Abstracts
Lourdes Agapito: Estimating Human 3D Pose from a Single Image
Simon Arridge: Compressed Sensing and Dynamic PhotoAcoustic Tomography
Angelica Aviles-Rivero:
The Complexity of Motion in Medical Imaging: From Variational Methods to Machine Learning
Abhir Bhalerao: Modelling and Learning Appearance Variation for Applications in Medical Imaging
Shape and appearance models have been used extensively for image segmentation in medical imaging for over two decades. This talk will present a new type of active appearance model which allows accurate and compact modelling of deformable anatomy. They are built with sparse landmark and wavelet features and can be used to generate samples from the training data. We show how such models can be used not only for segmentation but also for classification. We go on to demonstrate how landmark-free appearance models can built for pin-pointing evidence supporting a diagnosis with only weak supervision.
Joanna Collingwood: Imaging trace metals in the human brain
Iron and other metals can inflence MRI parameters, creating scope for quantitative clinical imaging. However, obtaining specific information is very challenging. By contrast, XRF imaging has a high degree of specificity, but cannot be use in the clinic, only with postmortem specimens. In this talk I will introduce our approaches to processing and analysing these two contrasting but very complementary types of data to discover more about trace metal distributions in the human brain.
Ron Kimmel: Learning Invariants and Representation Spaces of Shapes and Forms
We study the properties of the Laplace Beltrami Operator (LBO) in processing and analyzing geometric information. The decomposition of the LBO at one end, and the heat operator at the other end provide us with efficient tools for dealing with images and shapes. Denoising, segmenting, filtering, exaggerating are just few of the problems for which the LBO provides an efficient solution. We review the optimality of a truncated basis provided by the LBO, and a selection of relevant metrics by which such optimal bases are constructed. Specific example is the scale invariant metric for surfaces that we argue to be a natural selection for the study of articulated shapes and forms.
In contrast to geometry understanding there is a the emerging field of deep learning. Learning systems are rapidly dominating the areas of audio, textual, and visual information analysis. Recent efforts to convert these successes over to geometry processing indicate that encoding geometric intuition into modeling, training, and testing is a non-trivial task. It appears as if approaches based on geometric understanding do not align with those of data-heavy computational learning. We propose to unify these two methodologies by computationally learning geometric representations and invariants and thereby take a small step towards a new perspective on geometry processing.
Examples of shape matching, facial surface reconstruction from a single image, reading facial expressions, shape representation, and finally definition and computation of invariant operators and signatures will be presented.
Masanori Mishima: Label-free analysis of cell division dynamics in C. elegans embryos
Live fluorescence microscopy with fluorescent protein-tagging is now a standard approach for studying the dynamics of cellular and tissue processes. The specific and high-contrast images obtained by this method are suitable for automated computational analysis. On the other hand, this technique has an inevitable limitation in that it requires genetic manipulation. In addition, as a downside of the high specificity, to observe multiple proteins/structures at the same time, multiple markers need to be expressed. This is possible but can be demanding even in genetically-tractable model organisms such as C. elegans.
In contrast, in ideal situations, conventional transmission light microscopy such as differential interference contrast (DIC) microscopy allows a human eye to observe multiple subcellular structures without any fluorescent labelling in an almost non-invasive manner. For example, in transparent C. elegans embryos, DIC microscopy allows us to detect interphase nuclei, centrosomes, mitotic spindle and metaphase chromosomes, and thus monitor the progress of cell division. However, limited contrast and specificity of label-free images have been major obstacles to automated analysis.
In this talk, automated tracking of the spindle poles by conventional segmentation and determination of the timing of anaphase onset by deep learning will be presented. More challenging tasks to which recent advancements in segmentation techniques should be applied will also be discussed.
Marco Palombo: Non-invasive brain tissue microstructure imaging using diffusion MRI
Characterising the microstructure of an organ non-invasively using molecular diffusion measurements represents a major challenge in medical imaging and life science. Here, we introduce concepts in diffusion magnetic resonance to non-invasively extract structural properties of the brain tissue, from macroscopic tissue features to microscopic cellular morphology.
We show how diffusion-weighted magnetic resonance imaging (DW-MRI) allows to quantify neurites (axons and dendrites) complex morphology in terms of their density and orientation distribution at meso/macroscopic scale. We survey the clinical applications of one of the most popular DW-MRI approach to quantify neurite orientation dispersion and density: the neurite orientation dispersion and density imaging (NODDI) [1].
Lastly, we show how the short and long-range morphology of neurons and astrocytes can affect the diffusion of cell-specific metabolites as measured by diffusion weighted magnetic resonance spectroscopy (DW-MRS) at ultra-high magnetic field gradients and ultra-long diffusion times. We provide evidence that adequate computational modelling strategies seem to allow for the extraction of morphological parameters, including the fine structure, length, and complexity of both astrocytic and neuronal processes [2,3], opening the way towards the first non-invasive histology of the brain tissue.
[1] Zhang, H., Schneider, T., Wheeler-Kingshott, C. A., & Alexander, D. C. (2012). NODDI: practical in vivo neurite orientation dispersion and density imaging of the human brain. Neuroimage, 61(4), 1000-1016.
[2] Palombo, M. et al. (2016). New paradigm to assess brain cell morphology by diffusion-weighted MR spectroscopy in vivo. Proceedings of the National Academy of Sciences, 113(24), 6671-6676.
[3] Palombo, M.,Ligneul, C., Hernandez-Garzon, E., & Valette, J. (2017). Can we detect the effect of spines and leaflets on the diffusion of brain intracellular metabolites?. NeuroImage. In press. doi.org/10.1016/j.neuroimage.2017.05.003