, we present COCO-Stuff - the largest existing dataset with dense stuff and thing annotations. Existing datasets are much smaller and were made with expensive polygon-based annotation. We use a very efficient stuff annotation protocol to densely annotate 164K images. Using this new dataset, we provide a detailed analysis of the dataset and visualize how stuff and things co-occur spatially in an image. We revisit the question whether stuff or things are easier to detect and which is more important based on visual and linguistic analysis.

Die Dissertation untersucht ob partizipative Governance ein effektives Mittel ist um lokale Regierungen in Guatemala dazu zu bewegen ihren Wählern gegenüber verstärkt Rechenschaft abzulegen und den Haushalt mehr an den Bedürfnissen der Armen auszu-richten. Das erste Papier bereitet die wissenschaftlichen Erkenntnisse zur Wirkung von und den Bedingungen für effektive partizipative Governance auf. Das zweite Papier stellt ein neues Verfahren zur Kalibrierung qualitativer Interviewdaten für fuzzy-set Qualitative Comparative Analysis (fsQCA) vor. In einer qualitativ-vergleichenden Ana-lyse von zehn ländlichen Gemeinden untersucht das dritte Papier wie sich effektive par-tizipative Governance, politischer Wettbewerb und Zugang zu lokalen Medien auf die Haushaltsgestaltung einer Lokalregierung auswirken. Das vierte Papier analysiert mit derselben Methode die Bedingungen für effektive partizipative Governance. Das fünfte Papier präsentiert eine Fallstudie zweier Gemeinden und diskutiert Politikoptionen für die Implementierung von partizipativer Governance in Guatemala. Die Ergebnisse der Papiere zeigen, dass effektive partizipative Governance in Kombination mit starkem politischem Wettbewerb zu einer armutsorientierteren Ausrichtung öffentlicher Ausga-ben in den zehn Gemeinden führt, da Wähler besser informiert sind. Jedoch deuten die Ergebnisse auch darauf hin, dass partizipative Governance wegen des geringen Grades zivilgesellschaftlicher Organisation, des niedrigen Bildungsniveaus und hoher Armut in Guatemala nicht effektiv implementiert wird. Partizipative Governance kann also lokale Regierungen dazu bewegen Rechenschaft abzulegen und den Haushalt armutsorientier-ter zu gestalten. Ihre effektive Implementierung wird jedoch in Guatemala lange dauern und einen hohen Ressourceneinsatz erfordern. Daher sollten politische Entscheidungs-träger und Geber auch die Stärkung anderer Informations- und Rechenschaftslegungs-mechanismen, wie der Gemeinderäte, in Betracht ziehen. / This thesis analyses whether participatory governance is an effective means for increas-ing local government accountability and for making local government spending more responsive to the needs of the poor in rural Guatemala. The first paper evaluates the scientific evidence on the impact of and the conditions for effective participatory gov-ernance. The second paper presents a new technique for calibrating qualitative interview data for fuzzy-set Qualitative Comparative Analysis (fsQCA). In a qualitative compara-tive analysis of ten rural Guatemalan municipalities the third paper examines how effec-tive participatory governance, competitive elections, and access to local media influence the allocation of local government spending. The fourth paper analyses the conditions for effective participatory governance with the same empirical method. The fifth paper presents a comparative case study of two municipalities and discusses policy options for implementing participatory governance in Guatemala. Overall, the papers’ findings show that effective participatory governance is sufficient for local government responsiveness in the study area when it is combined with competitive elections, because it increases voter information about local government performance. Yet, the findings also suggest that it will be difficult to implement participatory governance effectively in Guatemala due to the low degree of civil society organization, the low level of education of the population and the high level of poverty. The conclusion drawn from these findings is that effective participatory governance arrangements can make local governments more accountable and responsive, but that it will require much time and resources to implement them. Policy makers and donors should therefore also consider strengthening other information mechanisms, as well as existing accountability mechanisms, such as elected Municipal Councils.