gpu-based deconvolution freeware?
Locked
|
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****
Hi everyone!
Does anyone know of any free GPU-based deconvolution software packages? Figured it’s worth asking before shelling out multiple thousands of dollars… Thanks! Uri
|
 
|
Aymeric FOUQUIER D'HEROUËL |
Re: gpu-based deconvolution freeware?
|
|
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****
Hi Uri,
We have developed a CUDA-based one, available on https://github.com/bobpepin/YacuDecu. It has both a Matlab and a C interface, but no dedicated GUI. Give it a try, might save you a buck.
Aymeric
Dr. Aymeric d'Hérouël
Research Associate UNIVERSITÉ DU LUXEMBOURG LUXEMBOURG CENTRE FOR SYSTEMS BIOMEDICINE Campus Belval | House of Biomedicine II 6, avenue du Swing L-4367 Esch-sur-Alzette T +352 46 66 44 6723 F +352 46 66 44 3 6723 This message is confidential and may contain privileged information. It is intended for the named recipient only. If you receive it in error please notify me and permanently delete the original message and any copies.
|
|
|
JOEL B. SHEFFIELD |
Re: gpu-based deconvolution freeware?
|
|
In reply to this post by Uri Manor
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****
Depending on your needs, there are several packages available for ImageJ. I have had pretty good success using the ones developed by Bob Dougherty that you can look up on the ImageJ plugins resource. JoelJoel B. Sheffield, Ph.D Department of Biology Temple University Philadelphia, PA 19122 Voice: 215 204 8839 e-mail: [hidden email] URL: http://tinyurl.com/khbouft On Thu, Oct 20, 2016 at 3:43 AM, Uri Manor <[hidden email]> wrote: ***** To join, leave or search the confocal microscopy listserv, go to: http://lists.umn.edu/cgi-bin/ |
|
|
George McNamara |
Re: gpu-based deconvolution freeware?
|
|
In reply to this post by Uri Manor
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****
Hi Uri, On 10/20/2016 2:43 AM, Uri Manor wrote:
***** To join, leave or search the confocal microscopy listserv, go to: http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on http://www.imgur.com and include the link in your posting. ***** Hi everyone! -- George McNamara, PhD Houston, TX 77054 [hidden email] https://www.linkedin.com/in/georgemcnamara https://works.bepress.com/gmcnamara/75/ http://www.ncbi.nlm.nih.gov/myncbi/browse/collection/44962650 |
|
|
Cole, Richard W (HEALTH) |
Re: gpu-based deconvolution freeware?
|
|
In reply to this post by Uri Manor
*****
To join, leave or search the confocal microscopy listserv, go to: http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on http://www.imgur.com and include the link in your posting. ***** Uri, I agree with George, try both “free” and commercial software with your data. Clarrie Brown (PMID: 24688321 PMCID: PMC3942261 DOI: 10.7171/jbt.14-2501-002) and Paul Goodwin (PMID: 24974028 DOI: 10.1016/B978-0-12-420138-5.00010-0) have published papers on the quantitative deconvolution in the last couple of years. Autoquant does have a GPU based package (release at Neuro I believe)-- no financial interest. I would caution that no upfront cost does not always equate to “free” and I will leave it at that. My 2¢ Rich Richard Cole Research Scientist V Director: Advanced Light Microscopy & Image Analysis Core Wadsworth Center Research Assistant Professor Dept. of Biomedical Sciences School of Public Health State University of New York 120 New Scotland Avenue, Albany N.Y. 12208 518-474-7048 Phone 518-408-1730 Fax Website http://www.wadsworth.org/research/cores/alm twitter.com/microscopejock From: George McNamara [mailto:[hidden email]] Sent: Thursday, October 20, 2016 9:38 AM Subject: Re: gpu-based deconvolution freeware? ***** To join, leave or search the confocal microscopy listserv, go to: http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on http://www.imgur.com and include the link in your posting. ***** Hi Uri, I strongly encourage you to demo both of the commercial systems, microvolution.com and SVI.nl Huygens GPU deconvolution. I am impressed by microvolution - and not just because my images are on their gallery and home page: fast and quantitative (disclosure: no financial interest). They could set you up so that anyone with microscopy data at Salk could batch process on a server/cluster. I assume SVI.nl can batch GPU deconvolve. You can ask each company which has quantitative algorithms, but best to test both speed and accuracy yourself (accuracy includes both correct structures, avoiding incorrect structures, and conservation of energy). There was a thread here on the confocal listserv -- the last message in that thread was: Subject: Re: GPU-based deconvolution From: George McNamara mailto:[hidden email] Reply-To: mailto:[hidden email] Date: Fri, 22 Aug 2014 08:17:43 -0500 Content-Type: text/plain Part of the text pasted later. A summer student in 2013 did a comparison of several deconvolvers - see https://works.bepress.com/gmcnamara/57/ This was before Bruce&Butte became http://www.microvolution.com and SVI.nl added GPU deconvolution to Huygens. I do not know if AutoQuant has GPU (I don't see any google hits for them with gpu deconvolution). 3i's systems page does not mention GPU on their systems page, https://www.intelligent-imaging.com/systems.php their founders discovered the T-cell immune synapse because of image processing widefield imaging, so they've been doing deblurring/deconvolution a long time. // Zanella 2013 Sci Rep DOI: 10.1038/srep02523 Schmid 2015 Bioinformatics doi: 10.1093/bioinformatics/btv387 UNC - Clarity Deconvolution Library https://github.com/CISMM/Clarity Part of the 2014 thread: If COSMOS is no longer being developed by Prof. Preza, all > the more reason > for you to adopt it! It is using a GNU General Public License, > http://cirl.memphis.edu/cosmos_faqs.php > > Likewise the Clarity deconvolution library at UNC is no > longer being > developed, consider taking over that. > > Adam Hoppe told me several months ago that he is > continuing to work on > 3DFSR, so consider contacting him at > http://www.sdstate.edu/chem/ > faculty/adam-hoppe/ > > I forgot to mention in yesterday's email the > BoaDeconvolution -- already > Parallel Deconvolution -- at > http://pawliczek.net.pl/deconvolution/ > http://onlinelibrary.wiley.com/doi/10.1002/jemt.20773/abstract there's also potential for Intel's Knights Landing CPU, http://www.nextplatform.com/2016/06/20/intel-knights-landing-yields-big-bang-buck-jump/ especially if Salk already has or is getting a KNL cluster (i.e. for genomics, single cell RNAseq, etc, that has compute resources available for it). enjoy, George On 10/20/2016 2:43 AM, Uri Manor wrote: ***** To join, leave or search the confocal microscopy listserv, go to: http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on http://www.imgur.com and include the link in your posting. ***** Hi everyone! Does anyone know of any free GPU-based deconvolution software packages? Figured it’s worth asking before shelling out multiple thousands of dollars… Thanks! Uri Uri Manor, Ph.D. DIRECTOR WAITT ADVANCED BIOPHOTONICS CORE PH (858) 453-4100 x2119 E mailto:[hidden email] W http://bpho.salk.edu -- George McNamara, PhD Houston, TX 77054 mailto:[hidden email] https://www.linkedin.com/in/georgemcnamara https://works.bepress.com/gmcnamara/75/ http://www.ncbi.nlm.nih.gov/myncbi/browse/collection/44962650 |
|
|
Brian Northan |
Re: gpu-based deconvolution freeware?
|
|
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****
Hi Rich "I would caution that no upfront cost does not always equate to “free” and I will leave it at that." Keep in mind there is a lot of great work being done in the open source community, and it needs support. Open source deconvolution is still lagging behind the commercial products, in terms of usability, however there are efforts underway to address that. One effort is a project called ImageJ ops, led by Kevin Eliceiri and Loci, it is basically a flexible math and image processing library for ImageJ, it will eventually include Deconvolution, but it's still a work in progress. I stress that closed source solutions have their weaknesses too, especially in terms of testing and documentation. Only licensed users of commercial products can run tests, while anyone can run tests on open source products. For example the testing and benchmarking of the commercial GPU products leaves a lot to be desired. Autoquant GPU - http://www.mediacy.com/autoquantx3/modules It looks like they compared a XEON e5506 (8 year old processor??) to new high(ish) end GPUs. These comparisons really should be done with state-of-the-art hardware on both ends at the same price point. It seems the AQI-GPU version got apr. 3x speed up. However there are no hard numbers reported. It's unclear what image sizes were used. Huygens GPU - https://svi.nl/HuygensGPU It looks like they compared with a Intel Core i5. Again it would be nice to see a benchmark for a high end CPU configuration. They got up to 29x with high end GPUs. I'm not sure what size of image they used either, and if these benchmarks are relevant to all image sizes. They do seem to have an app that lets you benchmark yourself (if you have a license, I assume). Microvolution - http://www.microvolution.com/technology They show comparisons against "mystery" company A and B and report speed increases up to 200x. Keep in mind they may not be comparing the same algorithm. Huygens and Autoquant both implement extended version of MLE with many constraints. Microvolution _may_ be a bare bones implementation. Huygens (I believe) and Autoquant also use "acceleration". 10 iterations of an accelerated MLE algorithm, is not the same as 10 iterations of straight forward MLE. Also the "missing filaments" example is _potentially_ deceiving, depending on whether they are showing a slice, or a 3D projection. If the former the filaments could simply be moving in and out of that plane. Brian On Fri, Oct 21, 2016 at 8:05 AM, Cole, Richard W (HEALTH) <[hidden email]> wrote: ***** |
|
|
George McNamara |
Re: gpu-based deconvolution freeware?
|
|
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****
Hi Brian, I suggest you buy 3 NVidia Pascal TITAN X or GTX 1080 Founders Edition cards, or wait for January 2017 for GTX 1080i, for one of your PC's (four if you can find a PC that fits them all -- or use your institutions GPU cluster -- hopefully they will have installed state of the art) and arrange one week demo licenses from each of AutoQuant, SCI and Microvolution. Find out from each what settings result in their most quantitative data, run the tests, and post the raw data and deconvolved data, and timings online. http://www.geforce.com/hardware/10series/titan-x-pascal http://www.geforce.com/hardware/10series/geforce-gtx-1080 You are welcome to use data I posted online, such as the single molecule FISH data at https://works.bepress.com/gmcnamara/64/ (I have data sets, and some results on several pages). I lobbied a Microvolution executive to work with Cell Image
Library http://www.cellimagelibrary.org/ to enable anyone who
uploads their data to CIL with "CC0" (copyright zero) to get
quantitative deconvolution for free. For anyone who does not want
to go "CC0" there could be a payment mechanism with CIL and the
vendor splitting the revenue. Has not happened so far. Of course
it does not have to be CIL - Brian, you could set up your own
"Cloud" to deconvolve and display people's data. Maybe Jeff Bezos
will donate to you and exabyte of AWS storage, and 1000 Pascal
TITAN X to you. So, I will update my proposal for each of the GPU deconvolution companies, and freeware initiatives (anyone with Intel Knights Landing/Corner/Hills or Xeons are welcome too): ==> Please work with Cell Image Library and/or similar light
microscope data sites to enable users who choose "CC0" to get
quantitative deconvolution for free (and data and results
permanently archived publicly available at the site). I also have suggestions for all vendors and freewarers in microscopy deconvolution: 1. implement quantitative joint spatial deconvolution and spectral unmixing - unlimited channels, unlimited pixels, voxels (and yes, that will ultimately mean quantitative stitching), a'la: Three-Dimensional Reconstruction of Three-Way FRET Microscopy
Improves Imaging of Multiple Protein-Protein Interactions. 2. quantitative brightfield "quantitative phase imaging"
deconvolution (which the founders of AutoQuant demonstrated a long
time ago). QPi, QPm, shown long ago by Keith Nugent et al in IATIA
(now Ultima Capital). Some other companies and labs have published
3. quantitative brightfield histology "stains" deconvolution (spatial and spectral).
I also encourage all the microscope manufacturers - both the "big 4" and the many others -- to enable "instant gratification" quantitative spatial deconvolution and spectral unmixing in their imaging products. Sincerely, George On 11/18/2016 5:06 AM, Brian Northan
wrote:
***** To join, leave or search the confocal microscopy listserv, go to: http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on http://www.imgur.com and include the link in your posting. ***** -- George McNamara, PhD Houston, TX 77054 [hidden email] https://www.linkedin.com/in/georgemcnamara https://works.bepress.com/gmcnamara/75/ http://www.ncbi.nlm.nih.gov/myncbi/browse/collection/44962650 |
|
|
Brian Northan |
Re: gpu-based deconvolution freeware?
|
|
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****
Thanks for the info George By the way, my perspective is from that of a developer, not a user. I should have mentioned that I occassionally contribute to ImageJ, and use ImageJ based code as libraries for my other projects. My long term goal is to help improve ImageJ deconvolution. I've been quite busy the last few months with other projects, so haven't had the chance to work on deconvolution lately. However I'd love to see ImageJ get more funding, it would help all of us. On Fri, Nov 18, 2016 at 10:22 AM, George McNamara <[hidden email]> wrote:
|
| Free forum by Nabble | Edit this page |