FundRef: 501100006446 , 501100006445
ISNI: 0000000419371290
RRID: RRID:SCR_011747 , RRID:nlx_152513
Wikidata: Q144488
FundRef: 501100006446 , 501100006445
ISNI: 0000000419371290
RRID: RRID:SCR_011747 , RRID:nlx_152513
Wikidata: Q144488
Deep neural networks (DNNs) have achieved outstanding performance and broad implementation in computer vision tasks such as classification, denoising, segmentation and image synthesis. However, DNN-based models and algorithms have seen limited adaptation and development within radiomics which aim to improve diagnosis or prognosis of cancer. Traditionally, medical practitioners have used expert-derived features such as intensity, shape, textual, and others. We hypothesize that, despite the potential of DNNs to improve oncological classification performances in radiomics, a lack of interpretability of such models prevents their broad utilization, performance, and generalizability. Therefore, the INFORM consortium proposes to investigate explainable artificial intelligence (XAI) with a dual aim of building high performance DNN-based classifiers and developing novel interpretability techniques for radiomics. First, in order to overcome the limited data typically available in radomic studies, we will investigate Monte Carlo methods and generative adversarial networks (GAN) for realistic simulation that can aid building and training DNN architectures. Second, we tackle the interpretability of DNN-based feature engineering and latent variable modeling with innovative developments of saliency maps and related visualization techniques. Both supervised and unsupervised learning will be used to generate features, which can be interpreted in terms of input pixels and expert-derived features. Third, we propose to build explainable AI models that incorporate both expert-derived and DNN-based features. By quantitatively understanding the interplay between expert-derived and DNN-based features, our models will be readily understood and translated into medical applications. Fourth, evaluation will be carried out by clinical collaborators with a focus on lung, cervical and rectal cancer. These proposed DNN models, specifically developed to reveal their innerworkings, will leverage the robustness and trustworthiness of expert-derived features that medical practitioners are familiar with, while providing quantitative and visual feedback. Overall, our methodological research will advance interpretability of feature engineering, generative models, and DNN classifiers with applications in radiomics and broad medical imaging. With this project we aim at maximizing the impact on the patient management of ML and DL techniques by developing novel methods to facilitate training of decision-aid systems for clinical treatment strategies optimization. The methodological approaches we propose in this specific area will play a major role in facilitating the acceptability of DL-based decision-aid systems relying on medical imaging for oncology. The proposed validated predictive models in various cancer types within the context of this project might subsequently be used to drive future prospective clinical studies in which patients could be offered alternative treatment strategies based on the results of these predictive models. Such a clinical and social potential is further enhanced by the public-private collaboration proposed in this project, where the developed methodologies will find their way in products. The multidisciplinarity of INFORM is key to meet the target challenges and achieve the proposed goals. All partners have their individual world-leading qualifications and additional scientific expertise providing all the prerequisites for the efficient implementation of INFORM’s approach. The successful implementation of this project will have a large and prolonged impact both in the Medical/Oncology and the Computing/ Artificial Intelligence field of predictive radiomics model, as well as the same methodology could be extended to other diagnostic and therapeutic medical applications.
<script type="text/javascript">
<!--
document.write('<div id="oa_widget"></div>');
document.write('<script type="text/javascript" src="https://www.openaire.eu/index.php?option=com_openaire&view=widget&format=raw&projectId=chistera____::78d1c2100cb1035b5c225b2d54641d99&type=result"></script>');
-->
</script>
<script type="text/javascript">
<!--
document.write('<div id="oa_widget"></div>');
document.write('<script type="text/javascript" src="https://www.openaire.eu/index.php?option=com_openaire&view=widget&format=raw&projectId=chistera____::78d1c2100cb1035b5c225b2d54641d99&type=result"></script>');
-->
</script>
The nematic phase (N) is the least ordered liquid crystal phase, and in which the long axes of the rod-like molecules are more or less aligned in the same direction, known as the director, whereas their centres of mass are randomly distributed. This phase is easily replicated by throwing a handful of matches into a box and shaking it. Providing there are enough matches, then, for packing reasons, they will all line-up in the same direction and effectively you have a nematic phase. Providing the matches were thrown into the box randomly, there will be an equal number of matches pointing with their heads in one direction as in the other. This is exactly the case for the conventional nematic phase, and the molecules are equally likely to be pointing in either direction along the director, and the phase is described as being non-polar. The conventional N phase underpins liquid crystal display technology which has a market value predicted to grow to almost $200 billion by 2025. Over 100 years ago, it was first suggested that a nematic phase could exist in which all the molecules could align in the same direction. This is the molecular equivalent of taking the matches, throwing them into the box, shaking it, and discovering that all the matches now lay with their heads pointing in the same direction. This is known as polar ordering and the phase is called the ferroelectric nematic (NF) phase. Very recently a new nematic phase was discovered having remarkable properties, and it has been suggested that this is the long sought after NF phase. This has the potential to be a hugely significant discovery from both fundamental and technological viewpoints. The polar ordering in the NF phase makes it vastly more sensitive to an electric field than the conventional N phase, and this will dramatically improve the performance of liquid crystal display devices in terms of both speed and power consumption. In addition, the study of this new phase has the potential of generating transformative new fundamental chemistry, physics and biology. For example, it was predicted over forty years ago that the NF phase, in order to reduce electrostatic energy, will twist giving a polar cholesteric phase, the spontaneous chirality being controlled through steric and electrostatic interactions between achiral molecules. Such an observation could have huge significance in understanding the origins of chirality. It has been proposed that on cooling the conventional N phase into the NF phase, the molecular dipoles will align spontaneously in a single direction. At this point there is a strong tendency towards crystallisation. If this can be suppressed, however, equal numbers of domains having opposite polarisations should form, separated by domain walls. The application of an electric field will remove this degeneracy and domains having favourable polarity will grow and unfavourable will shrink. The aim of this programme is to begin to understand what molecular features are required to observe the NF phase. Some time ago computer simulations suggested that an asymmetric or tapered shape combined with a longitudinal dipole moment promote polar order, and the very early experimental data available support this view. To achieve our aim, we will need to enhance our understanding of how to manipulate liquid crystallinity though molecular electrostatic and steric interactions. This programme has the very real potential to deliver materials that will lead to transformative new fundamental chemistry, physics and biology, and new technologies including the next generation of display devices.
<script type="text/javascript">
<!--
document.write('<div id="oa_widget"></div>');
document.write('<script type="text/javascript" src="https://www.openaire.eu/index.php?option=com_openaire&view=widget&format=raw&projectId=ukri________::97bcb3e6f2161ce01e2b416f0d1f6cc1&type=result"></script>');
-->
</script>
<script type="text/javascript">
<!--
document.write('<div id="oa_widget"></div>');
document.write('<script type="text/javascript" src="https://www.openaire.eu/index.php?option=com_openaire&view=widget&format=raw&projectId=ukri________::97bcb3e6f2161ce01e2b416f0d1f6cc1&type=result"></script>');
-->
</script>
Databases are irreplaceable in our modern information society. Classically, information has been stored in rigidly structured databases employing the relational data model and the query language SQL. This has led to highly optimised relational database management systems such as Oracle or Mircrosoft SQL that have large-scale industrial deployment. Underpinning much of the success of these systems has been their close connection with their mathematical foundations within relational algebra and mathematical logic. However, the internet has dramatically changed our understanding of data and databases: i) data on the Web is inherently hierarchical and arranged in a network/graph; and ii) the decentralised nature of the Web means that data comes from various heterogenous sources, is often incomplete or unreliable and has no uniform structure. Still, Web data retains some structure and, consequently, semi-structured data seeks to understand what structure persists and how it can be utilised. Examples of widespread, industrially-used data models are XML, JSON and RDF. Mathematically, semi-structured data is usually represented as unranked labelled trees where nodes are accessed via the parent, child and sibling relations or labelled graphs where nodes are accessed via the edge relation. Data elements are stored at the nodes. Due to the special features of Web data mentioned above, query languages for semi-structured data face significantly greater challenges compared to those for data stored in a relational database: i) queries must navigate the path structure within a tree or graph and query the data elements found along such paths to explore important properties of the data; and ii) one must add a common vocabulary to the data - often via an ontology - that provides a logical layer for integrating semantically related data from heterogeneous sources. But as semi-structured data models have become increasingly sophisticated and expressive - e.g. in order to model the uncertainty attached to the data - the development of matching query and ontology languages have struggled to keep pace. Indeed, while there is a large body of work on specific semi-structured data models and their corresponding query languages there is currently no comprehensive theory that both accounts for existing semi-structured data formats and their query languages, and is able to guide their extension to the next generation of semi-structured data formats. For example, while the widely used query language XPath has been recently extended from XML to graph data, there is currently no agreed mechanism to further extend XPath to graph data with uncertainty. Our central insight is that coalgebra provides the right level of abstraction to underpin a comprehensive theory of query and ontology languages for semi-structured data. This is because i) coalgebra generalises - via coalgebraic modal logic - the usual modal logics used for traversing trees and graphs; and ii) coalgebra generalises - via coalgebraic logic programming - standard rule-based ontology languages which are based upon logic programming.
<script type="text/javascript">
<!--
document.write('<div id="oa_widget"></div>');
document.write('<script type="text/javascript" src="https://www.openaire.eu/index.php?option=com_openaire&view=widget&format=raw&projectId=ukri________::770e90536dd1bb0cf7f9b862477a297b&type=result"></script>');
-->
</script>
<script type="text/javascript">
<!--
document.write('<div id="oa_widget"></div>');
document.write('<script type="text/javascript" src="https://www.openaire.eu/index.php?option=com_openaire&view=widget&format=raw&projectId=ukri________::770e90536dd1bb0cf7f9b862477a297b&type=result"></script>');
-->
</script>
We live in an information age, when computers and the software that drives them permeate every aspect of our society. There are two fundamentally important aspects of computation. - One concerns the resources needed to perform computational tasks: how many computational steps are needed, how much computer memory, etc. - The other concerns our ability to master the staggering complexity of the computer systems we create and use. The only way of managing this complexity is to use principles of modularity and abstraction, so that at each step of our design and construction of the system, we see only a very limited piece, whose complexity we can master. While the study of each of these aspects of computing has been greatly advanced as computer science has developed, currently we have a very limited understanding of how they relate to each other. Building on our previous work, this project aims to greatly enhance our common understanding of these issues, and to develop new mathematical tools and methods for studying computation based on this. This can lead in turn to new possibilities for fundamental advances in the field.
<script type="text/javascript">
<!--
document.write('<div id="oa_widget"></div>');
document.write('<script type="text/javascript" src="https://www.openaire.eu/index.php?option=com_openaire&view=widget&format=raw&projectId=ukri________::57a7d3465edaac527e20cb54436e1f2c&type=result"></script>');
-->
</script>
<script type="text/javascript">
<!--
document.write('<div id="oa_widget"></div>');
document.write('<script type="text/javascript" src="https://www.openaire.eu/index.php?option=com_openaire&view=widget&format=raw&projectId=ukri________::57a7d3465edaac527e20cb54436e1f2c&type=result"></script>');
-->
</script>
Abstracts are not currently available in GtR for all funded research. This is normally because the abstract was not required at the time of proposal submission, but may be because it included sensitive information such as personal details.
<script type="text/javascript">
<!--
document.write('<div id="oa_widget"></div>');
document.write('<script type="text/javascript" src="https://www.openaire.eu/index.php?option=com_openaire&view=widget&format=raw&projectId=ukri________::7d3c2de104b18d9e0437c88fdcfee626&type=result"></script>');
-->
</script>
<script type="text/javascript">
<!--
document.write('<div id="oa_widget"></div>');
document.write('<script type="text/javascript" src="https://www.openaire.eu/index.php?option=com_openaire&view=widget&format=raw&projectId=ukri________::7d3c2de104b18d9e0437c88fdcfee626&type=result"></script>');
-->
</script>