Abstract
The current report presents the development and application of a novel methodological approach for computer-based methods of processing and analysis of proliferative tissues labeled by ABC-peroxidase method using 3, 3′-diaminobenzidine (DAB) as chromogen. This semiautomatic method is proposed to replace the classical manual approach, widely accepted as gold standard. Our method is based on a visual analysis of the microscopy image features from which a computational model is built to generate synthetic images which are used to evaluate and validate the methods of image processing and analysis. The evaluation allows knowing whether the computational methods applied are affected by the change of the image characteristics. Validation allows determining the method’s reliability and analyzing the concordance between the proposed method and a gold standard one. Additional strongness of this new approach is that it may be a framework adaptable to other studies made on any kind of microscopy.
Highlights
Pure visual image analysis of immunocytochemical reactions is very prone to subjective bias, since it is a comparative process that can be influenced by the presence of background information, or other image artifacts
We show the outcome obtained using this methodology to replace a visual classic procedure of image analysis by a semiautomatic one, which is applied to the identification of DAB labeled nuclei for the cellular proliferation study in olfactory epithelium of Rhinella arenarum [10]
Olfactory neurons are the second nuclei layer, located approximately in the middle of the epithelium, some of which may be ciliated or present microvilli [22]. These are bipolar cells that have a unique thin dendrite, which reaches the outer surface of the epithelium, where their ends form a mosaic with external terminals supporting cells
Summary
Pure visual image analysis of immunocytochemical reactions is very prone to subjective bias, since it is a comparative process that can be influenced by the presence of background information, or other image artifacts. Quantification avoids the anecdotal analysis by means of a general model of the data representation [1, 2] This means that it is possible to diminish both the bias and error, based on a purely visual classification obtaining, from a set of objects already classified by one or more expert users, statistics parameter estimators. These parameters could represent some of the features of the objects to be classified. Using these estimators and their variation ranges, it is possible to apply the visual method and following, to perform the classification according to the statistic information previously analyzed.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
