False positive detection

The above detection procedure can produce two kinds of errors: missed detection when a fiber is not found and false positive when an image feature is wrongly considered to be a fiber. False positive is considered a worse problem because it is most likely to introduce bias in size distribution statistics, as most false positives are of a small size. Missed detection is only detrimental if its rate is size dependent, which does not appear to be the case.
In order to minimize the number of false positives, we apply three additional tests on the detected fibers. First, each individual fiber should have a d/D ratio close to 0.6. Secondly, fibers should be mostly surrounded by endoneurium, not by other fibers. This removes common false positives located in the space between 3 neighboring axons. Finally, isolated features are discarded.