GPU Computing for Medical Image Analysis

GPU Computing has drawn many attractions from diverse fields in either academic or industry. Based on its own physical design, the increasing of Arithmetic Logic Units and reducing Control Units force programmers to rethink about their problems and make insights that be able to map those problems onto GPU architecture, where thousands of massively parallel threads are greedy to race. For instance, Deep Convolutional Neural Network could be impractical to be processed without the support from the GPUs.  In HVCL, we are interested in using multiGPU-cluster to solve several Sparsity problems such as Sparsity Transforms like Wavelet, Sparsity Reconstruction in Compressed Sensing MRI using data-driven approach.

This research direction is centered around GPU-accelerated computing on biomedical image processing such as classical smoothing methods for undersampled MRI data using sparsifying transforms such as Wavelet transform (ICIP 2014, MICCAI 2015, TPDS 2016), or recently leveraging data-driven machine learning methods to deliver a higher quality of the reconstructed MRI (ISBI 2016, MICCAI 2016).

The developing algorithms focused on unsupervised learning techniques by introducing a fast alternating method for reconstructing highly undersampled dynamic MRI data using convolutional sparse coding which can be 2D (ISBI 2016) or 3D (MICCAI 2016). The proposed solution leverages Fourier Convolution Theorem to accelerate the process of learning a set of filters and iteratively refine the MRI reconstruction based on the sparse codes found subsequently. In contrast to conventional Compressed Sensing methods which exploit the sparsity by applying universal transforms such as wavelet and total variation (MICCAI 2015), the current approach extracts and adapts the related information directly from the MRI data using compact shift-invariant filters. The reconstruction outperforms CPU implementation of the state-of-the-art dictionary learning-based approaches by up to two orders of magnitude.

In those venues of publications, MICCAI, stands for Medical Image Computing and Computer Assisted Intervention, is among the top premier international conferences for medical image analysis with an overall acceptance rate of 25%. The prospective outcomes are also expected to publish in similar titles such as ISBI, MICCAI, TMI, MedIA, etc.

Related publications:

  • [PDF] [DOI] T. M. Quan and W. Jeong, “Compressed sensing dynamic MRI reconstruction using GPU-accelerated 3D convolutional sparse coding,” in Proceedings of the 19th international conference on medical image computing and computer-assisted intervention (MICCAI), Springer International Publishing, 2016, pp. 484-492.
    [Bibtex]
    @incollection{quan_compressed3d_2016,
    series = {Lecture {Notes} in {Computer} {Science}},
    title = {Compressed Sensing Dynamic {MRI} Reconstruction Using {GPU}-accelerated {3D} Convolutional Sparse Coding},
    copyright = {©2016 Springer International Publishing Switzerland},
    isbn = {978-3-319-46726-9},
    url = {http://dx.doi.org/10.1007/978-3-319-46726-9_56},
    language = {en},
    number = {9351},
    booktitle = {Proceedings of the 19th international conference on Medical image computing and computer-assisted intervention ({MICCAI})},
    publisher = {Springer International Publishing},
    author = {Quan, Tran Minh and Jeong, Won-Ki},
    year = {2016},
    doi = {10.1007/978-3-319-46726-9_56},
    keywords = {Artificial Intelligence (incl. Robotics), computer graphics, Health Informatics, Image Processing and Computer Vision, Imaging / Radiology, Pattern Recognition},
    pages = {484--492},
    }

 

  • [PDF] [DOI] T. M. Quan and W. K. Jeong, “A fast discrete wavelet transform using hybrid parallelism on GPUs,” IEEE transactions on parallel and distributed systems, vol. 27, iss. 11, pp. 3088-3100, 2016.
    [Bibtex]
    @ARTICLE{quan_fast_2016,
    author={T. M. Quan and W. K. Jeong},
    journal={{IEEE} Transactions on Parallel and Distributed Systems},
    title={A Fast Discrete Wavelet Transform Using Hybrid Parallelism on {GPU}s},
    year={2016},
    volume={27},
    number={11},
    pages={3088-3100},
    keywords={discrete wavelet transforms;graphics processing units;optimisation;parallel processing;CPU;{GPU} DWT methods;{GPU} optimization strategies;{GPU}-based discrete wavelet transform;Haar DWT;ILP maximization;acceleration techniques;computationally-intensive problem acceleration;fast discrete wavelet transform;graphics processing unit;hybrid parallelism;mixed-band memory layout;multilevel transform;single fused kernel launch;time-critical applications;Acceleration;Discrete wavelet transforms;Graphics processing units;Parallel processing;Registers;{GPU} computing;Wavelet transform;bit rotation;hybrid parallelism;lifting scheme},
    doi={10.1109/TPDS.2016.2536028},
    ISSN={1045-9219},
    month={Nov},}

 

  • [PDF] [DOI] T. M. Quan and W. K. Jeong, “Compressed sensing reconstruction of dynamic contrast enhanced MRI using GPU-accelerated convolutional sparse coding,” in 2016 IEEE 13th international symposium on biomedical imaging (ISBI), 2016, pp. 518-521.
    [Bibtex]
    @INPROCEEDINGS{quan_compressed_2016,
    author={T. M. Quan and W. K. Jeong},
    booktitle={2016 {IEEE} 13th International Symposium on Biomedical Imaging ({ISBI})},
    title={Compressed sensing reconstruction of dynamic contrast enhanced {MRI} using {GPU}-accelerated convolutional sparse coding},
    year={2016},
    pages={518-521},
    keywords={Convolution;Convolutional codes;Dictionaries;Encoding;Fourier transforms;Image reconstruction;Magnetic resonance imaging;Compressed Sensing;Convolutional Sparse Coding;{GPU};{MRI}},
    doi={10.1109/ISBI.2016.7493321},
    month={April},}

 

  • [PDF] [DOI] T. M. Quan, S. Han, H. Cho, and W. Jeong, “Multi-GPU Reconstruction of Dynamic Compressed Sensing MRI,” in Proceedings of the 18th international conference on medical image computing and computer-assisted intervention (MICCAI), Springer International Publishing, 2015, pp. 484-492.
    [Bibtex]
    @incollection{quan_multi_2015,
    series = {Lecture {Notes} in {Computer} {Science}},
    title = {Multi-{GPU} {Reconstruction} of {Dynamic} {Compressed} {Sensing} {MRI}},
    copyright = {©2015 Springer International Publishing Switzerland},
    isbn = {978-3-319-24573-7 978-3-319-24574-4},
    url = {http://link.springer.com/chapter/10.1007/978-3-319-24574-4_58},
    abstract = {Magnetic resonance imaging (MRI) is a widely used in-vivo imaging technique that is essential to the diagnosis of disease, but its longer acquisition time hinders its wide adaptation in time-critical applications, such as emergency diagnosis. Recent advances in compressed sensing (CS) research have provided promising theoretical insights to accelerate the MRI acquisition process, but CS reconstruction also poses computational challenges that make MRI less practical. In this paper, we introduce a fast, scalable parallel CS-MRI reconstruction method that runs on graphics processing unit ({GPU}) cluster systems for dynamic contrast-enhanced (DCE) MRI. We propose a modified Split-Bregman iteration using a variable splitting method for CS-based DCE-MRI. We also propose a parallel {GPU} Split-Bregman solver that scales well across multiple {GPU}s to handle large data size. We demonstrate the validity of the proposed method on several synthetic and real DCE-MRI datasets and compare with existing methods.},
    language = {en},
    number = {9351},
    urldate = {2015-10-12},
    booktitle = {Proceedings of the 18th international conference on Medical image computing and computer-assisted intervention ({MICCAI})},
    publisher = {Springer International Publishing},
    author = {Quan, Tran Minh and Han, Sohyun and Cho, Hyungjoon and Jeong, Won-Ki},
    year = {2015},
    doi = {10.1007/978-3-319-24574-4\_58},
    keywords = {Artificial Intelligence (incl. Robotics), computer graphics, Health Informatics, Image Processing and Computer Vision, Imaging / Radiology, Pattern Recognition},
    pages = {484--492},
    file = {Full Text PDF:files/227/Quan et al. - 2015 - Multi-{GPU} Reconstruction of Dynamic Compressed Sen.pdf:application/pdf;Snapshot:files/234/978-3-319-24574-4_58.html:text/html}
    }

 

  • [PDF] [DOI] T. M. Quan and W. Jeong, “A fast Mixed-Band lifting wavelet transform on the GPU,” in IEEE International Conference on Image Processing, 2014, pp. 1238-1242.
    [Bibtex]
    @INPROCEEDINGS{quan_fast_2013,
    AUTHOR="Tran Minh Quan and Won-Ki Jeong",
    TITLE="A Fast {Mixed-Band} Lifting Wavelet Transform on the {GPU}",
    BOOKTITLE="{{IEEE} International Conference on Image Processing}",
    PAGES="1238-1242",
    DAYS=27,
    MONTH=oct,
    YEAR=2014,
    KEYWORDS="Mixed-band, Wavelet, Denoising, {GPU}, Parallel Computing, Compressive
    Sensing, MRI",
    DOI = {10.1109/ICIP.2014.7025247},
    ABSTRACT="Discrete wavelet transform (DWT) has been widely used in many image
    compression applications, such as JPEG2000 and compressive sensing MRI.
    Even though a lifting scheme has been widely adopted to accelerate DWT,
    only a handful of research has been done on its efficient implementation on
    many-core accelerators, such as graphics processing units ({GPU}s). Moreover,
    we observe that rearranging the spatial locations of wavelet coefficients
    at every level of DWT significantly impairs the performance of memory
    transaction on the {GPU}. To address these problems, we propose a mixed-band
    lifting wavelet transform that reduces uncoalesced global memory access on
    the {GPU} and maximizes on-chip memory bandwidth by implementing in-place
    operations using registers. We assess the performance of the proposed
    method by comparing with the state-of-the-art DWT libraries, and show its
    usability in a compressive sensing (CS) MRI application."
    }