Assay Option: FAINT-EDGE-DETECTION
As the experiment progresses, the tubes become thinner until the point when the tubes visually disappear. During this transition the human reader can find faint tubes that are beyond the capability of automated analysis. In the image shown below, the red boxes indicate areas where the tubes are in various stages of disappearing from the image.
The automated system must also face a large variety of input images of varying resolution and quality. An aggressive linear edge detection can introduce un-wanted artifacts in noisy images. So we have chosen to be less aggressive on faint edge detection, by default, to provide the best possible result for the maximum number of researchers. However, we have provided an option to increase the aggressiveness of the faint edge detection.
In this example, a manual analysis indicates the presence of 21 closed loops.
In this example, the default analysis mode (less aggressive faint linear edge detection) will produce the following output, finding a total of 15 closed loops:
To increase the aggressiveness of the faint linear edge detection, add this text to the experiment-set description field:
With this option enabled, the following output is produced, finding a total of 22 closed loops.
While the total loops in this analysis is one more than the manual analysis, there are artifacts which missed one loop and created false loops.
- loop 10 from the manual analysis was lost (-1)
- number 20 from the manual analysis was split into two by the image margin (+1)
- number 12 from the manual was lost (-1),
- number 16 from the manual was split into three due to the margin and bubble (+2)
- number 10 from the manual was lost (-1)
for a net gain of 1.