I want to use a High Performance Computing cluster having several CPU, GPU and FPGA for performance enhancement of biomedical signal processing algorithms which are generally hugely time consuming for a sequential computing unit.
We have an algorithm for time-frequency analysis using wavelets for applications in ECG and EEG. We want to "parallelise", because it takes too much time to run, mainly when trying to do zoom.
We developed some algorithms for real-time signal extraction from univariate and multivariate data stream time series. You might have a look here:
http://www.statistik.tu-dortmund.de/sfb823-dp2011.html
There you find a link to our discussion paper (no. 28/11):
Matthias Borowski, Roland Fried: "Robust repeated median regression in moving windows with data-adaptive width selection"
Monte Carlo Simulations for simulating light propagation in tissue springs to mind
http://www.lth.se/fileadmin/atomfysik/Biophotonics/Education/prahl89.pdf
http://omlc.ogi.edu/software/mc/
Good luck,
Bryan
We have an algorithm for time-frequency analysis using wavelets for applications in ECG and EEG. We want to "parallelise", because it takes too much time to run, mainly when trying to do zoom.
DNA and protein sequencing are very computationally challenging. Google Smith-Waterman algorithm and maybe GPU GEMS 4. Our code is freely available on Nvidia's web site if you are using GPUs.
Dear Deba,
Parallel programming is really well appreciated in medical imaging operations. For example to segmentate magnetic resonance images, or to detect and measure volumes, metabolic functions, etc. Any transformation/computation that has to be done to a medical image is a potential application of parallel computing. In fact, there exist a wide variety of challenging solutions written un "cuda", a programming language thought for parallel processing under the newest GPUs, like NVIDIA, etc.
The most challenging applications needing parallel computing are related to 3D and 4D medical image analysis, including modalities like Computed Tomography, Magnetic Resonance and Ultrasound images. Algorithms like segmentation, enhancement and visualization require high computational processes and it is always necesary to have parallel implementations. Development of fast implementations of those algorithms would be great contributions.
Thanks a lot to everyone for the fantastic response. However, I've chosen computation of multi-channel correlation dimension D2 algorithm for parallelizing as of now. This is used in EEG, HRV or EMG data interpretation.
Hello everyone, Mr Bryan Hennelly i was looking the links you sent above and i am interested in your mcml code. Nevertheless i am not familiar with your work. I wonder if you could be so kind to provide me with an input file that needs 20-50 seconds in order to be processed, so i can experiment with your code, because the template.mci you published has very little execution time.
- Badges
- Science topic
- Similar topics
- Biological Science
- Physiology
- Biosignals
- Biomedical Signal Processing