Synapse Web
About Anatomy Links Publications Search Tools

IGL Trace FAQ

Jump to Issue: General | Image Input | Calibration | Contour Tracing | 3D Output
General:
Will IGL Trace run on a Mac?

IGL Trace should be compatible with Virtual PC with Windows, a Mac program available from Apple Computer, Inc., that allows you to run PC programs.

Image Input:

How do I import a whole stack of images into IGL Trace?

If the stack of images is all in one file then you can try to use the Generic File Import from the Study Menu.

If you have a set of files to import, try using the Convert program from the tools page. This allows you to convert the images into a file format recognizable to Trace while simultaneously renaming the files according to IGL Trace section numbering conventions.

Can color images be used in IGL Trace?

Trace will display 8-bit indexed color images. Formats that will display in color include BMP, TIFF, and GIF. Be warned that Convert will try to generate 8-bit grayscale output, so if you need to use Convert to rename your 8-bit color images make sure you use the Only Rename Files option.

Why can't I import my images?

The image I/O library used with IGL Trace cannot handle all varieties of images files. If the file cannot be imported at all, try converting to a different file format, such as Windows Bitmap, before importing.

An entire series can be easily converted to a compatible grayscale format by using the Convert program from the tools page.

Is there a maximum number of sections that can be made into a study?

The maximum section number is 999999.

When I open my study or try to Page Up, I get an error: "Not a Valid Section", but I can GoTo my sections just fine from the Section menu. What's wrong?

Your section numbers are outside the range allowed by IGL Trace. Sections must be from 0 to 999999, and there must not be any other characters in the number, such as letters (99a) or leading zeros.

To fix your section numbering painlessly, use Convert from the tools page.

Is there a maximum file size for a study?

There is no maximum file size for section images or contour files as far as I know. But for a PC with a modest amount of memory (<64Mb) and slow disk drive, working with large section images (>>1Mb) will be very slow.

Calibration:
During the calibration routine, there is a prompt for dpi but I'm unable to find an explanation of this in the help routines. What do I need to do there?

The dpi fields are just comment fields. Feel free to type any information you want there. The actual dpi is not used in the calibration calculation but it's nice to keep a record of it so that additional material can be scanned at the same resolution later.

Contour Tracing:
Will IGL Trace work with a PenTablet monitor, such as the WACOM or the new Sony LX960?

We have used IGL Trace with the WACOM pen and tablet, but not with a tablet monitor. In principle, IGL Trace should work with any device that directly substitutes for the mouse.

Why aren't my traces being saved?

Check the file system. If the disk is full, or if the files and/or directories have Read Only status, then IGL Trace will not modify or write contour files. Unfortunately, IGL Trace issues no error messages when a write fails.

Is there a maximum number of different contours that I can make and follow for each study? The name table appears to have a maximum of about 20.

There is no maximum number of contours within a section, as far as I know. The contours for each section are stored in a separate file, so total numbers of contours in the study should not limited either.

The contour name table is just a convenient place to store frequently used contour name/color combinations and then access them by keystrokes. It is limited by the number of keys on the keyboard.

There can be as many different contour names as you want in a study, but only contours which share the same contour name are considered to be the same object for purposes of 3D reconstruction and measurement.

Also, contour names are limited to 8 characters.

Contours drawn by hand contain many minute loops that cause the 3D Surface output to be erroneously covered with large exterior triangles. These loops are very hard to see and it is very labor intensive to find and correct them. Is there an easier way to get rid of them?

Loops generate errors in the surface because the algorithm is no longer able to correctly distinguish "interior" from "exterior" contours. Recall that clockwise contours mark the "exterior" or outer boundary of the object while counterclockwise contours mark "interior" holes. A contour with a loop will not be interpreted correctly.

Loops are generated when the hand jiggles during tracing. This is particularly problematic when tracing on high resolution images. It can help to adjust mouse sensitivity or to zoom in when tracing so that jiggles don't result in large loops. Also, try tracing with high zoom, so that the section pixels are at least as large as the screen pixels. This minimizes the amount of deviation introduced by inadvertent jiggles.

Small loops can usually be eliminated by setting the "Max Error" to a large value when generating a 3D Surface. Setting this to 5 pixels, for example, makes it unlikely that loops smaller than 5 pixels in diameter will be retained in output. You can also try using "Shrink Each", "Expand Each" and "Force Clockwise" in the 3D Surface dialog. These parameters were added to provide easier ways for overcoming problems with traces such as loops and orientations.

When I zoom in to a region in my image and then use Fit Window to Image, contours are misplaced. How can I get around this problem?

After Fit Window to Image is used on a subimage, contours will be misplaced. Resize the window slightly by dragging on the lower righthand corner and the contours should return to their proper positions.

How are wildfire contours generated?

When a section is first opened it is assigned an arbitrary threshold of 160, but the user can select any value from 0-255 by using the slider in the Contrast Dialog. In the default mode, wildfires will expand into all four-connected pixels that are greater than the threshold. By using the Configure-Wildfires submenu the user can change the mode of expansion. "More" changes the algorithm to expand into all eight-connected neighbors. "Less" changes the algorithm to expand into only those four-connected pixels that also have a neighboring above-threshold pixel.

3D Output:
Will I be able to easily change colors in a 3D rendering environement?

Yes, you can import your reconstruction into some other rendering program and apply whatever coloring or texturing you want. Alternatively, colors can be changed in the VRML file as described in the next answer.

When I generate output from the 3D Menu, all I get is a text file displayed in Notepad. How do I see the 3D image?

You'll need a VRML viewer to see the 3D output. The text displayed in Notepad is the actual VRML that is generated. A VRML viewer (several are listed on the tools page) can convert this text file into a 3D image.

You can also configure IGL Trace to output the VRML directly to your VRML viewer instead of to Notepad. (See the directions in IGL Trace's Help.)

Displaying the Trace output in CosmoPlayer the image is too small and off screen. In fact it is even hard to find the contour set. How can I more easily see my output?

Trace generates 3D coordinates (in microns) based on the image size and the section thickness and section number. Thus, an object can be very far away from the origin (to which CosmoPlayer is nominally oriented) when the section thickness is 100 microns and the object starts at section 20. Likewise, a distance of 200 microns within a section would put the object far off to one side. One solution is to enter units in millimeters even though the program says microns. This would probably bring everything back into view.

Alternatively, you could use another program to view your reconstructions. CosmoPlayer is probably not the optimal viewing tool, especially when viewing 3D lines. Try instead GLView, listed under VRML viewers on the tools page, or 3Space Assistant which will automatically center the object in the field of view.

The *.wrl files created by IGL Trace are quite large. When importing them into Crossroads and "converting" them again to VRML 1.0 format they tend to become much smaller. Is this just a format question, or am I loosing needed information in the conversion?

Crossroads eliminates the normal vectors (which are used by some 3D viewers to improve smoothing). If you see no difference in the two cases then you don't need them. You can prevent them by selecting Normals->None in the Surface dialog.

I use the Trace Regions command to create many contours on each section. 3D Studio MAX is slow to load one reconstruction and very slow to merge two. How can I make reconstructions which are quicker to load and manipulate in 3dsmax?

You can limit the size of the surface reconstruction (i.e. the number of planar faces) by increasing the Max. Error in pixels allowed during surfacing. Using Trace Regions creates very pixelated contours that probably need to be simplified during surfacing. Try setting the Max. Error in the Surface dialog to 5 or 10 pixels. Also, turn off the generation of surface Normals to reduce the size of the file.

You may also be able to reduce the number of extraneous contours by adjusting the threshold (in the Contrast dialog) prior to Trace Regions.

When I export one particular object as a 3D surface, I keep getting a fatal error: "acos: DOMAIN error", which terminates the program. Is this something I can fix?

This error may be coming from the computation of normal vectors, try turning off normal generation by selecting None of the Normals box in the Surface dialog.

If you can isolate the contours which are causing the problem, you may be able redraw them to eliminate the problem.

I have reconstructed two objects and I would like to display them both but with different colors. How should I proceed? Is it enough to define a contour using another color for the pencil ? Or is there any special procedure ?

IGL Trace only outputs colored 3D objects when more than one object is being created, e.g. from the 3D->Lines menu.

When a VRML object is created from the 3D->Surfaces menu it is a single object and has the default VRML color which is gray.

You can color this object by inserting a material descriptor in the VRML (.wrl) file. You can also color objects and combine them into a single VRML scene by using the program MergeWRL avaliable on the tools page.

I have a whole bunch of objects all named the same. When I export as a 3D surface, the surfacing algorithm connects all these objects together. Do I need to go back and rename all these individually (and then export them all individually) to get them to split up? Or is there something else I can do?

Yes, the program will link together contours with the same name because it thinks they're the same object. You might be able to avoid renaming everything by renaming on only those that are close together and performing surfacing over narrow ranges of sections.

How can I make my 3D output be an open tube rather than a closed tube?

To open the tube, simply uncheck the "Show upper horizontal faces" and the "Show lower horizontal faces" items on the 3D->Surface dialog box.

When I view my 3D reconstruction the orientation of the reconstruction is rotated 180 degrees. Why?

The orientation in the viewer program depends on how it interprets the positive z-direction. In Trace, section 1 is assumed zero (in z-direction) and each subsequent section is interpreted as an increment in the positive z-direction.

There is also a discrepancy between the handed-ness of the coordinate systems that will be corrected in future versions.

Is there an easy way to print my 3D images?

To get a high-quality result, you'll want import the VRML file into a 3D renderer program. If you just want to save what you see in your browser/VRML, I suggest a screen capture utility such as obtained by using the Print Screen button on your keyboard. The Print Screen button copies the screen as a bitmap to the clipboard which you can then import into another program for editing and printing.

What program should I use to do 3D rendering?

We have used 3D Studio MAX for all of our rendering. It runs under Windows NT and appears to be fully compatible with the output generated by IGL Trace. It imports VRML 1.0 objects directly. 3D Studio seems to support all the features we need and can certainly handle large 3D scenes.

More limited rendering is available from a VRML viewer such as Cosmoplayer or the free GLView on the tools page.

When a .wrl object created by Slabs is imported into trueSpace, each slab is treated as a separate object, which makes it very hard to change the material, color, etc. Is there some way for me to modify the VRML code to make sure that all these slabs are grouped together?

Try passing the slabs .wrl through MergeWRL. This will add an outer separator that trueSpace should respect. Corssroads will also do a similar thing.

The images on your website look smoother than the output I'm getting, were they postprocessed with another program or is there some other trick to it?

The 3D images on our website were generated with 3D Studio MAX. Smooth shading in this program is very good. We also use texturing to help hide the transitions between surface patches. Take a look at your reconstructions using something like 3Space Assistant (see VRML Viewers on the tools page) to see if they appear smoother.

If the problem is with IGL Trace 3D->Surface output, try increasing the Max Error parameter, and/or shrinking and expanding the contours. There may be large effects based on how the contours are generated or drawn.

How can I get a three dimensional scale bar for my reconstructions?

To generate a scale bar, just draw one into two sections and surface it. This is easy provided images are calibrated.

I would like to create a movie in an .AVI file from the animations obtained with CosmoPlayer. How can I do this? What kind of software should I use?

We use 3D Studio MAX to generate our animations. It designed to produce nice .AVI files.

I find that the 3D surfaces are too huge for my rendering program. What can I do?

If the problem with surfacing is the number of surface patches produced, make sure that you set the Max. Error value high in the Surfacing dialog.

Thanks to everyone who contributed questions!

Last Updated: 4/25/02