Tuning parameters of the video tracking window

Following successful specimen detection (Image analysis), fine-tune the video tracking algorithms in this window. Here, users adjust segmentation methods (e.g., Frame, Threshold, and Slope), define spatial constraints like the maximal growth factor, and apply post-processing filters to eliminate noise and refine detection accuracy (Figures 5–7). By iteratively testing tracking parameters and validating results through visual feedback, researchers ensure reproducible quantification of temporal changes such as cell migration, colony expansion, or morphological shifts.

Detailed description

**Figure 6:** Cellects video tracking window

Arena to analyze:

This arena number selects a specific arena in the current folder. The user can choose an arena, click Detection to load and analyze it, then Read results.

Note

Cellects automatically names the arena by their position (left to right, top to bottom).
For single arena setups, use 1.
Post processing triggers Detection, which in turn triggers Load One arena.
Videos can be saved (as .h5 files) for later analysis using the Advanced parameter Keep unalteredvideos.

Maximal growth factor:

This is the maximum allowable proportion of image area that may be covered by specimen movement between frames. Adjust accordingly:

Increase if specimen size is underestimated.
Decrease if specimen size is overestimated.

Note

Precisely, this defines an upper bound on relative coverage changes between sequential images.

Temporal smoothing:

Applies temporal smoothing to reduce noise and highlight long

-term trends by averaging pixel intensity changes. Use when analyzing slope -based segmentation results.

Note

This uses a moving window algorithm on pixel intensity curves over time.
Excessive iterations produce constant values, preventing accurate detection.

Segmentation method:

Cellects includes five video tracking options:

Frame option: Applies the image analysis algorithm frame by frame, without temporal dynamics.
Threshold option: Compares pixel intensity with the average intensity of the whole image at each time step.
Slope option: Compares pixel intensity slopes with an automatically defined threshold.
T and S option: logical AND of threshold and slope options.
T or S option: logical OR of threshold and slope options.

Note

Selecting the Compute all options before dunning Detection allows method comparison. Onceanalysis completes. Once the analysis completed, select one option and click Read.
Computing only one option is faster and requires less memory.
When Heterogeneous background or Grid segmentation has been selected in the image analysiswindow, only the Frame option remains available.

Load one arena:

Clicking this button loads the arena associated with Arena to analyze. The center of the window displays the first frame of that arena's video. Click Read to review the full video.

Detection:

Detection applies a (or all) segmentation methods to one arena. Once finished, click Read to view the detection result. If correct, answer Done to proceed with tuning parameters for post processing.

Read:

Clicking Read starts the video display corresponding to the current state of the analysis.

**Figure 7:** Cellects video tracking window during detection visualization

Specimen activity:

The behavior of the specimen(s) changes how Cellects post processes the data (after video segmentation):

move → Specimen(s) can move from one place to another in the arena but are not expected to grow. The status of an area (specimen or background) does not depend on where the specimen(s) were previously.
grow → Specimen(s) only grow, they cannot leave an area. The previous position of the specimen(s) is used to detect its current position.
move and grow → Specimen(s) are expected to move and grow. This feature use the previous position of the specimen(s) to evaluate growth and the pixel intensity history to evaluate when they are left.

Fading detection:

Fading detection monitors when specimens leave previously occupied areas, useful for moving organisms rather than static growth. Uncheck this option if not needed. Set a value between minus one and one to control sensitivity:

Near minus one: Minimal false removal of specimen traces.
Near one: High risk of over -removal from all areas.

Post processing:

Post processing applies detection algorithms with additional enhancements:

Binary operations: opening, closing, logical ops.
Fading detection* tracking: when specimen(s) may leave areas (optional).
Correct errors around initial shape: when the contour of the initial position of the specimen is hard to detect (optional).
Connect distant shapes: when the specimen's heterogeneity create wrong disconnections in the video detection (optional).
Prevent fast growth near periphery: when arena's border (typically petri dishes) may be wrongly detected as specimen (optional).

Note

Once Post processing works, the user can click “Done” to Step 2: Tune fading and advancedparameters to improve Post processing, and then Run All arenas.

Save one result:

Complete the current video analysis by clicking this button for single arena processing. Saving includes:

Calculating all selected descriptors (.csv) per frame.
Generating validation videos for detection verification.
Storing configuration parameters for reproducibility.

Note

This action will overwrite results and validation data for the current arena.

**Figure 8:** Cellects video tracking window, running all arenas

Run All:

Apply validated parameters to all arenas by clicking Run All. This action:

Generates full -resolution video outputs (storage -intensive)
Processes videos sequentially with real time visualization
Calculates selected descriptors for each frame
Produces validation content at multiple intervals
Preserves current configuration settings

Save all choices:

Clicking Save all choices writes/updates configuration files to preserve analysis parameters for future replication.