Patient-specific lung model development, clinical validation and prediction of ventilatory response
Anaesthesiology
Medical Physics, Biomedical Technology
Final Report Abstract
Mechanical ventilation is crucial for ARDS patients but carries the risk of ventilator-induced lung injury (VILI). Despite some advances in lung-protective strategies, progress has stagnated for two decades, primarily due to limited insights into lung (patho-)physiology and its individual manifestation in patients. This project sought to fill this gap by developing a patientspecific, spatially resolved computational lung model aiming to unravel an individual’s lung (patho-) physiology. This objective required the incorporation of two prevalent lung damage mechanisms, over-straining and cyclic recruitment and derecruitment (RD, opening/closing) processes, and to make it applicable to a patient's respiratory mechanics and heterogeneous lung pathology. A prospective observational study was conducted to validate the model, including 10 ARDS patients with present thoracic CT scans. Ventilator data, as well as electrical impedance tomography (EIT) data, were collected during predefined ventilation maneuvers and routine clinical practice. As part of the project, we developed two new approaches for patient-specific modeling of ARDS lungs including RD dynamics. The first approach is based on an established RD model, which was previously only applied to animal data and parameterized purely statistically. It was possible to further develop this model, apply it to patients and identify a mechanically based parameter set that enabled the reproduction of respiratory measurements. This was a major step, but it also showed that for a truly predictive model application, further research into the biophysical fluid properties and complex interactions in damaged lung tissue is required. Therefore, in addition to the originally planned work program, we have developed a completely new RD modeling approach as an alternative and also a parameterization algorithm tailored to it. Using image data and specific ventilation maneuvers of a patient, the algorithm calibrates the model and overcomes the problems identified. The application of the model to a patient showed a high degree of agreement with measured ventilation curves. In the final prospective clinical observational study, a total of 10 patients with ARDS and current CT imaging of the lungs were included in the period between June 2021 and September 2022. As part of the (currently ongoing) evaluation, the global ventilation variables simulated with patient-specific lung models of the 10 included ARDS patients are compared with clinically measured data and the lung models are validated locally using EIT measurements. Preliminary results from the patients studied are promising. For EIT processes, we also developed an efficient open-source software tool for generating virtual EITs from the simulation results of the respective lungs and for comparing clinical and simulated EIT images using innovative evaluation methods.
Publications
-
Patientenspezifische Entwicklung eines Lungenmodells, klinische Validierung und Vorhersage der Beatmungsantwort. German Clinical Trials Register: DRKS00017151.
Becher, T.
-
A next-generation discontinuous galerkin fluid dynamics solver with application to high-resolution lung airflow simulations. Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis, 1–15.
Kronbichler, Martin; Fehn, Niklas; Munch, Peter; Bergbauer, Maximilian; Wichmann, Karl-Robert; Geitner, Carolin; Allalen, Momme; Schulz, Martin & Wall, Wolfgang A.
-
An approach to study recruitment/derecruitment dynamics in a patient‐specific computational model of an injured human lung. International Journal for Numerical Methods in Biomedical Engineering, 39(9).
Geitner, Carolin M.; Becher, Tobias; Frerichs, Inéz; Weiler, Norbert; Bates, Jason H. T. & Wall, Wolfgang A.
-
Holo Shed: Generate EIT images from simulation output. Software tool.
Holo Shed
-
Pressure‐ and time‐dependent alveolar recruitment/derecruitment in a spatially resolved patient‐specific computational model for injured human lungs. International Journal for Numerical Methods in Biomedical Engineering, 40(2).
Geitner, Carolin M.; Köglmeier, Lea J.; Frerichs, Inéz; Langguth, Patrick; Lindner, Matthias; Schädler, Dirk; Weiler, Norbert; Becher, Tobias & Wall, Wolfgang A.
-
Solving Bayesian inverse problems with expensive likelihoods using constrained Gaussian processes and active learning. Inverse Problems, 40(9), 095008.
Dinkel, Maximilian; Geitner, Carolin M.; Robalo, Rei Gil; Nitzler, Jonas & Wall, Wolfgang A.
