User:TheLarry/Notebook/Larrys Notebook/2009/11/12

{| width="800"
 * style="background-color: #EEE"|[[Image:owwnotebook_icon.png|128px]] Tracking
 * style="background-color: #F2F2F2" align="center"|  |Main project page
 * style="background-color: #F2F2F2" align="center"|  |Main project page


 * colspan="2"|
 * colspan="2"|

Generate Data
I want to find out how well my tracking software works. To do this i need to generate my own movie and then try the tracking software on it. So i need a way to match the .pngs I am looking at. To match the background noise is easy. I take a sample of the background make a histogram of the pixel intensity. Fit that to a Gaussian and then create Gaussian noise with the same standard deviation and mean.

To create the microtubule is a bit more complicated. After talking with Koch the best way to do this is to take a cylinder (maybe a rectangle on a 2-d axis not sure yet) and place on fluorescence dots on it. Then i'll have to smear those dots. To do this I THINK I need to convolute the pixel the dot is actually in with a Gaussian function. Labview has a convolution .vi so that isn't much of a problem. I am still not sure if this is what i do but i am pretty sure this is how it works. I think the Gaussian function is the point spread function. After I have what 1 dot will look like i can randomly place dots on the microtubule. I am a bit fuzzy on how i'll do this. This will definitely be concentration dependent. I am just not sure right now how to randomly choose where they go. But I can worry about that when i get up to it. The first thing i gotta do is smear this dot.

To make sure i have the Gaussian right I some how made my way over to 3d surfaces. And it is pretty fun.

But now that i know my Gaussian formula is right (i make mistakes when i plug in complicated formulas into the formula node so i needed to double check) I can convolute this with my point.

I don't know what values i'll use for the sigmas (both x and y) but hopefully inspiration will come.

I think i am being stupid. The Gaussian may be what i am looking for. That is what a point looks like so it might be correct to model each dot as the Gaussian. I am not sure and have some reading to do. I just need to find something to read. For right now I'll go with thinking the Gaussian is what i want. Next what do i do about the standard deviation for the Gaussian and what is the maximum intensity of the dot.

I talked with Andy and he agreed that the Gaussian is what i want. It is what a point looks like in the microscope. So now i need to know what the spread is. How many pixels does this fluorescence spread into? Also i need to know how bright it gets. The spread will have to be changed into pixels and the max intensity will have to be changed into bits. I am not sure how to do the bit part but i'll figure something out when i get there.

Andy said that the fluorescence is Rhoadmine 6-G and we got it from Cytoskeleton. Also that the microtubules are 17% labeled. I am still not sure how to randomly place these down but i'll figure it out when i think about it closely.

I am starting to think that this spread is something people don't print. I don't know why unless it changes from microscope to microscope. I don't know. Is it something we'll have to get experimentally? I am not sure. Something i just fit? But how? I am not sure yet.

Apparently I can't open .pdfs any more. Hooray! Fuckett I am outta here before I get frustrated. Why can't i look at pdfs? Fuck you that's why, computer! Fuck you


 * }