Practical Imaging Guides & Tutorials?

[Kalen K]

Are there any detailed yet practical imaging guides or tutorials? Overall, what are some of the best resources for an 3D imaging? I am aware of the resources on http://rsbweb.nih.gov/ij/docs/index.html

*Issue:
In general, It seems that imaging information is either too basic/vague, or too mathematical/abstract. Thus, making it difficult to achieve usable data for the average user, especially for 3D biological image sets.
When multiple functions or algorithms are available, how does one know which to use? What is the best place to find the advantages/disadvantages of similar functions or procedures?

*Basic Imaging Steps:
In general, these are the basic imaging steps as I understand it. Please correct me if I'm wrong.

1) Normalization: normalize image intensity
2) Filtering: filter image by removing background noise
What function should be used?
3) Selection: select the region of interest using thresholding or manual selection to create a binary image or ROI selection
How is this correctly achieved for 3D z-stacked images? How does one fix the selected 'islands'? ...watershedding? What if the entire region of interest can't be selected, that is thresholding misses some regions?
3) Calibration: calibrate image using scale bar or other methods.
5) Quantification: calculate the desired parameter

*Data Type:
I'm currently working with z-stack confocal microscopy data of neurons.


I'm sure the above questions are broad, but any direction to help me further learn proper biological imaging techniques would be very appreciated.

-Kalen

[Thomas Boudier]

Hi,

I've written a small tutorial for imagej conf 2008 :

http://imagejdocu.tudor.lu/doku.php?id=tutorial:working:3d_image_processing_and_analysis_with_imagej

do not hesitate if you have comments or suggestions to improve it.

Best,

Thomas


Kalen K a écrit :

> Are there any detailed yet practical imaging guides or tutorials? Overall, what are some of the best resources for an 3D imaging? I am aware of the resources on http://rsbweb.nih.gov/ij/docs/index.html
>
> *Issue:
> In general, It seems that imaging information is either too basic/vague, or too mathematical/abstract. Thus, making it difficult to achieve usable data for the average user, especially for 3D biological image sets.
> When multiple functions or algorithms are available, how does one know which to use? What is the best place to find the advantages/disadvantages of similar functions or procedures?
>
> *Basic Imaging Steps:
> In general, these are the basic imaging steps as I understand it. Please correct me if I'm wrong.
>
> 1) Normalization: normalize image intensity
> 2) Filtering: filter image by removing background noise
> What function should be used?
> 3) Selection: select the region of interest using thresholding or manual selection to create a binary image or ROI selection
> How is this correctly achieved for 3D z-stacked images? How does one fix the selected 'islands'? ...watershedding? What if the entire region of interest can't be selected, that is thresholding misses some regions?
> 3) Calibration: calibrate image using scale bar or other methods.
> 5) Quantification: calculate the desired parameter
>
> *Data Type:
> I'm currently working with z-stack confocal microscopy data of neurons.
>
>
> I'm sure the above questions are broad, but any direction to help me further learn proper biological imaging techniques would be very appreciated.
>
> -Kalen
>
>

--
   /**********************************************************/
      Thomas Boudier, MCU Université Pierre et Marie Curie,
      IFR 83. Bat B 7ème étage, porte 706D, Jussieu.
      Tel : [phone]  Fax : [phone]
/*******************************************************/

[Joel Sheffield]

HI,

For general background, you might take a look at
http:\\micro\magnet\fsu.edu\digitalimaging -

> Are there any detailed yet practical imaging guides or tutorials?
> Overall, what are some of the best resources for an 3D imaging? I am
> aware of the resources on http://rsbweb.nih.gov/ij/docs/index.html
>
> *Issue:
> In general, It seems that imaging information is either too
> basic/vague, or too mathematical/abstract. Thus, making it difficult
> to achieve usable data for the average user, especially for 3D
> biological image sets.
> When multiple functions or algorithms are available, how does one
> know which to use? What is the best place to find the
> advantages/disadvantages of similar functions or procedures?
>
> *Basic Imaging Steps:
> In general, these are the basic imaging steps as I understand it.
> Please correct me if I'm wrong.
>
> 1) Normalization: normalize image intensity
> 2) Filtering: filter image by removing background noise
> What function should be used?
> 3) Selection: select the region of interest using thresholding or
> manual selection to create a binary image or ROI selection
> How is this correctly achieved for 3D z-stacked images? How does one
> fix the selected 'islands'? ...watershedding? What if the entire
> region of interest can't be selected, that is thresholding misses
> some regions?
> 3) Calibration: calibrate image using scale bar or other methods.
> 5) Quantification: calculate the desired parameter
>
> *Data Type:
> I'm currently working with z-stack confocal microscopy data of
> neurons.
>
>
> I'm sure the above questions are broad, but any direction to help me
> further learn proper biological imaging techniques would be very
> appreciated.
>
> -Kalen

--
Joel B. Sheffield, Ph.D.
Biology Department, Temple University
1900 North 12th Street
Philadelphia, PA 19122
[hidden email]  
[phone], fax [phone]
http://astro.temple.edu/~jbs

[John Oreopoulos]

Dear Kalen, your questions are very broad, but very important at the same time. I can understand your frustrations about the information out there either being too basic or too advanced - I thought the same thing when starting out my graduate work, and eventually after much searching I was able to build up a good library of book chapters and journal articles that land somewhere in that intermediate level.

You posted your questions to the ImageJ listsever (as opposed to the confocal microscopy listserver), so it sounds like your currently more interested in the analysis of biological images instead of their acquisition. But image analysis and manipulation should never be a substitute or remedy for proper image acquisition practices. How you acquire your images can greatly affect how you analyze them and the quality of the analysis.

The modern fluorescence microscope is a bit a of a beast that can provide a wealth of multi-dimensional data in a single image set (3 space dimensions [x,y,z], time, intensity, wavelength, lifetime, polarization, even the rate at which pixel intensities fluctuate contains valuable information about the fluorescence probes present in the sample). These dimensional parameters can combined to create additional parameters; for example, two space dimensions can give you areas, three space dimensions can give you volumes, space and time can be combined to create trajectories, velocities, and speed etc. Intensity is usually used to infer probe concentration (but see notes below regarding the dangers associated with this), and when intensity is combined with wavelength, you can test whether two spectrally distinct probes colocalize in two separate spectral channels. Intensity combined with space dimensions can be used to assess what I call "morphological descriptors" that quantify various shapes and sizes, distances between objects, image patterns, etc. These are usually complicated and require some kind of automated algorithm. Finally there are the so-called "F-techniques" that make use of intensity, time, wavelength, polarization and/or lifetime to access information that reports on the local molecular environment of the probes to infer binding, conformational changes, and other molecular interactions (FRET, FRAP, FLIM, FCS, etc.).

In terms of books that deal with the acquisition of biological images using fluorescence microscopy, I recommend the following:

Confocal Microscopy for Biologists (http://www.amazon.com/Confocal-Microscopy-Biologists-Alan-Hibbs/dp/0306484684)
Fundamentals of Light Microscopy and Electronic Imaging (http://www.amazon.ca/Fundamentals-Light-Microscopy-Electronic-Imaging/dp/047125391X)
Digital Microscopy 3rd Edition: Methods in Cell Biology (http://www.amazon.ca/Digital-Microscopy-Methods-Cell-Biology/dp/0123740258/ref=sr_1_1?ie=UTF8&s=books&qid=1270651214&sr=1-1)

In those books there are a few chapters that deal with how to handle 3D image sets as well.

Here are some good journal articles that are a bit more accessible:

Pearson, H., Top tips for taking images. Nature, 2007. 447(7141): p. 140-140.
Brown, C.M., Fluorescence microscopy - avoiding the pitfalls. Journal of Cell Science, 2007. 120(10): p. 1703-1705.
North, A.J., Seeing is believing? A beginners' guide to practical pitfalls in image acquisition. Journal of Cell Biology, 2006. 172(1): p. 9-18.

The bottom line is, it's important to understand the factors that affect image acquisition and image quality (diffraction, noise, sampling, background, detector saturation, etc) and how these factors in turn affect the measurements you plan to make using your images post-acquistion. There's a lot of potential data in a single image, and it's up to you to figure out how to best parse the dimensions mentioned above into a quantifiable form that can then be fed into an image processing algorithm.

Having said all that, now I can try to address the image analysis component of your questions. You asked:

> When multiple functions or algorithms are available, how does one know which to use? What is the best place to find the advantages/disadvantages of similar functions or procedures?

There is no simple answer to this question since it will depend on what you are imaging, how you've prepared and labeled it, and what information you want to get out of your images. Unfortunately, there is no simple recipe for processing biological confocal microscopy images. You need to specify exactly what it is you want to study. What are the questions about your particular biological sample that you want to answer with optical microscopy? Once you tell everyone here on this listserver (or the confocal microscopy listserver) what that is, someone might be able to give you a more specific direction (most likely other people before you have already tried to do the same thing you're thinking of doing, and perhaps an image processing algorithm has been developed). By stating the specific questions and goals about the imaging experiment, you'll also be able to determine the limitations of the strategy (see everything I said above about image acquisition) and if optical imaging is the best technique to address the issue in the first place.

You also provided a list of "basic imaging processing steps". I think people will have differing opinions on this. I am of the philosophy that if you just want to show an image in a publication (without quantifying it in some way, ie: the image just shows some obvious spatial pattern to prove a statement in the text like, "we observed a strong staining of the nucleus under conditions x and y"), you should adjust it as little as possible. At a minimum, you should subtract the background (measured as the average pixel value from a boxed region of interest that contains no cells) and adjust the brightness and contrast so that the features of interest show up well in printed form, and put a scale bar on there to let everyone know about the image size. If you're showing multiple images for comparison, try to present them with equal brightness and contrast settings.

Any additional filtering beyond that is up to you, but you have to be careful that additional filtering doesn't inadvertently create image artifacts. That can be done only if you understand how the filter works and what it's used for. You should be able to justify why additional filtering is required. Some people will also argue that it is necessary to deconvolve confocal microscopy images (see Jim Pawley's "Handbook of Biological Confocal Microscopy"), but this has not become a mainstream practice yet (either due to ignorance, laziness, or lack of computer resources to perform this processing). The main point I'm trying to make here is that you want to avoid accidental "image fraud". Some very clear guidelines have recently been given about this:

Rossner, M., How to guard against image fraud. Scientist, 2006. 20(3): p. 24-25.
Pearson, H., Image manipulation: CSI: cell biology. Nature, 2005. 434(7036): p. 952-953.
Rossner, M. and K.M. Yamada, What's in a picture? The temptation of image manipulation. Journal of Cell Biology, 2004. 166(1): p. 11-15.
Cromey, D.W., Digital Image Ethics. Microscopy Today, 2009. doi: 10.1017/S1551929509000431
http://swehsc.pharmacy.arizona.edu/exppath/micro/digimage_ethics.php#recommended

You can't go wrong as long as you state explicitly somewhere in your manuscript/publication what manipulations you performed on your images to create your data and why you made those manipulations.

When it comes to quantification, the last step in your list, most often people are talking about quantifying pixel intensities to infer probe concentration or quantifying the amount of probe colocalization in two different spectral channels. The former is surprisingly non-trivial since there are many factors that affect the absolute intensity of any given pixel in your images. Even if you prepare all of your samples on the same day, the same way, and examine them with the same instrument, even then relative comparisons of pixel intensities can be tricky. I highly recommend reading Jim Pawley's "39 Steps":

Pawley, J., The 39 steps: A cautionary tale of quantitative 3-D fluorescence microscopy. Biotechniques, 2000. 28(5): p. 884-888.

Here are the best articles I've come across related to accurate colocalization and general image quantification:

Waters, J.C., Accuracy and precision in quantitative fluorescence microscopy. Journal of Cell Biology, 2009. 185(7): p. 1135-1148.
Zinchuk, V., O. Zinchuk, and T. Okada, Quantitative colocalization analysis of multicolor confocal immunofluorescence microscopy images: Pushing pixels to explore biological phenomena. Acta Histochemica Et Cytochemica, 2007. 40(4): p. 101-111.
Wolf, D.E., C. Samarasekera, and J.R. Swedlow, Quantitative analysis of digital microscope images, in Digital Microscopy, 3rd Edition. 2007, Elsevier Academic Press Inc: San Diego. p. 365-396.
Bolte, S. and F.P. Cordelieres, A guided tour into subcellular colocalization analysis in light microscopy. Journal of Microscopy-Oxford, 2006. 224: p. 213-232.
Costes, S.V., et al., Automatic and quantitative measurement of protein-protein colocalization in live cells. Biophysical Journal, 2004. 86(6): p. 3993-4003.


John Oreopoulos, BSc,
PhD Candidate
University of Toronto
Institute For Biomaterials and Biomedical Engineering
Centre For Studies in Molecular Imaging

On 2010-04-07, at 2:34 AM, Kalen K wrote:

> Are there any detailed yet practical imaging guides or tutorials? Overall, what are some of the best resources for an 3D imaging? I am aware of the resources on http://rsbweb.nih.gov/ij/docs/index.html
>
> *Issue:
> In general, It seems that imaging information is either too basic/vague, or too mathematical/abstract. Thus, making it difficult to achieve usable data for the average user, especially for 3D biological image sets.
> When multiple functions or algorithms are available, how does one know which to use? What is the best place to find the advantages/disadvantages of similar functions or procedures?
>
> *Basic Imaging Steps:
> In general, these are the basic imaging steps as I understand it. Please correct me if I'm wrong.
>
> 1) Normalization: normalize image intensity
> 2) Filtering: filter image by removing background noise
> What function should be used?
> 3) Selection: select the region of interest using thresholding or manual selection to create a binary image or ROI selection
> How is this correctly achieved for 3D z-stacked images? How does one fix the selected 'islands'? ...watershedding? What if the entire region of interest can't be selected, that is thresholding misses some regions?
> 3) Calibration: calibrate image using scale bar or other methods.
> 5) Quantification: calculate the desired parameter
>
> *Data Type:
> I'm currently working with z-stack confocal microscopy data of neurons.
>
>
> I'm sure the above questions are broad, but any direction to help me further learn proper biological imaging techniques would be very appreciated.
>
> -Kalen

[Kalen K]

Thank you all for your suggestions! I now have some great resources to build upon my imaging knowledge. Image analysis can be a difficult learning process due to its complexity and ease of unintentional errors. It can also be a 'distraction' against the biologist’s main interest of focusing on the biology.

Thomas, I have used your tutorial before, and found it to be helpful. I probably need to revisit it in greater detail, and will keep you informed of any comments. Would the ImageJ wiki be a good place to add information and resources on image analysis to help people like me? Or is the wiki specifically for ImageJ?

Thanks Joel for that great website. I have my work cut out for me.

Thank you John for your wonderful input and suggestions, you raise some excellent points that will make me a better scientist. It's comforting to hear that I am not alone in my problem. One of the goals as a scientist is to have accurate data to obtain correct measurements which is only possible with a true understanding of image acquisition and analysis. Your resources and input will help me achieve this goal.


-Kalen

[Marko Usaj]

Dear community!

I recently started to investigate cell cytoskeleton. I finally got first images of it. Is here any good practice, tutorial, etc. for achieving good and quality images of cytoskeleton in terms of image acquisition, preprocessing, analysis... I heard in past that deconvolution helps...

Any quick tips of those who works on such images to right directions will be very appreciated.

Many thanks all the best to all.

Marko
 
-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of John Oreopoulos
Sent: Wednesday, April 07, 2010 11:32 PM
To: [hidden email]
Subject: Re: Practical Imaging Guides & Tutorials?

Dear Kalen, your questions are very broad, but very important at the same time. I can understand your frustrations about the information out there either being too basic or too advanced - I thought the same thing when starting out my graduate work, and eventually after much searching I was able to build up a good library of book chapters and journal articles that land somewhere in that intermediate level.

You posted your questions to the ImageJ listsever (as opposed to the confocal microscopy listserver), so it sounds like your currently more interested in the analysis of biological images instead of their acquisition. But image analysis and manipulation should never be a substitute or remedy for proper image acquisition practices. How you acquire your images can greatly affect how you analyze them and the quality of the analysis.

The modern fluorescence microscope is a bit a of a beast that can provide a wealth of multi-dimensional data in a single image set (3 space dimensions [x,y,z], time, intensity, wavelength, lifetime, polarization, even the rate at which pixel intensities fluctuate contains valuable information about the fluorescence probes present in the sample). These dimensional parameters can combined to create additional parameters; for example, two space dimensions can give you areas, three space dimensions can give you volumes, space and time can be combined to create trajectories, velocities, and speed etc. Intensity is usually used to infer probe concentration (but see notes below regarding the dangers associated with this), and when intensity is combined with wavelength, you can test whether two spectrally distinct probes colocalize in two separate spectral channels. Intensity combined with space dimensions can be used to assess what I call "morphological descriptors" that quantify various shapes and sizes, distances between objects, image patterns, etc. These are usually complicated and require some kind of automated algorithm. Finally there are the so-called "F-techniques" that make use of intensity, time, wavelength, polarization and/or lifetime to access information that reports on the local molecular environment of the probes to infer binding, conformational changes, and other molecular interactions (FRET, FRAP, FLIM, FCS, etc.).

In terms of books that deal with the acquisition of biological images using fluorescence microscopy, I recommend the following:

Confocal Microscopy for Biologists (http://www.amazon.com/Confocal-Microscopy-Biologists-Alan-Hibbs/dp/0306484684)
Fundamentals of Light Microscopy and Electronic Imaging (http://www.amazon.ca/Fundamentals-Light-Microscopy-Electronic-Imaging/dp/047125391X)
Digital Microscopy 3rd Edition: Methods in Cell Biology (http://www.amazon.ca/Digital-Microscopy-Methods-Cell-Biology/dp/0123740258/ref=sr_1_1?ie=UTF8&s=books&qid=1270651214&sr=1-1)

In those books there are a few chapters that deal with how to handle 3D image sets as well.

Here are some good journal articles that are a bit more accessible:

Pearson, H., Top tips for taking images. Nature, 2007. 447(7141): p. 140-140.
Brown, C.M., Fluorescence microscopy - avoiding the pitfalls. Journal of Cell Science, 2007. 120(10): p. 1703-1705.
North, A.J., Seeing is believing? A beginners' guide to practical pitfalls in image acquisition. Journal of Cell Biology, 2006. 172(1): p. 9-18.

The bottom line is, it's important to understand the factors that affect image acquisition and image quality (diffraction, noise, sampling, background, detector saturation, etc) and how these factors in turn affect the measurements you plan to make using your images post-acquistion. There's a lot of potential data in a single image, and it's up to you to figure out how to best parse the dimensions mentioned above into a quantifiable form that can then be fed into an image processing algorithm.

Having said all that, now I can try to address the image analysis component of your questions. You asked:

> When multiple functions or algorithms are available, how does one know which to use? What is the best place to find the advantages/disadvantages of similar functions or procedures?

There is no simple answer to this question since it will depend on what you are imaging, how you've prepared and labeled it, and what information you want to get out of your images. Unfortunately, there is no simple recipe for processing biological confocal microscopy images. You need to specify exactly what it is you want to study. What are the questions about your particular biological sample that you want to answer with optical microscopy? Once you tell everyone here on this listserver (or the confocal microscopy listserver) what that is, someone might be able to give you a more specific direction (most likely other people before you have already tried to do the same thing you're thinking of doing, and perhaps an image processing algorithm has been developed). By stating the specific questions and goals about the imaging experiment, you'll also be able to determine the limitations of the strategy (see everything I said above about image acquisition) and if optical imaging is the best technique to address the issue in the first place.

You also provided a list of "basic imaging processing steps". I think people will have differing opinions on this. I am of the philosophy that if you just want to show an image in a publication (without quantifying it in some way, ie: the image just shows some obvious spatial pattern to prove a statement in the text like, "we observed a strong staining of the nucleus under conditions x and y"), you should adjust it as little as possible. At a minimum, you should subtract the background (measured as the average pixel value from a boxed region of interest that contains no cells) and adjust the brightness and contrast so that the features of interest show up well in printed form, and put a scale bar on there to let everyone know about the image size. If you're showing multiple images for comparison, try to present them with equal brightness and contrast settings.

Any additional filtering beyond that is up to you, but you have to be careful that additional filtering doesn't inadvertently create image artifacts. That can be done only if you understand how the filter works and what it's used for. You should be able to justify why additional filtering is required. Some people will also argue that it is necessary to deconvolve confocal microscopy images (see Jim Pawley's "Handbook of Biological Confocal Microscopy"), but this has not become a mainstream practice yet (either due to ignorance, laziness, or lack of computer resources to perform this processing). The main point I'm trying to make here is that you want to avoid accidental "image fraud". Some very clear guidelines have recently been given about this:

Rossner, M., How to guard against image fraud. Scientist, 2006. 20(3): p. 24-25.
Pearson, H., Image manipulation: CSI: cell biology. Nature, 2005. 434(7036): p. 952-953.
Rossner, M. and K.M. Yamada, What's in a picture? The temptation of image manipulation. Journal of Cell Biology, 2004. 166(1): p. 11-15.
Cromey, D.W., Digital Image Ethics. Microscopy Today, 2009. doi: 10.1017/S1551929509000431
http://swehsc.pharmacy.arizona.edu/exppath/micro/digimage_ethics.php#recommended

You can't go wrong as long as you state explicitly somewhere in your manuscript/publication what manipulations you performed on your images to create your data and why you made those manipulations.

When it comes to quantification, the last step in your list, most often people are talking about quantifying pixel intensities to infer probe concentration or quantifying the amount of probe colocalization in two different spectral channels. The former is surprisingly non-trivial since there are many factors that affect the absolute intensity of any given pixel in your images. Even if you prepare all of your samples on the same day, the same way, and examine them with the same instrument, even then relative comparisons of pixel intensities can be tricky. I highly recommend reading Jim Pawley's "39 Steps":

Pawley, J., The 39 steps: A cautionary tale of quantitative 3-D fluorescence microscopy. Biotechniques, 2000. 28(5): p. 884-888.

Here are the best articles I've come across related to accurate colocalization and general image quantification:

Waters, J.C., Accuracy and precision in quantitative fluorescence microscopy. Journal of Cell Biology, 2009. 185(7): p. 1135-1148.
Zinchuk, V., O. Zinchuk, and T. Okada, Quantitative colocalization analysis of multicolor confocal immunofluorescence microscopy images: Pushing pixels to explore biological phenomena. Acta Histochemica Et Cytochemica, 2007. 40(4): p. 101-111.
Wolf, D.E., C. Samarasekera, and J.R. Swedlow, Quantitative analysis of digital microscope images, in Digital Microscopy, 3rd Edition. 2007, Elsevier Academic Press Inc: San Diego. p. 365-396.
Bolte, S. and F.P. Cordelieres, A guided tour into subcellular colocalization analysis in light microscopy. Journal of Microscopy-Oxford, 2006. 224: p. 213-232.
Costes, S.V., et al., Automatic and quantitative measurement of protein-protein colocalization in live cells. Biophysical Journal, 2004. 86(6): p. 3993-4003.


John Oreopoulos, BSc,
PhD Candidate
University of Toronto
Institute For Biomaterials and Biomedical Engineering
Centre For Studies in Molecular Imaging

On 2010-04-07, at 2:34 AM, Kalen K wrote:

> Are there any detailed yet practical imaging guides or tutorials? Overall, what are some of the best resources for an 3D imaging? I am aware of the resources on http://rsbweb.nih.gov/ij/docs/index.html
>
> *Issue:
> In general, It seems that imaging information is either too basic/vague, or too mathematical/abstract. Thus, making it difficult to achieve usable data for the average user, especially for 3D biological image sets.
> When multiple functions or algorithms are available, how does one know which to use? What is the best place to find the advantages/disadvantages of similar functions or procedures?
>
> *Basic Imaging Steps:
> In general, these are the basic imaging steps as I understand it. Please correct me if I'm wrong.
>
> 1) Normalization: normalize image intensity
> 2) Filtering: filter image by removing background noise
> What function should be used?
> 3) Selection: select the region of interest using thresholding or manual selection to create a binary image or ROI selection
> How is this correctly achieved for 3D z-stacked images? How does one fix the selected 'islands'? ...watershedding? What if the entire region of interest can't be selected, that is thresholding misses some regions?
> 3) Calibration: calibrate image using scale bar or other methods.
> 5) Quantification: calculate the desired parameter
>
> *Data Type:
> I'm currently working with z-stack confocal microscopy data of neurons.
>
>
> I'm sure the above questions are broad, but any direction to help me further learn proper biological imaging techniques would be very appreciated.
>
> -Kalen