Article Preview
TopIntroduction
The damage suffered by different organs is often caused or accompanied by impaired peripheral nerve function. Therefore, in different fields of medicine, the researchers are interested in studying the changes of the peripheral nerve due to pathologies or injuries. This is why experimental models of peripheral nerve research based on morphometric methods have become very attractive for neuromorphologists (Badia et al., 2010; Helvacioglu et al., 2016; Moiseev et al., 2019). Additionally, any set of nerve fibers consists of subpopulations that differ in terms of morphometric parameters, functions (sensory or motor nerve fibers) and conduction velocity. Nevertheless, there is no general clustering approach based on a set of morphological parameters able to identify groups of myelinated nerve fibers (MNF) according to the commonly accepted physiological classification (Prodanov et al., 2010). The problem of morphometric and functional based classification of MNF in neuromorphology involves several aspects. It is related to unsupervised classification in terms of obtaining the maximally different groups (Nath et al., 2014). On the other hand, MNF are of similar morphology, but form natural biological clusters depending on their physiological features. Additionally, the number of the morphometric parameters is reduced, and these parameters can be determined based on the cross-section of MNF. Furthermore, there is the impossibility of physiological methods be used for assessing nerve conduction velocity in morphological studies. Another significant limitation is the huge biological variability in the measurements, which is associated to peculiarities of the histological process of the tissue (Arbuthnott et al., 1980; Hopkins et. al., 2015). Due to these factors, the use of the traditional clustering approach is impracticable, which motivates to search for a clustering approach for MNF groups separation related to the commonly accepted physiological classification based on morphometric parameters used in the neuromorphological experimental research.
The novelty of the proposed algorithm is that it was developed according to the morphology and function of myelinated nerve fibers. Hence, data regarding the well accepted physiological classification was analyzed, compared with nerve fibers microanatomy research and then a general pattern was obtained. Afterwards, this pattern was implemented with statistical apparatus. It should be noticed that the proposed algorithm was not aimed to identify groups which maximally differentiate their characteristics as happens in traditional clustering analysis, but to determine the clusters that have high similarity as to physiological types of MNF.
In this article, an algorithm for MNF clustering is proposed. The new algorithm is based on model-based clustering using finite mixture models, performed in few phases by increasing the number of morphometric parameters in order to obtain a classification, which, with a high level of probability, represents the physiological features of different types of MNF of the rat sciatic nerve. The proposed clustering approach allows to estimate with high level of probability the functional changes of peripheral nerve fibers based on the estimation of the sets of their morphometric parameters in experimental neuromorphology.
The remaining of this article is organized as follow: in the next section, a review of works related to peripheral nerve with focus on morphometry of nerve fibers is presented; then, section “Materials and methods” contains information about the experimental setup and the description of the proposed algorithm; in “Results” section, data of descriptive statistic and clustering solution, obtained with the proposed algorithm and also with the traditional one-step clustering based approach, are presented. Finally, the obtained results and their discussion as to the probability of biological nature representation of the nerve fibers population are presented. The conclusions are drawn in the final section.