A line scan is a one-dimensional image, either in the x or y direction. The time between adjacent line scans is the inverse of the sampling frequency. Do you know what that is? The line scan is usually displayed as a quasi-2D image but the 2nd dimension is time and not an x, y, or z coordinate.. Also, do you want to determine how intensity changes over time for each pixel in the line, a few adjacent pixels averaged together, or all pixels in the line averaged together? I would normally choose one of the latter two options. However, that decision depends on the heterogeneity of the structures/metabolic events occurring underneath your line. Since I don't know which CLSM you are using, I can only guess about how your line scans are displayed and how the analysis for them should be set up.

In imageJ you can manually specify the pixel dimensions of your scan image (which is probably in TIF format), typically shift+p is the shortcut. Here you can also specify the pixel unit as sec or msec, and manually adjust the pixel values to match the actual duration of the line scan. You need to have the line scan image rotated so that it reads from left to right (meaning individual lines are vertical). The scan duration based on your manually adjusted properties will be visible at the top of the image frame (in essence, only the x-axes pixel dimensions are interesting). It is important that you know the correct overall line scan duration to know what the correct pixel values must be (maybe your acquisition software reports this?), as this cannot be simply estimated from pixel numbers and dwell time due to line fly-back time and line shift time. 

Once you have the line scan image with the correct overall duration, and still with time on the left-to-rigth axis, you can draw a square covering the part of the trace you want to analyse and press 'plot profile'. This will give you numerical data in the form of a plot, which you can analyse in imageJ or other software, e.g. by fitting a mono-exponential to the recovery part of the plot to extract tau.

You can easily record the main parts of this work flow as a macro in imageJ, especially if you could get away with using the same line scan settings (at least scan duration) for all experiments.

Prudhvi, reading your question, my impression is that you might not be getting the required information from the imaging software. What software are you using?

As Gordon and Jan point out, the value you urgently need is the time-per-pixel (aka pixel dwell time). Depending on the software you are using, there are several ways in which you might be able to access this. Some programs let you set the actual pixel dwell time directly in the image/acquisition settings. Often this is also stored in the image header or meta information. Other programs will allow you to set the time-per-line, in which case you simply need to divide this line duration by the number of pixels. 

ImageJ is a good software for the initial quick check of the data, but from my experience, I would recommend extracting the intensity traces directly in the software environment that you will be using for further analysis. Python, Igor Pro, Matlab etc. all do a good job at this. Skipping ImageJ in the data analysis makes automation of the data processing easier.

Good luck!

Find whether your lineScan period is the product of Npixel*pixelTime (depends on what your scope understands under the term "dwell time"). If yes, just find the center of each stripe, take the horizontal coordinate and divide it by the number of Npixel (pixels recorded in one passage through the line, total number of columns of the stripe-like data). Finally, multiply this number * scanLinePeriod and add the product to each neuron time trace. 

I would reccoment you Two-Photon Processor, but it's a whil I have improved the available version and it is probably not there. In 2017 we will possibly release the second version including processing of GCaMP-filled neurons, but we have to debug it first.

Hello Gordon, Jan, Christian and Ondrej!

Thank you all for your answers. For more detailed information on how my analysis looks like, I've attached few figures here. Let me know if it helps in understanding my question better.

1) I am using Olympus multiphoton microscope for acquisition of my traces and ImageJ for the analysis. 

2) The image 'LineScan' shows the traces in red and green channels respectively with a line width of 10 on the green trace.

3) The image 'MetaData" shows the raw information of the image acquired. Here I am interested in the time per line option which I would use eventually in the setting up the scale.

4) In the image 'SetScale', I have set the values according to the metadata. (Please let me know if I have made any mistake in this)

5) Image 'PlotProfile' is the final plot of the trace.

Now, the question is, the value of time per line changes depending on the length of the line itself and it is a bit hard to maintain the same length of the line while scanning for all the experiments. This in turn leads to misalignment in the time points of all the traces.

Thank you so much for your time.

P.S. I am a novice to both live imaging and using two-photon microscope. All your suggestions are highly appreciated. Thank you!

Honestly, I guess That nobody cares about Precise length of the line or the linesPerSecond. We used to publish just saying "We used LineScan, sampling frequency 90-110Hz (mean + -std)".

Unfortunately, I can not recommend anything more, because I do not know Olympus scope at all.

It seems Olympus has changed their linescan implementation since I last used it.

If you want to average traces from different lines (although I see no reason to do that), you would need to ensure that the time per line will need to be constant. If you don't need to do that, it really doesn't matter. You analyze your traces to get the data you need (amplitude, rise times, decay times, etc.) and then continue with that.

For display purposes you will need to scale the traces to compensate for the timing difference. If you are using Igor Pro, that is done for you automatically, based on the scaling information you provide. With Matlab and Python you will need to figure out how to do this.

Hope that helps!

To me, it looks like you filled the ImageJ frame properties correctly, and the ImageJ graph duration (x axis) seems to correspond well with the 1545 ms frame duration in your meta data.

I am not sure what you specifically want to analyse?

Do you need to setup a fully automated procedure for 100s or 1000s of experiments, or can you get away with more manual analyses and aligment of data?

What are you trying to extract from the data - are you interested in rise time, decay kinetics and/or amplitude? 

In general, I don't agree at all that "nobody cares about the precise length of the line or the LinesPerSecond". If you are looking at subtle differences between groups, then this could have a large impact on your results and distort your analyses. This will of course depend on the expected, and observed, differences between groups.

Thank you, Jan for your discussion.

I do need to setup a fully automated procedure for 100s of line traces, but at this point, it is not the priority as I can get away with the manual analysis.

My question is more about the alignment of data from different traces. I am trying to look at the decay kinetics (the evacuation time) in particular. You are absolutely right in saying that the length of the line is significant when looking at subtle differences between groups. This is the goal of my study too. I am looking for the differences in the decay kinetics between two genotypes if any. 

So, how do I align multiple traces with different time points (as the length of the line is different for different traces) into an accurate representation and analysis?

Thank you once again for your time. 

If you can live with a somewhat manual analysis process, simply adjust the length of the line that defines your profile plot to start at the max intensity peak, so it only describes the decay part of the data. Here I would again suggest you rotate the image left, so that time is not on the y-axis but on the x-axis. Then you can use the rectangular tool instead of the line tool to define the region of the intensity plot, which is easier and avoids complications of a tilted line.

One important notion here is that if you use a single-exponential fit function, then this is independent on 100% accurate alignment of data: the tau will be the same no matter where you measure it on the decay curve (if the data actually follow a mono-exponential).

Basically, you draw a rectangular ROI across the decay part of the image, plot the profile, save the profile (as text), and open the .txt plot into the fitting tool of imageJ (analyze> tools> curve fitting), where you chose the monoexponential (or define yourself whatever function you want). The whole process is much faster than it sounds, especially if you use shortcut keys, record macros, and/or use the ROI manager tool, so you can fairly easy analyse tens to hundreds of recordings.