Objective
Manual analysis of 3D medical and seismic data is very work intensive and difficult. In the TriTex project, leading European research groups from the medical and seismic data analysis domains will join effort to produce new automation methodologies and tools. Important properties with respect to finding oil and gas and performing medical diagnosis (e.g. detecting tumours, Alzheimers-, and Creutzfeld-Jakob diseases) are mapped by the texture content and automation may be obtained by 3D texture recognition. The major tasks involved in the project will be the development of new texture attributes suited for the 3D data along with techniques for real-time signal content navigation and visualisation. The developed algorithms will be evaluated in large-scale tests on real-world data. Expected results are new algorithms leading to a significant market potential. Automated tools have the potential to reduce analysis cost and increase the understanding. Increased understanding will lead to increased diagnosis success and oil and gas exploitation / finding rate.
Objectives:
The TriTex project aims at improving the European competitiveness in the oil and gas industry and building a European excellence and competitiveness on medical image database query and retrieval. Joint benefits will be achieved through joint developments in the two domains using common technology. This will allow development of better solutions and a quicker market penetration, which are targets clearly to the benefit of the European citizen.
The project aims at improving and automating the current elaborate and difficult manual approaches by mapping important properties with 3D texture information. It can furthermore be noted that the manual techniques are to a large extent limited by large data volumes and the 2D nature of the display technology, problems that a 3D automated system are overcoming. Scientific data analysts will be enabled by the tools to meet the increasing demands for reduced turnaround time and increased efficiency.
The project will only be successful if it can match or beat the interpretation quality obtained by manual interpretation, with the effort commonly used in such situations. The consortium has a solid foundation in all elements from end-user, through industrial R&E, to academic research.
Work description:
Essential elements of the project are development of 3D texture attributes and real-time indexing and data navigation techniques with an appropriate human interface and visualisation. Texture recognition is an established field in digital image processing. Common elements in most approaches to texture recognition are attribute extraction and clustering or classification (depending on whether the classes are known a priori). The most commonly applied attribute extractors are digital filters with non-linearity and smoothing, statistical measures, and model based approaches.
There are two problems with most texture recognition approaches regarding the data types addressed in this project. First they are developed for and applied to 2D data sets (images). One of the tasks will be to extend existing 2D approaches to 3D. Next the data have some properties that may need special attention. For example, the orientation of the pattern may be irrelevant; i.e. two patterns may be considered identical even if they are tilted relative to each other. Hence it might be necessary to develop tailor-made attribute extraction schemes.
The next main element of the research is development of query by content strategies and techniques. A proper query by content approach will allow the end-user to navigate effortlessly in huge data repositories, taking advantage of knowledge from prior examples and having a self-learning system.
To fully utilise the results, adequate visualisation is necessary. Visualisation of 3D data is a non-trivial task that will be carefully examined. Special care must be taken to ensure maximum added value for the end-user.
The methodology will be tested on large-scale real-world medical and seismic data. The results will be examined carefully. Iterations over tool development and large-scale tests might be necessary.
Milestones:
March 2001: Requirements and Dissemination and Use Plan.
June 2001: Large-scale test data selected.
October 2001: Feasibility and literature study.
April 2002: Working concept prototype.
October 2002: Algorithms and method selection.
January 2003: Algorithms/prototypes.
October 2003: Large-scale experiments; result assessment.
October 2003: Technical report, patent applications, and publications.
Fields of science (EuroSciVoc)
CORDIS classifies projects with EuroSciVoc, a multilingual taxonomy of fields of science, through a semi-automatic process based on NLP techniques.
CORDIS classifies projects with EuroSciVoc, a multilingual taxonomy of fields of science, through a semi-automatic process based on NLP techniques.
- natural sciencescomputer and information sciencesdata science
- social sciencesmedia and communicationsgraphic design
- engineering and technologyelectrical engineering, electronic engineering, information engineeringelectronic engineeringcontrol systems
- natural sciencescomputer and information sciencesdatabases
- humanitieslanguages and literatureliterature studies
You need to log in or register to use this function
Call for proposal
Data not availableFunding Scheme
CSC - Cost-sharing contractsCoordinator
TANANGER
Norway