The dream is to provide a set of tools that support exploration and analysis of new image data in the context of what is already known, to plan experiments which resolve uncertainty, and to discover new and significant relationships.
Studying the function and dysfunction of the cell is at the centre of modern medicine, life science research, the pharmaceutical/biotechnology industry and more recently agriculture. It is one in which Europe is particularly well placed. The interactions of cells, cellular components and molecules is an immensely data rich undertaking, especially since the availablily of genetic information. This has led to the emergence of the new field of system biology, which treats the cell as system with complex adaptive control mechanisms to be understood under various conditions.
Life scientists talk of bioinformatics as the discipline which will help them make sense of this information. Various genomes have now been sequenced and there a wide range of public domain databases of gene and protein data.
However bioinformatics has so far concentrated on methods able to parse one dimensional sequences of data such as DNA though they are starting to include rudimentary machine learning techniques (PCA, genetics algorithms, neural networks, SVM, etc)
Since the invention of the microscope however, the study of the cell has always relied upon the image, and techniques to analyse this body of data are still very weak and limited to basic image processing operations.
Scientists and biologists working to understand the functioning of cells have access to a range of tools to examine the cell in its various states, the richest of which is probably the confocal microscope. This is often used in conjunction with various dyes which highlight active structure or events.
This instrument provides XYZ image sequences in time, but also across various excitation and emission wavelengths (to visualise the dyes) and across cells within a population or tissue of mixed cells, resulting in a 4D to 7D dataset. Europe is at the forefront of this work with all major instruments originating here
Other relevant tools include MRI as commonly used in diagnostic medical work and to which computer vision, albeit not cognitive vision, has contributed greatly in recent years. Again Europe is well placed with 2 of the 4 producers.
Many classic imaging problems are absent in this domain, since the objects are well illuminated, while surface reflection and occlusion effects are not present. However variable shape and deformation can be observed. Robust segmentation and classification are essential.
The key research issues for vision include:
hypothesis guided imaging and data extraction event recognition and categorisation integration with external datasets reasoning under uncertainty reasoning across scales and levels of abstraction tracking variable objects over long periods of time processing very large data sets modelling instrument deficiencies focus of attention control fusion of information from complementary sources high dimensional data visualisation and exploration
Many of these are common with other activities within the cognitive vision field.
The dream is to provide a set of tools that support exploration and analysis of new image data in the context of what is already known, to plan experiments which resolve uncertainty, and to discover new and significant relationships.