Stochastic modeling of the human middle ear dynamics under pathological conditions.

Background: Although the dynamics of the middle ear (ME) have been modeled since the mid-twentieth century, only recently stochastic approaches started to be applied. In this study, a stochastic model of the ME was utilized to predict the ME dynamics under both healthy and pathological conditions.

Methods: The deterministic ME model is based on a lumped-parameter representation, while the stochastic model was developed using a probabilistic non-parametric approach that randomizes the deterministic model. Subsequently, the ME model was modified to represent the ME under pathological conditions. Furthermore, the simulated data was used to develop a classifier model of the ME condition based on a machine learning algorithm.

Results: The ME model under healthy conditions exhibited good agreement with statistical experimental results. The ranges of probabilities from models under pathological conditions were qualitatively compared to individual experimental data, revealing similarities. Moreover, the classifier model presented promising results.

Discussion: The results aimed to elucidate how the ME dynamics, under different conditions, can overlap across various frequency ranges. Despite the promising results, improvements in the stochastic and classifier models are necessary. Nevertheless, this study serves as a starting point that can yield valuable tools for researchers and clinicians.