Track workflow: Detection
Procedure
- Click a few worms to define a size range. The worms are displayed with an overlay model.
- The detection parameters are set by example based on the worms you clicked. See Detection settings below to learn about modifying the parameters.
- Click the button to obtain a preview of the detected worms in the window.
- If you're not satisfied with the results, click more worms in the window to modify the parameters.
- If you're satisfied, start tracking.
- Select the icon in the toolbar.
- Click the unwanted worm to select it.
- Press the Delete key.
Sometimes a worm fails detection because its shape does not conform to the detection parameters or it is overlapping itself.
Rather than reset the parameters, trace the worm manually:
- Click the icon
- Click along the worm's centerline from head to tail to draw the worm.
- Right-click at the tail to end the drawing. An green overlay model is placed over the worm.
If you right-click away from the worm, even slightly, the model will be incorrect. See an example of an incorrect model here:
Detection settings
Detect worms: Use AFTER clicking worms to define a size range. Detects worms within the field of view according to the parameters set under Detection Parameters. Worms that don't conform to these parameters are not detected.
Delete worms: Deletes models of detected worms.
Detection parameters
If checked, worms touching the edge of the image are included in the detection if they are valid.
Minimum/maximum area of valid worms.
Minimum/maximum length of valid worms.
Minimum/maximum width of valid worms.
Minimum/maximum width to length ratio of valid worms.
Minimal degree of fitness between newly detected worms and image (i.e., the degree to which the contour of the model aligns with the contour of the worm binary image).
Minimal degree of fitness between registered worms and image. The value is typically lower than detection fit because worm registration occurs in complex tracking scenarios (e.g., entanglement or occlusion) where the fit between model and image is usually lower.
Advanced
WormLab regularly performs an image-wide detection sweep to detect new worms entering the field of view. This setting determines how often this detection process is performed. A higher number of frames between detection increases processing speed, but delays the detection of worms entering the field of view.
To restrict tracking to worms selected manually, set the frequency to 0 (zero).
Our default worm model assumes that worms maintain a constant length. Selecting this parameter activates length elasticity with the assumption that worm length is variable.
Length Fitting is useful when the change in worm length must be captured accurately or when detection accuracy of the worm extremities is insufficient.
Systematically adjusts worm width along the length of the worm. Selecting this parameter will decrease processing speed.
The whole plate tracking mode uses a different model designed for low magnification and a large field of view.
For low resolution, worms typically only occupy a few pixels in the image, so accurate worm shape or interaction interpretation is not possible. In this mode, you sacrifice accuracy and tracking robustness for an increase in processing speed.
Notably absent from this mode is tracking robustness to worm interaction and self-overlap. For example, when two worms touch one another, WormLab drops their associated tracking , resuming it only when interaction ends.
Worm Shape
The number of iterations used in the worm-fitting process. With very big worms or high-magnification images, this number may need to be increased.
You may need to increase this value if you have a sequence of low frame rate images or worms that move unusually fast.
The number of points used to describe the worm median axis. Reducing the number of points increases processing speed. We recommend a minimum value of 31.
When enabled, this option maintains a uniform width profile along the worm body. This option is especially useful when dealing with low resolution images where the worm radius is less than 7 or 8 pixels. Note that this mode does not allow automatic detection of the head and the tail. If width fitting is enabled, WormLab attempts to fine-tune the width profile of the worm while maintaining the uniformity of the main body. If the fitting is successful head and tail can theoretically be distinguished from one another automatically.