Estimation de la pression partielle artérielle du dioxyde de carbone par des paramètres noninvasifs

Abstract

Patients suffering from a respiratory disorder are in general mechanically ventilated. The purpose of this ventilation is to provide the oxygen to the organs to insure their proper function, and to eliminate the carbon dioxide produced by these organs. In order to efficiently choose the ventilator settings, the clinicians rely on vital signs and arterial partial pressure of the patients. Acquiring the arterial partial pressure directly from the arteries is very painful and can destroy the blood vessels and the nerves in the area of the puncture. Therefore, we try to find a reliable and less dangerous way to estimate these partial pressures. The arterial oxygen partial pressure (PaO2) is easily estimated by the transcutaneous oxygen saturation (SpO2) measured with a spirometer, while the arterial blood CO2 pressure (PaCO2) is more challenging to estimate. The main purpose of this work is to find a noninvasive reliable technique to rapidly estimate the PaCO2 with a better performance than the already in-use techniques such as the PaCO2 estimation with end-tidal CO2 (PetCO2) and the minute ventilation volume (Vmin) evolution. To do so, a high-resolution database in Sainte-Justine University Hospital was used. This database has all patients’ data in the Pediatric Intensive Care Unit (PICU). The MultiLayer Perceptron (MLP), a type of the neural networks, was used to do the estimations. The model is calibrated by a previously performed arterial blood gas test. The data were separated into four groups : [0h, 2h], [2h, 6h], [6h, 12h] and [12h, 24h] depending on the time gap between the calibrating test and the time of estimation. To test the efficiency of the developed approach, the percentage of estimations with an error less than 5 mmHg was used since a difference of less than 5 mmHg is considered medically insignificant. The obtained results show that the approach developed in this work is better than the traditional techniques that are used in clinical practice nowadays.

Date	26 Mar 2020
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Rita Noumeir (Supervisor) & Philippe Jouvet (Co-supervisor)