Dear all,
I need to quantify the amount of aggregation of a certain protein in nuclei. Visually the difference is obvious but how to put that into numbers? I thought about some kind of wavelet analysis since the fluctuations in the aggregated state have a different frequency but I didn't find the right tool or the right parameters. Has anybody encountered a similar problem? https://www.dropbox.com/sh/jknltkw2jmyeg1x/AAApo_VMu2-fe3ssvxZJXZu3a?dl=0 Best regards, Thomas ……………………………….. Thomas Lendl Image Analysis/FACS Specialist IMP Research Institute of Molecular Pathology Campus-Vienna-Biocenter 1 1030 Vienna, Austria -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html |
Hi Thomas,
if you limit the analysis to just the protein content, a simple coefficient of variation, aka relative standard deviation, of the intensity should already be able to highlight the difference in a numerical manner. I hope it helps. Giovanni -----Original Message----- From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Lendl,Thomas Sent: Freitag, 26. April 2019 11:17 To: [hidden email] Subject: Feature description of condensation Dear all, I need to quantify the amount of aggregation of a certain protein in nuclei. Visually the difference is obvious but how to put that into numbers? I thought about some kind of wavelet analysis since the fluctuations in the aggregated state have a different frequency but I didn't find the right tool or the right parameters. Has anybody encountered a similar problem? https://www.dropbox.com/sh/jknltkw2jmyeg1x/AAApo_VMu2-fe3ssvxZJXZu3a?dl=0 Best regards, Thomas ……………………………….. Thomas Lendl Image Analysis/FACS Specialist IMP Research Institute of Molecular Pathology Campus-Vienna-Biocenter 1 1030 Vienna, Austria -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html |
Hi Giovanni,
Thank you for the suggestion! I already tried that and there are differences in CV but not so pronounced. Both types show coarser intensity fluctuations (due to nucleoli and uneven intensity) that mess up the standard deviation. -----Original Message----- From: ImageJ Interest Group <[hidden email]> On Behalf Of Cardone, Giovanni Sent: Freitag, 26. April 2019 11:38 To: [hidden email] Subject: Re: Feature description of condensation Hi Thomas, if you limit the analysis to just the protein content, a simple coefficient of variation, aka relative standard deviation, of the intensity should already be able to highlight the difference in a numerical manner. I hope it helps. Giovanni -----Original Message----- From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Lendl,Thomas Sent: Freitag, 26. April 2019 11:17 To: [hidden email] Subject: Feature description of condensation Dear all, I need to quantify the amount of aggregation of a certain protein in nuclei. Visually the difference is obvious but how to put that into numbers? I thought about some kind of wavelet analysis since the fluctuations in the aggregated state have a different frequency but I didn't find the right tool or the right parameters. Has anybody encountered a similar problem? https://www.dropbox.com/sh/jknltkw2jmyeg1x/AAApo_VMu2-fe3ssvxZJXZu3a?dl=0 Best regards, Thomas ……………………………….. Thomas Lendl Image Analysis/FACS Specialist IMP Research Institute of Molecular Pathology Campus-Vienna-Biocenter 1 1030 Vienna, Austria -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html |
Hi Thomas,
have you considered asking this question on the forum? :) https://forum.image.sc I'd imagine you'll get a broader range of suggestions there, in particular if you're not limited to using ImageJ (1.x). There are a bunch of texture-based image feature sets (e.g. Haralick, Zernike) that are more easily accessible in other tools. CellProfiler offers some of these measurements, and they're implemented in ImageJ-Ops, accessible either via scripting, or e.g. in KNIME via the Feature Calculator node. Cheers Jan On 26.04.19 12:25, Lendl,Thomas wrote: > Hi Giovanni, > > Thank you for the suggestion! I already tried that and there are differences in CV but not so pronounced. Both types show coarser intensity fluctuations (due to nucleoli and uneven intensity) that mess up the standard deviation. > > > -----Original Message----- > From: ImageJ Interest Group <[hidden email]> On Behalf Of Cardone, Giovanni > Sent: Freitag, 26. April 2019 11:38 > To: [hidden email] > Subject: Re: Feature description of condensation > > Hi Thomas, > > if you limit the analysis to just the protein content, a simple coefficient of variation, aka relative standard deviation, of the intensity should already be able to highlight the difference in a numerical manner. > > I hope it helps. > Giovanni > > > -----Original Message----- > From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Lendl,Thomas > Sent: Freitag, 26. April 2019 11:17 > To: [hidden email] > Subject: Feature description of condensation > > Dear all, > > I need to quantify the amount of aggregation of a certain protein in nuclei. Visually the difference is obvious but how to put that into numbers? I thought about some kind of wavelet analysis since the fluctuations in the aggregated state have a different frequency but I didn't find the right tool or the right parameters. Has anybody encountered a similar problem? > > https://www.dropbox.com/sh/jknltkw2jmyeg1x/AAApo_VMu2-fe3ssvxZJXZu3a?dl=0 > > Best regards, > Thomas > > ……………………………….. > Thomas Lendl > Image Analysis/FACS Specialist > > IMP > Research Institute of Molecular Pathology Campus-Vienna-Biocenter 1 > 1030 Vienna, Austria > > > > > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html > > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html > > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html > -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html |
In reply to this post by Lendl,Thomas
Good day Thomas,
you could try to enhance the differences between the states by bandpass-filtering. You may start with the setting "large=10" and "small=0" and everything else unchecked. With your images in 32bit I get StdDev_a = 9.99; StdDev_s = 8.01; If this isn't sufficient I would try evaluating the Fourier Power Spectra of both images and use the "Radial Profile"-plugin to get descriptors. Regards Herbie :::::::::::::::::::::::::::::::::::::::::: Am 26.04.19 um 12:25 schrieb Lendl,Thomas: > Hi Giovanni, > > Thank you for the suggestion! I already tried that and there are differences in CV but not so pronounced. Both types show coarser intensity fluctuations (due to nucleoli and uneven intensity) that mess up the standard deviation. > > > -----Original Message----- > From: ImageJ Interest Group <[hidden email]> On Behalf Of Cardone, Giovanni > Sent: Freitag, 26. April 2019 11:38 > To: [hidden email] > Subject: Re: Feature description of condensation > > Hi Thomas, > > if you limit the analysis to just the protein content, a simple coefficient of variation, aka relative standard deviation, of the intensity should already be able to highlight the difference in a numerical manner. > > I hope it helps. > Giovanni > > > -----Original Message----- > From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Lendl,Thomas > Sent: Freitag, 26. April 2019 11:17 > To: [hidden email] > Subject: Feature description of condensation > > Dear all, > > I need to quantify the amount of aggregation of a certain protein in nuclei. Visually the difference is obvious but how to put that into numbers? I thought about some kind of wavelet analysis since the fluctuations in the aggregated state have a different frequency but I didn't find the right tool or the right parameters. Has anybody encountered a similar problem? > > https://www.dropbox.com/sh/jknltkw2jmyeg1x/AAApo_VMu2-fe3ssvxZJXZu3a?dl=0 > > Best regards, > Thomas > > ……………………………….. > Thomas Lendl > Image Analysis/FACS Specialist > > IMP > Research Institute of Molecular Pathology Campus-Vienna-Biocenter 1 > 1030 Vienna, Austria > > > > > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html > > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html > > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html > -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html |
In reply to this post by Lendl,Thomas
Hi Thomas
Have you tried measuring the multifractal dimensions or the lacunarity? Those are in the Fraclac plugin --Larry Lawrence M. Anovitz Senior Research Scientist Geochemistry and Interfacial Sciences Group Chemical Sciences Division Oak Ridge National Laboratory MS 6110, PO Box 2008 1 Bethel Valley Rd. Oak Ridge, TN, 37831-6110 865-574-5034 On 4/26/19, 5:20 AM, "Lendl,Thomas" <[hidden email]> wrote: Dear all, I need to quantify the amount of aggregation of a certain protein in nuclei. Visually the difference is obvious but how to put that into numbers? I thought about some kind of wavelet analysis since the fluctuations in the aggregated state have a different frequency but I didn't find the right tool or the right parameters. Has anybody encountered a similar problem? https://www.dropbox.com/sh/jknltkw2jmyeg1x/AAApo_VMu2-fe3ssvxZJXZu3a?dl=0 Best regards, Thomas ……………………………….. Thomas Lendl Image Analysis/FACS Specialist IMP Research Institute of Molecular Pathology Campus-Vienna-Biocenter 1 1030 Vienna, Austria -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html |
In reply to this post by Lendl,Thomas
Hi Thomas,
you should surely follow the suggestions from the others on using higher order texture descriptors for the analysis. I just want to add a comment: if you observe some factors that interfere with your analysis, try to adapt the analysis accordingly. Don't expect that more sophisticated methods necessarily compensate for those intensity fluctuations. Since the intensity fluctuates in a single cell, you could reduce the scale of the analysis to a smaller size, excluding clear outliers. Additionally, you could use a more robust statistics. Using your data as an example, for each image I measured the coefficient of variation on 32 subregions randomly extracted from the cells (radius=20 pixels, arbitrary chosen), excluding only the most obvious nucleoli. Since the distribution of the measurements was not approximating a Gaussian, I took the median. The result is a coefficient of variation of 0.21 and 0.17 for the 'aggregate' and 'smooth' images, respectively, with a dispersion of 0.02 (median absolute deviation). This is why I suggested to try such a simple approach, before going for more complex approaches. -----Original Message----- From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Lendl,Thomas Sent: Freitag, 26. April 2019 12:25 To: [hidden email] Subject: Re: Feature description of condensation Hi Giovanni, Thank you for the suggestion! I already tried that and there are differences in CV but not so pronounced. Both types show coarser intensity fluctuations (due to nucleoli and uneven intensity) that mess up the standard deviation. -----Original Message----- From: ImageJ Interest Group <[hidden email]> On Behalf Of Cardone, Giovanni Sent: Freitag, 26. April 2019 11:38 To: [hidden email] Subject: Re: Feature description of condensation Hi Thomas, if you limit the analysis to just the protein content, a simple coefficient of variation, aka relative standard deviation, of the intensity should already be able to highlight the difference in a numerical manner. I hope it helps. Giovanni -----Original Message----- From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Lendl,Thomas Sent: Freitag, 26. April 2019 11:17 To: [hidden email] Subject: Feature description of condensation Dear all, I need to quantify the amount of aggregation of a certain protein in nuclei. Visually the difference is obvious but how to put that into numbers? I thought about some kind of wavelet analysis since the fluctuations in the aggregated state have a different frequency but I didn't find the right tool or the right parameters. Has anybody encountered a similar problem? https://www.dropbox.com/sh/jknltkw2jmyeg1x/AAApo_VMu2-fe3ssvxZJXZu3a?dl=0 Best regards, Thomas ……………………………….. Thomas Lendl Image Analysis/FACS Specialist IMP Research Institute of Molecular Pathology Campus-Vienna-Biocenter 1 1030 Vienna, Austria -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html |
Giovanni!
"[...] I suggested to try such a simple approach, before going for more complex approaches." Fully agreed and supported! Perhaps I wasn't clear enough in my post. What I've finally computed is the *global* StdDev, i.e. that of the sample images as such (no ROIs or subregions) and that's pretty simple. Of course we need to see more images to be sure that such analysis is generally applicable. Regards Herbie ::::::::::::::::::::::::::::::::::::::::::::::: Am 26.04.19 um 18:05 schrieb Cardone, Giovanni: > Hi Thomas, > > you should surely follow the suggestions from the others on using higher order texture descriptors for the analysis. I just want to add a comment: if you observe some factors that interfere with your analysis, try to adapt the analysis accordingly. Don't expect that more sophisticated methods necessarily compensate for those intensity fluctuations. > Since the intensity fluctuates in a single cell, you could reduce the scale of the analysis to a smaller size, excluding clear outliers. Additionally, you could use a more robust statistics. > Using your data as an example, for each image I measured the coefficient of variation on 32 subregions randomly extracted from the cells (radius=20 pixels, arbitrary chosen), excluding only the most obvious nucleoli. Since the distribution of the measurements was not approximating a Gaussian, I took the median. The result is a coefficient of variation of 0.21 and 0.17 for the 'aggregate' and 'smooth' images, respectively, with a dispersion of 0.02 (median absolute deviation). This is why I suggested to try such a simple approach, before going for more complex approaches. > > > > -----Original Message----- > From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Lendl,Thomas > Sent: Freitag, 26. April 2019 12:25 > To: [hidden email] > Subject: Re: Feature description of condensation > > Hi Giovanni, > > Thank you for the suggestion! I already tried that and there are differences in CV but not so pronounced. Both types show coarser intensity fluctuations (due to nucleoli and uneven intensity) that mess up the standard deviation. > > > -----Original Message----- > From: ImageJ Interest Group <[hidden email]> On Behalf Of Cardone, Giovanni > Sent: Freitag, 26. April 2019 11:38 > To: [hidden email] > Subject: Re: Feature description of condensation > > Hi Thomas, > > if you limit the analysis to just the protein content, a simple coefficient of variation, aka relative standard deviation, of the intensity should already be able to highlight the difference in a numerical manner. > > I hope it helps. > Giovanni > > > -----Original Message----- > From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Lendl,Thomas > Sent: Freitag, 26. April 2019 11:17 > To: [hidden email] > Subject: Feature description of condensation > > Dear all, > > I need to quantify the amount of aggregation of a certain protein in nuclei. Visually the difference is obvious but how to put that into numbers? I thought about some kind of wavelet analysis since the fluctuations in the aggregated state have a different frequency but I didn't find the right tool or the right parameters. Has anybody encountered a similar problem? > > https://www.dropbox.com/sh/jknltkw2jmyeg1x/AAApo_VMu2-fe3ssvxZJXZu3a?dl=0 > > Best regards, > Thomas > > ……………………………….. > Thomas Lendl > Image Analysis/FACS Specialist > > IMP > Research Institute of Molecular Pathology Campus-Vienna-Biocenter 1 > 1030 Vienna, Austria > > > > > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html > > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html > > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html > > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html > -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html |
Free forum by Nabble | Edit this page |