Speaker
Description
The nasal cavity is one of the most important organs of the human body. Its various functionalities are essential for the well-being of the individual person. It is responsible for the sense of smell, supports degustation, and filters, tempers, and moistens the inhaled air to provide optimal conditions for the lung. Diseases of the nasal cavity
like chronic rhinosinusitis, septal deviation, or nasal polyps may lead to restrictions or complete loss of these functionalities [1, 2]. A decreased respiratory capability, the development of irritations and inflammations, and lung diseases can be the consequences.
The shape of the nasal cavity varies from person to person with stronger changes being present in pathological cases. A decent analysis on a per-patient basis is hence crucial to plan for a surgery with a successful outcome. Nowadays diagnostic methods rely on morphological analyses of the shape of the nasal cavity. They employ methods of medical imaging such as computed tomography (CT) or magnetic resonance imaging (MRI), and nasal endoscopy [3]. Such methods, however, do not cover the fluid mechanics of respiration, which are essential to understand the impact of a pathology on the quality of respiration, and to plan for a surgery. Only a meaningful and physics-based diagnosis can help to adequately understand the functional efficiency of the nasal cavity, to quantify the impact of different pathologies on respiration, and to support surgeons in decision making. This work presents a data-processing pipeline for planning surgical interventions in the respiratory tract, which contains the following steps:
(i) Super-resolution of CT data in case of an insufficient resolution
(ii) Automatic machine learning (ML)-based extraction of the upper airway and in- and outflow regions from CT data
(iii) ML-based flow-field initialization in Computational Fluid Dynamics (CFD) simulation for accelerating the computation
(iv) Efficient high-fidelity simulations of respiratory flows including an automated analyses of relevant flow parameters
(v) ML-assisted surgery prediction for modifying the shape of the nasal cavity with the goal of optimizing its functionalities (suggestion of a surgery plan)
In many of these steps, ML techniques are employed to guarantee an automated or efficient usage of the pipeline. In (i), a super-resolution network (SRN) is used to increase the resolution of a CT recording, in case the resolution is too low for reliable CFD simulations [4]. Convolutional neural networks (CNNs) are employed in (ii) to extract the airway from the CT data, and to detect inflow and outflow regions automatically [5]. In (iii) - (iv), highly-resolved CFD simulations are conducted to yield accurate results for an analysis from a fluidmechanical point of view. To accelerate the simulation, the CFD solver starts from an approximated flow field generated by a physics-aware CNN (PA-CNN) in (iii) that integrates the Navier-Stokes equations into its loss function, before the simulation continues in (iv). In the final step (v), a reinforcement learning (RL) algorithm is employed to modify the airway [6]. After each modification, the algorithm receives feedback in terms of fluid mechanical parameters from CFD simulations. The final structure of the airway then functions as the proposed surgery.
References
[1] M. Damm, G. Quante, M. Jungehuelsing, E. Stennert, Impact of Functional Endo-
scopic Sinus Surgery on Symptoms and Quality of Life in Chronic Rhinosinusitis, The
Laryngoscope 112 (2) (2002) 310–315. doi:10.1097/00005537-200202000-00020.
[2] I. Croy, T. Hummel, A. Pade, J. Pade, Quality of Life Following Nasal Surgery, The
Laryngoscope 120 (4) (2010) 826–31. doi:10.1002/lary.20824.
[3] G. Scadding, P. Hellings, I. Alobid, C. Bachert, W. Fokkens, R. G. Wijk, P. Gevaert,
J. Guilemany, L. Kalogjera, V. Lund, J. Mullol, G. Passalacqua, E. Toskala,
C. Drunen, Diagnostic tools in Rhinology EAACI position paper, Clinical and Trans-
lational Allergy 1 (1) (2011) 2. doi:10.1186/2045-7022-1-2.
[4] X.Liu, M. Rüttgers, A. Lintermann, Using super-resolution networks with ct images
for increasing the accuracy of respiratory flow simulations, In preparation (2023).
[5] M. Rüttgers, M. Waldmann, W. Schr ̈oder, A. Lintermann, A machine-learning-based
method for automatizing lattice-boltzmann simulations of respiratory flows, Applied
Intelligence 52 (8) (2022) 9080–9100.
[6] M. Rüttgers, M. Waldmann, K. Vogt, J. Ilgner, W. S. and. A. Lintermann, Auto-
mated surgery planning for an obstructed nose, In preparation (2023).
| Keywords | Machine learning, Reinforcement learning, Surgery planning |
|---|
