 |
HiPiP : High Performance Image Processing |
 |
Work packages
WP2: Application Parallelisation
WP3: Server Based Computing
WP4: Demonstration
WP5: Exploitation, Dissemination and Project Management
Public deliverables
Project Summary
The trend and need in imaging to display very large 2-, 3- and 4 D (space + time) images incl. enhanced features drives conventional computing resources to the limit. Image processing is a demanding process that requires a large amount of computing resources. As a consequence there is a growing need for sophisticated real-time imaging during intervention to give the medical specialist direct feedback. The ongoing drive for cost reduction simply does not allow for an increase in equipment expenses. Hence, expensive dedicated hardware to improve the (real-time) performance is not an option.
In recent times improved imaging technology has offered high-quality images of the human body, allowing for new medical treatments. Likewise, electron microscopes allow lab researchers to obtain high-resolution structural information on sub-micron level. Now that these technologies are being applied in medical centres and in industrial quality labs there is a growing demand to use such imaging in an interactive, dynamic setting. This implies that the image processing time of these very large data sets needs to be reduced significantly, to allow for fast or even real time imaging.
HiPiP addresses this demand for shorter image processing times in the following areas:
- Detailed brain imaging.
- Real-Time Simulation for Radiotherapy
- Mass screening for cancer
- Minimal invasive interventions.
- Scanning Electron Microscopes.
HiPiP’s challenge in all three sectors is to achieve high throughput image processing of often very large and heterogeneous data sets. Significant improvements are needed in methodology and processing strategies to increase the speed to such a level that these novel applications can be really used by customers in their real-life cases.
The main objective of the HiPiP project is to develop novel methods for parallel image processing, improving the throghput time of algorithms to process large and heterogeneous data sets to make them usable for real-time or low-latency procedures. Easy parallel solutions exist for many basic image processing algorithms. Often many data points undergo the same algorithm independently of each other, leading to long processing throughput times. However, for complex image processing algorithms there is no simple parallel solution. Either because of data and operation dependency, because different scales of granularity need to be connected, or since it is unclear at which level of granularity an efficient parallel solution exist. These are the algorithms that HiPiP addresses.
The project addresses three main research topics:
- Parallel image processing on a multi-core standard processor
- Using the multi-core standard processor for all tasks during low latency image processing, including the background tasks
- Using a single multi-core standard processor server for several pieces of imaging equipment
In the last stage demonstrators will be built that show the feasibility complex medical image processing using standard multi core hardware. The project addresses the improvement of medical imaging for several situations, in minimal invasive intervention, and in biomedical research.
Project description
General goals
In recent times improved imaging technology has offered high-quality images of the human body, allowing for new medical treatments. Likewise, electron microscopes allow lab researchers to obtain high-resolution structural information on sub-micron level. Now that these technologies are being applied in medical centres and in industrial quality labs there is a growing demand to use such imaging in an interactive, dynamic setting. This implies that the image processing time of these very large data sets needs to be reduced significantly, to allow for fast or even real time imaging.
HiPiP addresses this demand for shorter image processing times in the following areas:
- Detailed brain imaging can significantly improve the treatment of diseases related to the brain and neural system, as well as tumours. Currently the processing of complex brain images takes days and is only applied in academic environments. Aim of HiPiP is to reduce the processing time of complex brain images from days to minutes, enabling this type of image processing to be applied in regular hospitals. This application will be validated within the related System@tic cluster in France.
- Real-Time Simulation for Radiotherapy can significantly improve the outcome of radiotherapy treatments. As the detailed delineation of the tumour s time consuming, only a partial tumour delineation can be iterated over with the patient present. There is a need to better utilise available patient imaging data and deliver more precise and performing treatments that maximise curative probability and minimise patient side effects. This application will be validated within the related System@tic cluster in France.
- Mass screening for cervix cancerto enable fast and systematic mass screening of a population at risk. Conventional cervix cancer screening is currently performed by human scoring, a time consuming tasks needing highly-trained personnel. A mass screening according to this model is unaffordable, since a technician can score about 80 samples per day. This application will be validated within the related System@tic cluster in France.
- Minimal invasive interventions (i.e. bringing in instruments into patient through a small opening) are increasingly replacing open surgery procedures, for instance in cancer and cardiovascular treatments. In contrast to open surgery, for minimal invasive interventions a large amount of imaging is needed during intervention. Different aspects of the biomedical data have to be presented promptly, in such a way that the medical professional is able to perform the intervention as fast and precise as possible, preferably on real-time basis. Philips Healthcare will validate this application in its medical imaging equipment.
- Scanning Electron Microscopes (SEMs) are used as nanotechnology tools to obtain high-resolution structural information on nm-scale samples, for example in the semiconductor or biotechnology industry. These sectors increasingly demand real-time 3D imaging of dynamic events, to allow for real-time sample handling and modification during inspection. FEI will validate this application in its SEMs.
HiPiP’s challenge in all three sectors is to achieve high throughput image processing of often very large and heterogeneous data sets. Significant improvements are needed in methodology and processing strategies to increase the speed to such a level that these novel applications can be really used by customers in their real-life cases.