COVID-19 therapy optimization by AI-driven biomechanical simulations

Abstract

The COVID-19 disease causes pneumonia in many patients that in the most serious cases evolves into the Acute Distress Respiratory Syndrome (ARDS), requiring assisted ventilation and intensive care. In this context, identification of patients at high risk of developing ARDS is a key point for early clinical management, better clinical outcome and optimization in using the limited resources available in the intensive care units. We propose an AI-based prognostic system that makes predictions of oxygen exchange with arterial blood by using as input lung Computed Tomography (CT), the air flux in lungs obtained from biomechanical simulations and Arterial Blood Gas (ABG) analysis. We developed and investigated the feasibility of this system on a small clinical database of proven COVID-19 cases where the initial CT and various ABG reports were available for each patient. We studied the time evolution of the ABG parameters and found correlation with the morphological information extracted from CT scans and disease outcome. Promising results of a preliminary version of the prognostic algorithm are presented. The ability to predict the evolution of patients' respiratory efficiency would be of crucial importance for disease management.

Fig. 1

Lung CT slice of a patient from the “Santa Croce e Carle” database. The yellow box highlights a lesion with ground glass opacities

Fig. 2

Example of the model for a patient of the “Santa Croce e Carle” dataset. The volumetric rendering shows the lungs, the trachea and the internal airways. The presence of GGO is represented as a solid object. The extraction of airways and parenchyma and the identification of lesions is the starting basis for biomechanical model and the subsequent simulations

Fig. 3

Example of results of biomechanical simulations for a patient presenting 28% of Covid lesions. In the left panel the Covid lesions are taken into account in the biomechanical model where in the right panel the lesions are substituted by a completely healthy parenchyma. The contours of the velocity magnitude (units in m/s and in logarithmic scale), are shown on a reference frontal slice and at the maximum inspiration during the breathing cycle

Fig. 4

Example of results of biomechanical simulations for the same patient of Fig. 3 reporting volumetric flow rate in the trachea (upper panels) and the mass flow rate (lower panel) in the healthy region of the parenchyma acquired over two respiratory cycles, each having a duration of 4s. The second respiratory cycle is used to compute ventilatory statistics. The left and right panels are for the case with Covid and without-Covid lesions, respectively, as in Fig. 3. The presence of 28% of lesions reduces the total mass flow rate by 42%

Fig. 5

pO${}_2$/FiO${}_2$ as a function of time for a patient of the “San Croce e Carle” database. A linear fit is superimposed

Fig. 6

a First two components in the PCA analysis described in the text. Each point corresponds to a patient. In purple there are the patients that needed intensive care at some point and/or died while in light blue there are the less severe cases. b Correlation plot between the pO${}_2$/FiO${}_2$ ratio at the CT time versus the COVID-19 lesion fraction