Below the list of projects cofunded by the María de Maeztu program (selected via internal calls, in this link the first one launched at the beginning of the program, and in this link the second one, launched in September 2016).

In addition, the program supported:

The detail of the internal procedures for the distribution of funds associated to the program can be found here

Related Assets:

Back

Bio Image and Signal Analysis

Bio Image and Signal Analysis

Bio Image and Signal Analysis
Web of the project for updated details https://www.upf.edu/en/web/bia

Web of the project for updated details: https://www.upf.edu/en/web/bia 

Researchers associated to [email protected] use data- and knowledge-driven (machine learning; image&signal analysis; computational modelling) algorithms to support both the extraction of pertinent information from the current (large dataset) information sources as well as for supporting optimal decision-making for biomedical research as well as patient treatment.

These approaches used allow to do supervised learning to suggest the most efficient approach for data analysis and decision processes to address a specific clinical question, as well as to perform unsupervised analysis of the data, enabling the suggestion of novel information content towards further hypothesis-driven assessment of the data thus supporting new discoveries in biomedical research as well as address clinic questions. Challenges include data dimensionality reduction and data-driven decision-making.

[email protected] is embedded in the BCN MedTech unit, which is a collaboration of the groups: PhySense (Sensing in Physiology and Biomedicine) – O. Camara/B. Bijnens; SIMBiosys  (Simulation, Imaging and Modelling for Biomedical Systems) – M.A. González Ballester /G. Piella; NTSA  (Nonlinear Time Series Analysis) – R. Andrzejak and BERG (Biomedical Electronics Research Group) – A. Ivorra. Given the strong interdisciplinarity, [email protected] further links to the groups working in Artificial Intelligence; Natural Language Processing; and Multimedia Analysis and Visualisation.

 

[email protected] aims at providing a comprehensive approach towards Image and Signal Analysis in Biomedical applications. Therefore, the programme encompasses the following complementary aspects (see figure):

  • Algorithmic research: Specific, interdisciplinary, research projects will be performed , predominantly in the Bio-Image Analysis field (microscopy and high-resolution medical imaging) and non-linear signal analysis.
  • Repository: High resolution imaging and signal datasets will be made available for the wider research community.
  • Analysis Tools: Tools developed within BCN MedTech as well as in the framework of [email protected] will be made available for researchers trough a platform for tool-sharing and tool/algorithm annotation.
  • Benchmarking: Benchmarking approaches, linking tools and datasets (both [email protected] as well as external ones) will be investigated and implemented.
  • Consultancy: Specialised (technical/processing) knowledge, especially on Bio-Image Analysis, will be made accessible for biomedical researchers in the Barcelona Area.
  • Training: [email protected] will put a special emphasis on the organisation of training events for a wide range of researchers and specialist. This will range from Bio-Image Analysis tools courses and tagatons for tools description; over student summer schools; going towards specialised (clinical/biomedical) workshops.
  • Networking: [email protected] will ensure wide dissemination in the research community trough active participation and organisation of networking events, linked to networking projects (e.g. COST project – NEUBIAS and COSMOS) and international and local organisations.

 

 

Some special focus will be on:

  • Generalisation of algorithms and automated analyses.

The large diversity applications and imaging/signal acquisition techniques requires the development of algorithms for analysis tasks to perform robustly and equally well under different scenarios. Most of the existing tools or methods have been tailored for specific applications or projects. Applying to other datasets typically requires parameter tuning or re-programming of the software. Manual software adaptations are tedious and raise obstacles for life scientists due to their lack of expertise in software engineering.

Machine learning methods provide an effective way to automate the analysis, as they seek to use intrinsic data structure and expert annotations to infer models that can be used to solve versatile data analysis tasks. By providing a general solution through learning processing rules from examples rather than relying on manual adjustments of parameters or pre-defined processing steps, machine learning methods are particularly more flexible than conventional image processing techniques for solving complex multi-dimensional data analysis tasks.

Furthermore, automation is playing an increasingly important role in image/signal processing and analysis. We expect that the techniques to be developed within [email protected] will be broad enough and can effectively address the parameter-tuning problem. Additionally, for reproducibility and easy access, it should be feasible to integrate them into the automation pipeline from the commonly used platforms in the community.

  • Networking through community support.

Every biologist/biomedical researcher now has the potential to investigate the multidimensional operation of biological systems. At the same time, they are struggling with the growing amount of data at hand and the augmenting complexity of image/signal analysis methods. At least to some extent, their ability for customization of image analysis workflows is required but this knowledge is not taught in biology curriculum in universities. Also, the number of tools is increasing, which inevitably adds further hurdles in their daily efforts to find matching tools to address their specific biological questions. Ultimately, automation of image analysis is becoming a necessity as highly-automated acquisition of bioimage data has become more accessible (e.g. high-throughput screening microscopy). For all these reasons, various training and support services should be provided, such as software usage support, training on bioimage analysis, techniques for common analysis topics: visualization, segmentation, colocalization, deconvolution, registration, morphology analysis, etc. Beyond education purposes, we believe that this will promote research projects and collaborations both locally and internationally. In return, these activities could lead to potential interesting research projects and collaboration in the future.

  • (Inter-) national collaboration.

To achieve these objectives, [email protected] will also seek external (technical- and application-wise) collaborations with both local institutes (e.g. IRB Barcelona, CRG, IBEC, CEXS) and international ones (e.g. University of Heidelberg, ETH, EMBL). Most of which we have worked or collaborated with previously.

To learn more:

 

Slides

 


Related Assets: