Installation

• I have most recently compiled this for Amira 5.3.2 on MacOS X 10.6
• I normally distribute this as a zip archive containing two sets of files in nested folders that match the main Amira folders:

share/resources/*.rc
lib/arch-MacX-Optimize/*.dylib
* The files in the zip archive need to be copied into the matching location in the Amira tree.

• You may need to move the tracing toolbar to a new location (e.g. from top to left hand side) for it to respond to key presses etc.
• Add points by pressing the Cmd tool on the Mac
• Use the right most 'curvy' add branch tool for adding points (this uses Felix's algorithm). Middle one uses a memory hungry ZIB algorithm and left one just adds points in a straight line!
• Select the correct parameters and note issues with 16 bit images (see below)
• Hold down the Alt key to temporarily switch to the Hand pointer that allows the scene to be rotated. Press Cmd as well to move the focus of the viewer laterally through the scene.

• Sometimes skeletonize will have difficulty following a branch which is very feint or most problematically has a break. In order to get round this, use the leftmost add point tool to add a few points close together to bridge the difficult region, then trace on with the regular tool (rightmost), select the whole segment and do a few rounds of fitting.
• It can be easier to see what is going on if you increase the contrast of the Orthoslice
  • Choose e.g. Mapping Type Histogram to do this
• Alternatively you can try reducing the External parameters (as described below); in the limit this makes the rightmost add point button behave like the left most add point button.

This is a somewhat underdocumented art! If your parameters are wrong you will end up that adding points will not follow the intensity path. The dialog box allows you to adjust a total of 7 parameters.

The first line contains so called internal factors. These relate to the fitting of the basic path across the tracing. Starting from the right, the simplest parameter is step. As you trace, points are added every step microns; Felix calls these points snaxels. We typically use 0.5 microns. The two other parameters loc and rad are coupling parameters. Basically they indicate how flexible the path is (loc) and how much the radius of one snaxel influences the next (rad). They are effectively dependent on step size. As you increase the step size you also need to decrease the coupling strength.

These control the interaction with the image data more directly and relate to how image gradient information is used to set the centreline and radius of the cylinders that are use to model each segment of a neuron. The bigger they are, the more the tracing will move, but also the more it may jump as the image gets noisier.

They will need to change when you change voxel/object sizes. When you use 16 bit data they will need to be decreased substantially (Felix suggests 50 fold eg down to 0.001). Otherwise the tracing will do a random walk all over the place. In general the lower these parameters, the more stable the fitting routines (and the less one expects unstable oscillations), but the slower the fits will be and the less likely they are to fix a major problem.

Felix came by 2012-05-09

For Jonny's stacks 16 bit, 0.46×0.46×1 µm voxels suggested

0.1 0.1 0.8

0.001 0.001
0.001 0.001

For fuzzy neurons (but they have to be pretty fuzzy) increase locCenter and decrease locOffset.

Jonny has been using the following parameters for 8bit stacks

0.5 0.5 0.5

0.02 0.02
0.02 0.02

Recommends checking Carsten Duch for lysine fixable biotin concentrations (which he used in Manduca with better fills - more even, better penetration). Used Linaris Gmbh neurobiotin. But could only find them using this with sharps.