Abstract
The notion of sparsity, namely the idea that the essential information contained in a signal can be represented with a small number of significant components, is widespread in signal processing and data analysis in general. Great progress for example in image compression and enhancement has been obtained by modeling signals as sparse in an appropriate domain. This success in applications has also been accompanied by tremendous theoretical developments. Recently, new sparsity models, new reconstruction algorithms and new theoretical insights into the sparse representation problem have been presented, leading to a new wave of activities in the area of sparse approximation and representation This special session touches on all the aspects of these new developments with contributions from some of the best researchers in this field.

See programme details for Sparsity 2.0: New Trends in Sparse Approximation, Sparsity-based Signal Models and Algorithms

Abstract
In recent years, the statistical community has become increasingly engaged in developing methodology for imaging data. Statistical and model-based methods for imaging have advantages in terms of characterizing uncertainty through probabilistic and likelihood-based models. In addition, model-based approaches can potentially provide an automated alternative to hand-designed features, automatically learning dictionaries to sparsely characterize the image and predict health outcomes. This session highlights recent developments in statistical methods for imaging, with a focus on dictionary learning and probabilistic modeling approaches, including Bayesian methods.”

See programme details for Statistical Methods in Imaging

Abstract
Optimization algorithms are being developed by researchers and used by practitioners in a growing number of fields, including engineering, signal processing, machine learning, applied mathematics, computer science, biology, astronomy and operations research. However, new applications bring about new challenges, the most prominent being scalability to problems of growing sizes. This session brings together several leading researchers in the area of scalable optimization algorithms, who will touch on topics such as proximal primal-dual methods, bilevel optimization, alternating minimization, iteratively reweighted least squares, semi-stochastic gradient descent (S2GD), and incremental majorization-minimization methods. Applications discussed will include image reconstruction and large-scale machine learning and signal processing.

See programme details for Modern Scalable Algorithms for Convex Optimization

Abstract
After three decades of relatively slow progress, radio interferometric imaging is suddenly experiencing a period of rapid innovation. This is driven by the new generation of the “Square Kilometre Array (SKA) pathfinder” telescopes, as well as the prospect of the SKA itself, for which it is clear that imaging will be a major performance bottleneck. Multiscale approaches based on the traditional CLEAN algorithm, as well as completely new algorithms based on compressive sensing and Bayesian formulations, have recently been proposed and implemented. This session will bring together many of the researchers responsible for this progress, who will discuss further developments in the field.

See programme details for Radio interferometric deconvolution and imaging techniques, from CLEAN to CS to Bayesian

Abstract
Data analysis in astronomy faces many challenges, at every stage from the construction of images, through the inference within the context of physical models, to the mining of increasingly large datasets. At all of these stages, astronomers have been responsible for introducing novel and innovative techniques, from the widespread use of Bayesian inference modelling, including hierarchical models and marginalised likelihood for model selection, through to the use of machine-learning techniques for data mining. In this session, we will explore many facets of the use of statistics in astronomy, illustrating with a number of case studies, and showing how the techniques may be applicable to alternative imaging and other fields.

See programme details for Astrostatistics: Bayes and machines

Abstract
This session will present the state-of-the-art in signal processing techniques in observational cosmology. The speakers will collectively cover aspects from the entire currently observable Universe, from the Cosmic Microwave Background, through the Epoch of Reionisation, to late Universe observations of Large Scale Structure and Weak Gravitational Lensing. Recent observations will be presented from new experiments including the optical ESO KiDS survey, the CMB ACT survey and SDSS.

See programme details for Signal Processing in Cosmology

Abstract
Medical imaging, both for diagnosis and image-guided therapies, has seen tremendous progress and growth.  This session will cover the recent advances in multiple modalities and give participants insight into unsolved problems that limit further advances.  Modalities covered will include MRI, tomosyntheses, interventional devices, PET-MR, and focused ultrasound.

See programme details for Successes and Opportunities in Medical Imaging

Abstract
Biomedical imaging especially diagnostic imaging has traditionally focused on generating and visualizing image contrast. Novel signal encoding and processing techniques have started to push the boundaries and offer additional information that may not be available immediately. These techniques may provide quantitative assessment of tissue properties, probe tissue microstructures, decode brain connectivity or enhance image resolution. This session showcases examples of such advancements in areas of MRI, ultrasonic imaging and optical imaging with presentations from some leading researchers in the respective areas.

See programme details for Neuro and Quantitative Imaging

Abstract
Medical imaging signals in modalities including MRI and CT, are often collected along multiple dimensions, for instance, the spatial, temporal, and spectral dimensions. Recently, additional degrees of freedom in the form of multiple receiver arrays, new spatial encoding spaces, and parametric dimensions have also been exploited to enrich the information content of medical imaging signals. By understanding the properties of the signals along each these dimensions, their relationships to one another, and any redundancies in content, medical imaging can be accelerated or enhanced, and used to obtain diagnostic information that cannot be collected in other ways. This session seeks to examine several recently proposed medical signal encoding and reconstruction techniques which take advantage of multidimensional signals to improve medical imaging. Both similarities across different imaging modalities as well as modality-specific capabilities will be highlighted in order to facilitate cross-modality transfer of ideas.

See programme details for Rapid and Multidimensional Imaging