Skip to main content

EXtreme-scale Analytics via Multimodal Ontology Discovery & Enhancement

Deliverables

1st dissemination, communication and exploitation report & outline for the following year

This report reviews all the dissemination, exploitation and communication activities performed during the year and it draws the outlines for the following year, divided per activity.

Algorithms for semantic segmentation and detection in histopathology images

Deep neural networks for detection and semantic segmentation of tissue regions in whole-slide images. Targets of trained networks will be released as a milestone during the project. Detection and segmentation algorithms will be implemented in a web-based platform available at Radboudumc, namely CIRRUS Pathology. Segmentation results will be produced in a format compatible with the in-house developed open-source platform ASAP [61], as well as CIRRUS Pathology, which will allow to inspect, modify and further process segmentation results as well as likelihood maps produced by neural networks. Novel techniques of weakly-supervised semantic in whole-slide images will be disseminated as scientific publications. The tool is tested in order to verify Milestone 8.

EXA MODE website

The web site contains the description of the project, a section with a list of publications (and also press releases and articles in the popular press) and a private section for the partners to exchange software, data and information. All public deliverables are available as pdf from the web page. The web site includes an RSS feeder with news on the project and it is updated regularly.

First set of annotated digital pathology data

The first set of annotated whole slide images are made available to the consortium. Such data are selected from the AOEC data and must include at least 100 annotated whole slide images in the final annotated dataset. The annotation requirements are defined by AOEC and MICROSCOPEIT together before the data annotation begins and are made available on the private section of the EXA MODE web page. The data annotations are performed using a software based on previous works of the partners HES-SO and MICROSCOPEIT.

First set of cured, publicly available multimodal and multimedia data

The first set of cured publicly available data include respectively at least 500 and 5’000 histopathology images and related text extracted from scientific literature and the web. The images and the related text include content that is relevant for histopathology diagnostic purposes and that can be used to train machine learning based algorithms.

Set of publicly available algorithms to separate compound images

The algorithms to separate compound images and link them to related text are presented into a scientific publication and they are publicly released on the EXA MODE website.

Final set of cured, publicly available multimodal and multimedia data

The final set of cured publicly available data include respectively at least 500 and 5’000 histopathology images and related text extracted from scientific literature and the web. The images and the related text include content that is relevant for histopathology diagnostic purposes and that can be used to train machine learning based algorithms.

First set of data curated and available

The first set of whole slide images and medical report are made available to the consortium. Such data are selected from the 600’000 AOEC and Radboudumc data according to the consortium requirements.

Searching for OpenAIRE data...

Publications

Generalizing convolution neural networks on stain color heterogeneous data for computational pathology

Author(s): Amjad Khan, Manfredo Atzori, Sebastian Otálora, Vincent Andrearczyk, Henning Müller
Published in: Medical Imaging 2020: Digital Pathology, 2020, Page(s) 26
DOI: 10.1117/12.2549718

A systematic comparison of deep learning strategies for weakly supervised Gleason grading

Author(s): Sebastian Otálora, Manfredo Atzori, Amjad Khan, Oscar Jimenez-del-Toro, Vincent Andrearczyk, Henning Müller
Published in: Medical Imaging 2020: Digital Pathology, 2020, Page(s) 20
DOI: 10.1117/12.2548571

Exploiting biomedical literature to mine out a large multimodal dataset of rare cancer studies

Author(s): Anjani K. Dhrangadhariya, Oscar Jimenez-del-Toro, Vincent Andrearczyk, Manfredo Atzori, Henning Müller
Published in: Medical Imaging 2020: Imaging Informatics for Healthcare, Research, and Applications, 2020, Page(s) 9
DOI: 10.1117/12.2549565

An Analysis of Query Reformulation Techniques for Precision Medicine

Author(s): Maristella Agosti, Giorgio Maria Di Nunzio, Stefano Marchesin
Published in: Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval, 2019, Page(s) 973-976
DOI: 10.1145/3331184.3331289

Probabilistic Word Embeddings in Neural IR - A Promising Model That Does Not Work as Expected (For Now)

Author(s): Alberto Purpura, Marco Maggipinto, Gianmaria Silvello, Gian Antonio Susto
Published in: Proceedings of the 2019 ACM SIGIR International Conference on Theory of Information Retrieval, 2019, Page(s) 3-10
DOI: 10.1145/3341981.3344217

Staining Invariant Features for Improving Generalization of Deep Convolutional Neural Networks in Computational Pathology

Author(s): Sebastian Otálora, Manfredo Atzori, Vincent Andrearczyk, Amjad Khan, Henning Müller
Published in: Frontiers in Bioengineering and Biotechnology, Issue 7, 2019, ISSN 2296-4185
DOI: 10.3389/fbioe.2019.00198

Deep learning-based retrieval system for gigapixel histopathology cases and the open access literature

Author(s): Roger Schaer, Sebastian Otálora, Oscar Jimenez-del-Toro, Manfredo Atzori, Henning Müller
Published in: Journal of Pathology Informatics, Issue 10/1, 2019, Page(s) 19, ISSN 2153-3539
DOI: 10.4103/jpi.jpi_88_18

Search Text to Retrieve Graphs: A Scalable RDF Keyword-Based Search System

Author(s): Dennis Dosso, Gianmaria Silvello
Published in: IEEE Access, Issue 8, 2020, Page(s) 14089-14111, ISSN 2169-3536
DOI: 10.1109/ACCESS.2020.2966823

Focal elements of neural information retrieval models. An outlook through a reproducibility study

Author(s): Stefano Marchesin, Alberto Purpura, Gianmaria Silvello
Published in: Information Processing & Management, 2019, Page(s) 102109, ISSN 0306-4573
DOI: 10.1016/j.ipm.2019.102109

Studying Public Medical Images from the Open Access Literature and Social Networks for Model Training and Knowledge Extraction

Author(s): Henning Müller, Vincent Andrearczyk, Oscar Jimenez del Toro, Anjani Dhrangadhariya, Roger Schaer, Manfredo Atzori
Published in: MultiMedia Modeling - 26th International Conference, MMM 2020, Daejeon, South Korea, January 5–8, 2020, Proceedings, Part II, Issue 11962, 2020, Page(s) 553-564
DOI: 10.1007/978-3-030-37734-2_45

A Framework for Citing Nanopublications

Author(s): Erika Fabris, Tobias Kuhn, Gianmaria Silvello
Published in: Digital Libraries for Open Knowledge - 23rd International Conference on Theory and Practice of Digital Libraries, TPDL 2019, Oslo, Norway, September 9-12, 2019, Proceedings, Issue 11799, 2019, Page(s) 70-83
DOI: 10.1007/978-3-030-30760-8_6