Session: 02-08-01: Advances in Human Modelling

Paper Number: 144335

144335 - Development and Comparative Analysis of Statistical Shape Models of the Nasosinusal Anatomy 

The human nose exhibits a large variation in shape among individuals. All these variants alter the airflow through the nasal cavity and can impact how we smell odours. To acquire a better understanding of physiological and pathological functioning, it is important to study the effects of these modifications. SSM, or Statistical Shape Modeling, is a widely used methodology for considering morphological differences within a population. The creation of an SSM opens many opportunities for further analysis of the geometrical differences that exist within the population: it can be useful to study the influence of morphology on different flow characteristics of the nasal flow, e.g., nasal resistance, heat transfer, and flow humidity, but also for clinical studies related to particle deposition, drug delivery, and disease diagnosis.

In the literature, just a few studies analyze nasal anatomy in its entirety, including paranasal sinuses, whose segmentation is particularly challenging due to their complex and variable anatomy. For this reason, despite the multiple fundamental roles paranasal sinuses play, they are often excluded from the 3D model generation.

Starting from these considerations, this work aims to create a highly accurate SSM for the nasosinusal complex and conduct sensitivity analyses to evaluate the impact of considering just specific subsets of anatomies instead of the entire population.

A traditional, thresholding-based segmentation approach for the whole nasal cavity geometry has been chosen, relying on Mimics software (Leuven, Belgium). From a dataset of 58 CT scans, 40 (with ages from 19 to 84; 20 males and 20 females) were considered qualitative enough from a technical point of view to be included in the model creation. Indeed, the quality of CT scans is crucial in the process, and only images with sufficient resolution and Signal-to-Noise Ratio (SNR) are suitable. Each anatomy was divided into four masks, containing the cavity, the frontal, the lateral, and the ethmoidal sinuses, respectively. After segmentation, the post-process has been performed using 3-Matic software (Leuven, Belgium). A sequence of wrap, remesh, and smooth algorithms has been applied to get the final 3D models. For SSM generation, a proper module available in 3-Matic software was exploited. Two different SSMs were obtained in our analysis. A “partitioned” one, starting from differentiated inputs for cavities and three pairs of sinuses, and a “merged” one, where all anatomical parts were merged for each patient before the SSM. A comparison of meshes was then performed, evaluating the DICE coefficient and Hausdorff distance as metrics.

Both generated SSMs showed good trends of mean and standard deviation of the variability interval, but the partitioned one had better overall results. To deepen this and related issues, further work is now in progress.

As a preliminary output, this study showed the feasibility of statistical analysis of the nasosinusal complex, pursuing two different approaches. The resulting SSMs can be considered a representative model of a varied population of patients. However, more studies on a more significant number of samples could be conducted to confirm the trends found. An interesting development of the work is related to the possibility of running computational fluid dynamics (CFD) studies on these models, evaluating the impact in terms of the model sensitivity for CFD output parameters, like velocities and pressures. This could be highly relevant also for studying smell perception mechanisms.

 

 

Presenting Author: Marco Rossoni Politecnico di Milano

Presenting Author Biography: Assistant Professor at the Dept. of Mechanical Engineering, Politecnico di Milano, Italy

Authors:

Michele Bertolini Politecnico di Milano
Silvia Tonghini Politecnico di Milano
Marco Rossoni Politecnico di Milano
Marina Carulli Politecnico di Milano
Susanne Weise Technischen Universität Dresden
Christian Jan Baldus Technischen Universität Dresden
Giorgio Colombo Politecnico di Milano
Monica Bordegoni Politecnico di Milano