Automatic Measurement of Inner Ear in Different Mammals¹

Abstract

Objective: The aim of this research is to develop an accurate and automatic measuring method based on the aid of centerline to construct three dimensional models of inner ear in different mammals and to assess the morphological variations. Methods: Three adult healthy mice, three adult guinea pigs, three adult mini pigs and one left temporal bone of human were included in this research. All 18 animal specimens and the human sample were scanned with the use of Micro-CT. After being segmented, three-dimensional models of the inner ear in different mammals were reconstructed using Mimics. A novel method with the use of centerline was established to estimate the properties of 3D models and to calculate the length, volume and angle parameters automatically. Results: Morphological models of inner ears in different mammals have been built, which describe detailed shape of cochlear, vestibule, semicircular canals and common crus. Mean value of lengths and volumes of the cochlear, lateral semicircular canal, superior semicircular canal and posterior semicircular canal, tended to increase with the body size of the mammals, showed the proximity to the human data in mini pig. The angles between the semicircular canal planes showed differences between mammals. The mean values of semicircular canals of mice and mini pigs closely resembled to human data in numerical assessment. Conclusion: The automatic measurement of the inner ear based on centerline builds an effective way to assess lengths, volumes and angles of three-dimensional structures. This study provides a theoretical basis for mechanical analysis of inner ear in different mammals and proves the similarity between mini pig and human.

Disturbance in the protein landscape of cochlear perilymph in an Alzheimer’s disease mouse model

“…This possible limitation is also related to the technical difficulty involved in our procedure and ethical concern for animal experiments. In a mouse cochlea, the volume of perilymph is extremely small, approximately 5 μL [60]. Therefore, collection of this fluid with minimal contamination of other fluids and cells requires a high level of skill.…”

Data Synthesis for Testing Black-Box Machine Learning Models

“…Their intent comes from the fact that a deeper neural layer captures complex features and therefore, its next layer can be considered as its summary. These testing criteria mainly focus on feed-forward neural networks, while DeepStellar [10] proposed the model-based testing criteria for recurrent neural networks.…”