Objective 1. Weakly-supervised knowledge discovery for exascale medical data.
– Weakly supervised information extraction, linking and retrieval in multimodal and multimedia medical data.
– Alignment of the extracted structures to authoritative knowledge sources (e.g. UMLS) – Knowledge structures navigation and refinement.
– Limited human interaction to perform data annotations.
– Multimodal & multimedia knowledge management
– Ground breaking extraction of valuable knowledge value from exascale heterogeneous data
– Faster deep analysis.
2. Develop extreme scale analytic tools for heterogeneous exascale multimodal & multimedia data.
– Develop and release tools for the homogenization of highly heterogeneous images (compound images, variable magnification, colour and contrast).
– Develop and release tools for weakly supervised extraction of multimodal semantic concepts from text and images.
– Develop and release tools to visualize, navigate and refine knowledge structures.
– Heterogeneous data homogenization.
– Efficient training of DNN and effective generalization to heterogeneous test data – Semantic data compression.
– Open source tools applicable to other domains.
– Extreme-scale multimodal & multimedia entity based analytics.
– Semantic multimodal & multimedia data management.
– Novel tools for multimodal & multimedia knowledge visualization, navigation and refinement.
3. Healthcare & industry decision-making adoption of extreme-scale analysis & prediction tools.
– Adoption of the results by industry and society.
– Partner companies can provide precise predictions and decision making support on exascale highly heterogeneous data.
– Pathologists can benefit of computer aided diagnosis systems, they can search and retrieve for specific terms or visual features in proprietary datasets and in scientific literature.
– Pathologists can easily share and retrieve portions of images.
– Multimodal & multimedia knowledge-based commercial products and services (for medicine as well as for other domains).
– Extraction of societal value from exascale, heterogeneous, medical data.
