Limits...
Real-time multi-view deconvolution.

Schmid B, Huisken J - Bioinformatics (2015)

Bottom Line: State-of-the-art multi-view (MV) deconvolution simultaneously fuses and deconvolves the images in 3D, but processing takes a multiple of the acquisition time and constitutes the bottleneck in the imaging pipeline.Here, we show that MV deconvolution in 3D can finally be achieved in real-time by processing cross-sectional planes individually on the massively parallel architecture of a graphics processing unit (GPU).Our approximation is valid in the typical case where the rotation axis lies in the imaging plane.

View Article: PubMed Central - PubMed

Affiliation: Max Planck Institute of Molecular Cell Biology and Genetics, 01307 Dresden, Germany.

No MeSH data available.


Related in: MedlinePlus

Plane-wise multi-view deconvolution concept and performance. (a) Concept of plane-wise deconvolution for two views. Each dataset is resliced into planes orthogonal to the microscope’s rotation axis. Datasets are deconvolved plane-by-plane. (b) Memory requirements for traditional 3D and our plane-wise multi-view deconvolution, for various data sizes and numbers of views, on a logarithmic scale. (c) Execution times for plane-wise multi-view deconvolution, implemented on GPU and CPU, and 3D deconvolution, with and without GPU support. Memory requirements for 3D deconvolution timings for the 20483 pixel dataset were beyond the capabilities of our workstation. (d–i) Resulting images of a 9 h post-fertilization transgenic Tg(h2afva:h2afva-mCherry) zebrafish embryo, using different methods (view along the rotational axis, scale bar 100 , 10  in the inset): (d, e) acquired raw data, (f–i) fusion performed by (f) averaging, (g) entropy-weighted averaging, (h) 3D multi-view deconvolution and (i) plane-wise multi-view deconvolution (10 iterations). (Dell T6100, Intel E5-2630 @2.3 GHz 2 processors, 64 GB RAM; Graphics card: Nvidia GeForce GTX TITAN Black)
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4595906&req=5

btv387-F1: Plane-wise multi-view deconvolution concept and performance. (a) Concept of plane-wise deconvolution for two views. Each dataset is resliced into planes orthogonal to the microscope’s rotation axis. Datasets are deconvolved plane-by-plane. (b) Memory requirements for traditional 3D and our plane-wise multi-view deconvolution, for various data sizes and numbers of views, on a logarithmic scale. (c) Execution times for plane-wise multi-view deconvolution, implemented on GPU and CPU, and 3D deconvolution, with and without GPU support. Memory requirements for 3D deconvolution timings for the 20483 pixel dataset were beyond the capabilities of our workstation. (d–i) Resulting images of a 9 h post-fertilization transgenic Tg(h2afva:h2afva-mCherry) zebrafish embryo, using different methods (view along the rotational axis, scale bar 100 , 10 in the inset): (d, e) acquired raw data, (f–i) fusion performed by (f) averaging, (g) entropy-weighted averaging, (h) 3D multi-view deconvolution and (i) plane-wise multi-view deconvolution (10 iterations). (Dell T6100, Intel E5-2630 @2.3 GHz 2 processors, 64 GB RAM; Graphics card: Nvidia GeForce GTX TITAN Black)

Mentions: The primary goal of MV fusion is the improvement of the poor axial resolution in a single 3D dataset using the superior lateral resolution of an additional, overlapping dataset, and not necessarily to improve resolution beyond the intrinsic lateral resolution. We therefore approximated the full 3D point spread function (PSF) with a 2D PSF, neglecting one lateral component (along the rotation axis), and processed each plane orthogonal to the rotation axis independently (Fig. 1a). Memory requirements were thereby reduced by the number of lines read out from the camera chip, i.e. typically 100–1000 fold (Fig. 1b). This allowed us to implement the entire MV deconvolution on a GPU. Taking advantage of three CUDA (Compute Unified Device Architecture) streams, we interleaved GPU computations with data transfers, such that not only expensive copying to and from GPU memory, but also reading and writing data from and to the hard drive came without additional cost (Supplementary Note S2). Compared with 3D MV deconvolution, with and without GPU support, we thereby reduced processing times by a factor of up to 25 and 75, respectively (Fig. 1c, Supplementary Table S1), while producing comparable results.Fig. 1.


Real-time multi-view deconvolution.

Schmid B, Huisken J - Bioinformatics (2015)

Plane-wise multi-view deconvolution concept and performance. (a) Concept of plane-wise deconvolution for two views. Each dataset is resliced into planes orthogonal to the microscope’s rotation axis. Datasets are deconvolved plane-by-plane. (b) Memory requirements for traditional 3D and our plane-wise multi-view deconvolution, for various data sizes and numbers of views, on a logarithmic scale. (c) Execution times for plane-wise multi-view deconvolution, implemented on GPU and CPU, and 3D deconvolution, with and without GPU support. Memory requirements for 3D deconvolution timings for the 20483 pixel dataset were beyond the capabilities of our workstation. (d–i) Resulting images of a 9 h post-fertilization transgenic Tg(h2afva:h2afva-mCherry) zebrafish embryo, using different methods (view along the rotational axis, scale bar 100 , 10  in the inset): (d, e) acquired raw data, (f–i) fusion performed by (f) averaging, (g) entropy-weighted averaging, (h) 3D multi-view deconvolution and (i) plane-wise multi-view deconvolution (10 iterations). (Dell T6100, Intel E5-2630 @2.3 GHz 2 processors, 64 GB RAM; Graphics card: Nvidia GeForce GTX TITAN Black)
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4595906&req=5

btv387-F1: Plane-wise multi-view deconvolution concept and performance. (a) Concept of plane-wise deconvolution for two views. Each dataset is resliced into planes orthogonal to the microscope’s rotation axis. Datasets are deconvolved plane-by-plane. (b) Memory requirements for traditional 3D and our plane-wise multi-view deconvolution, for various data sizes and numbers of views, on a logarithmic scale. (c) Execution times for plane-wise multi-view deconvolution, implemented on GPU and CPU, and 3D deconvolution, with and without GPU support. Memory requirements for 3D deconvolution timings for the 20483 pixel dataset were beyond the capabilities of our workstation. (d–i) Resulting images of a 9 h post-fertilization transgenic Tg(h2afva:h2afva-mCherry) zebrafish embryo, using different methods (view along the rotational axis, scale bar 100 , 10 in the inset): (d, e) acquired raw data, (f–i) fusion performed by (f) averaging, (g) entropy-weighted averaging, (h) 3D multi-view deconvolution and (i) plane-wise multi-view deconvolution (10 iterations). (Dell T6100, Intel E5-2630 @2.3 GHz 2 processors, 64 GB RAM; Graphics card: Nvidia GeForce GTX TITAN Black)
Mentions: The primary goal of MV fusion is the improvement of the poor axial resolution in a single 3D dataset using the superior lateral resolution of an additional, overlapping dataset, and not necessarily to improve resolution beyond the intrinsic lateral resolution. We therefore approximated the full 3D point spread function (PSF) with a 2D PSF, neglecting one lateral component (along the rotation axis), and processed each plane orthogonal to the rotation axis independently (Fig. 1a). Memory requirements were thereby reduced by the number of lines read out from the camera chip, i.e. typically 100–1000 fold (Fig. 1b). This allowed us to implement the entire MV deconvolution on a GPU. Taking advantage of three CUDA (Compute Unified Device Architecture) streams, we interleaved GPU computations with data transfers, such that not only expensive copying to and from GPU memory, but also reading and writing data from and to the hard drive came without additional cost (Supplementary Note S2). Compared with 3D MV deconvolution, with and without GPU support, we thereby reduced processing times by a factor of up to 25 and 75, respectively (Fig. 1c, Supplementary Table S1), while producing comparable results.Fig. 1.

Bottom Line: State-of-the-art multi-view (MV) deconvolution simultaneously fuses and deconvolves the images in 3D, but processing takes a multiple of the acquisition time and constitutes the bottleneck in the imaging pipeline.Here, we show that MV deconvolution in 3D can finally be achieved in real-time by processing cross-sectional planes individually on the massively parallel architecture of a graphics processing unit (GPU).Our approximation is valid in the typical case where the rotation axis lies in the imaging plane.

View Article: PubMed Central - PubMed

Affiliation: Max Planck Institute of Molecular Cell Biology and Genetics, 01307 Dresden, Germany.

No MeSH data available.


Related in: MedlinePlus