Skip to Main content Skip to Navigation
Theses

Interfaces cerveau-ordinateur pour améliorer l'identification de la dyspnée chez les patients ventilés artificiellement

Abstract : Almost half of the patients under mechanical ventilation (MV) in intensive care units (ICU) experience respiratory discomfort (or dyspnea). Communication impairments between patients and caregivers in ICU make recognition and evaluation of such a dyspnea difficult. The aim of this work was to develop brain-computer interfaces (BCI) to help caregivers recognizing respiratory discomfort under MV in ICU. In the DYSVENT study, dyspneic patients under MV in ICU were included. An electroencephalogram (EEG) was recorded at baseline and after optimization of ventilator settings. The EEG was analyzed to look for a pre-inspiratory potential (PIP) which is a sign of a ventilation-related cortical activity (VRCA), which is usually absent during spontaneous breathing. In the DYSPEV study, 2 BCIs based on steady-state visual evoked potentials (SSVEP) were tested on healthy volunteers: a BCI that detects dyspnea (D-BCI) and a BCI that quantifies the dyspnea in the form of a virtual visual analogic scale (LAS). Subjects were studies under various respiratory conditions: spontaneous breathing (SB), inspiratory threshold load (ITL) and resistive load (IRL), CO2 inhalation (CO2) and back to SB (SBWO). Many frequency sets were tested: 12/15Hz, 15/20Hz, 20/30Hz for D-BCI and high frequencies (13, 17, 19, 23 and 29 Hz) and low frequencies (41, 43, 47, 53 and 59 Hz) for the LAS. In the DYSVENT study, included patients (n = 47) were dyspneic in 73 to 89% of cases according to the evaluation method used (EVA or the hetero-evaluation scores IC-RDOS or RDOS). The optimization of the ventilator settings significantly improved their discomfort. At baseline 38% of the patients had a PIP on EEG versus 19% after ventilator optimization (p < 10-4). Predictive positive value of PIP detection to identify respiratory discomfort was 1.00 (95CI [0.91-1.00]) using IC-RDOS and 0.96 (95CI [0,87-1,00]) using RDOS. Predictive negative value was 0.37 (95CI [0.25-0.50]) using IC-RDOS and 0.31 (95CI [0.19-0.44]) using RDOS. In the DYSPEV study, healthy volunteers experienced respiratory discomfort during IRL, ITL and CO2 in the D-BCI group (30 subjects) and during ITL and CO2 in the LAS group (20 subjects, ITL condition not tested in this group) with VAS significantly higher than during SB. For the D-BCI the best frequency set was 20-30Hz with AUC 0.89 (95CI [0.80-0.90]) and low frequencies for the LAS with AUC 0.84 (95CI [0.83-0.85]). In the DYSVENT study, VRCA detection was insufficient to highlight situations at risk of respiratory discomfort under MV. The DYSPEV study made the proof of concept in healthy volunteers of using a SSVEP-based BCI to detect and quantify dyspnea. A BCI gathering both techniques could be developed to help caregivers recognizing and taking care of respiratory discomfort under MV in ICU.
Complete list of metadatas

Cited literature [241 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-02932185
Contributor : Abes Star :  Contact
Submitted on : Monday, September 7, 2020 - 3:40:34 PM
Last modification on : Wednesday, September 9, 2020 - 2:59:54 PM

File

these_campion_sebastien_2019.p...
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-02932185, version 1

Citation

Sébastien Campion. Interfaces cerveau-ordinateur pour améliorer l'identification de la dyspnée chez les patients ventilés artificiellement. Neurosciences [q-bio.NC]. Sorbonne Université, 2019. Français. ⟨NNT : 2019SORUS190⟩. ⟨tel-02932185⟩

Share

Metrics

Record views

59

Files downloads

12