Crypts detection in microscopic images using hierarchical structures

This paperpresentsan extended and improved version of an automatictechniquewhich robustlyidentifiestheepithelial nuclei (crypt)against interstitial nucleiinmicroscopic images taken from colontissues.The detection of thecrypt inner boundaryis performedusing the closing morphological hierarchy.The disadvantages of this approachrelated to the execution time andtheused memoryare highlighted andthe morphological pyramid is used insteaddue to itscomputational efficiency, thereducedamount of usedmemoryandthe increasedrobustness. An analysis ofthe two approaches isperformedconsidering thenumber of processed pixels, the used memory and the complexity.The outer border isdeterminedbytheepithelial nuclei overlapped bythe maximal isoline of the inner boundary.The percentage ofthe mis-segmented nucleiagainstepithelial nucleiper cryptis used toevaluatethe proposed methods.The limitationsare describedin orderto highlight the situations in which the current approachesdo not provide suitableresults.