No abstract

Over the last decade, the process of automatic image colorization has been of significant interest for several application areas including restoration of aged or degraded images. This problem is highly ill-posed due to the large degrees of freedom during the assignment of color information. Many of the recent developments in automatic colorization involve images that contain a common theme or require highly processed data such as semantic maps as input. In our approach, we attempt to fully generalize the colorization procedure using a conditional Deep Convolutional Generative Adversarial Network (DCGAN). The network is trained over datasets that are publicly available such as CIFAR-10 and Places365. The results between the generative model and traditional deep neural networks are compared.

Most traditional image registration algorithms aimed at aligning a pair of images impose well-established regularizers to guarantee smoothness of unknown deformation fields. Since these methods assume global smoothness within the image domain, they pose issues for scenarios where local discontinuities are expected, such as the sliding motion between the lungs and the chest wall during the respiratory cycle. Furthermore, an objective function must be optimized for each given pair of images, thus registering multiple sets of images become very time-consuming and scale poorly to higher resolution image volumes.Using recent advances in deep learning, we propose an unsupervised learning-based image registration model. The model is trained over a loss function with a custom regularizer that preserves local discontinuities, while simultaneously respecting the smoothness assumption in homogeneous regions of image volumes. Qualitative and quantitative validations on 3D pairs of lung CT datasets will be presented.

The mathematical modelling of biological systems has historically followed one of two approaches: comprehensive and minimal. In comprehensive models, the involved biological pathways are modelled independently, then brought together as an ensemble of equations that represents the system being studied, most often in the form of a large system of coupled differential equations. This approach often contains a very large number of tuneable parameters (> 100) where each describes some physical or biochemical subproperty. As a result, such models scale very poorly when assimilation of real world data is needed. Furthermore, condensing model results into simple indicators is challenging, an important difficulty in scenarios where medical diagnosis is required. In this paper, we develop a minimal model of glucose homeostasis with the potential to yield diagnostics for pre-diabetes. We model glucose homeostasis as a closed control system containing a self-feedback mechanism that describes the collective effects of the physiological components involved. The model is analyzed as a planar dynamical system, then tested and verified using data collected with continuous glucose monitors (CGMs) from healthy individuals in four separate studies. We show that, although the model has only a small number (3) of tunable parameters, their distributions are consistent across subjects and studies both for hyperglycemic and for hypoglycemic episodes.

A novel technique of building an anglellength representation of planar polygons is introduced. The object's positional data is slowly acquired by robotic tactile sensors and a neural network is then used to recognise the shape. The method proposed is straight-forward, and seems to work well on simple shapes. It is rotational and shift invariant and can also be made scale invariant.