From Utah Center for Neuroimage Analysis

Liz, Pavel, Melissa and I (Chi-Bin) had a productive meeting on 11/21/05 (~45 minutes). Here's a summary, including some followup from after the meeting:

Contents 1 Datasets 2 Preprocessing issues 3 Annotation tools 4 Other

Datasets

• There are two datasets at present, one of a wildtype fish (9d1) and another of a mutant fish (8d3). Each set comprises three large stacks of slices, which are generally consecutive (although there is a 15-slice gap in one case). We had given these to Tolga on DVD and I believe they are on a SCI server. Resolution is 26.7 x 26.7 x 40 nm.

• We are gearing up to generate some new datasets in the next few months, aiming to increase lateral resolution to 12.5 nm.

Preprocessing issues

• There are some obvious defects in individual slices--either large pieces of plastic obscuring large areas of the images, or occasionally small shreds of plastic. The large pieces seem to be correlated with overall shift and deformation of the image. In the short term we can just ignore the defective slices; in the longer term we may want to add some masking tools and deformation corrections so that we can use the data that is there.

• There is a ripple visible if you play through the slices; ripples are ~40 pixels in period, running slightly diagonally left-to-right. Pavel and Liz are going to check in frequency space and see if this could perhaps be removed with a notch filter.

From discussion Talk:Nov 21 (Chi-Bin, Liz, Pavel, Melissa)#Ripples: Tolga and Ross have looked at this ripple and don't believe it is worth taking out, as it is unlikely to affect tracking. At present this seems to be a dead issue.

• Cropping--Melissa has a cropped ROI (500 x 900 pixels) that is the minimal box enclosing the optic tract. She will send this to Liz. She will also send an even smaller ROI (~200 x 200 pixels) that encloses a useful subset of the dataset and can be used for test runs.

Follow-up: As of 12/2/05, Melissa tried to post a cropped dataset on the Chien lab FTP server, but Liz was unable to download this file. Instead, Melissa will mark the crop box on a single slice and email this to Liz.

• We were impressed by the UINTA preprocessing that Tolga showed at our meeting last week, but wanted to know whether this is practical to use. What is the processing time for a 500 x 900 image? Would this be the same for every slice, or could succeeding slices be done more quickly using the parameters found for the original slice?

Follow-up: The answer to the last question seems to be no. See Talk:Nov 21 (Chi-Bin, Liz, Pavel, Melissa)#UINTA.

Annotation tools

• Melissa has been able to use ImageViewer to track axons throughout the entire stack (~2-3 hours), and to export this data as a text file. We would like a couple of small improvements to the ImageViewer annotation tool
  • Larger cursor--e.g. 3 x 3 pixels instead of 1 pixel
  • Multiple colors--e.g. up to 10, to track multiple axons
  • Reimportation of annotation--this would let you do one axon one day, another axon another day, and compare the results.

Follow-up: 12/2/05: Paul modified ITK ImageViewer to make ZebraViewer, which does all of the things that we asked for.

Other

We discussed 3D visualization tools to look at these axon tracks, but decided for now to stick with quick and dirty solutions (e.g. Excel, Graphing Calculator, or GNUplot). See gnuplot for instructions on how to download Gnuplot for Mac OS X and use it to display axon tracks.