Machine learning predicts lung recruitment in acute respiratory distress syndrome using single lung CT scan

Pennati, Francesca; Aliverti, Andrea; Pozzi, Tommaso; Gattarello, Simone; Lombardo, Fabio; Coppola, Silvia; Chiumello, Davide

doi:10.1186/s13613-023-01154-5

Annals of Intensive Care

Table 2 Subsets of the most informative variables selected according to the frequency with which they were chosen after repeating the least absolute shrinkage and selection operator (LASSO) algorithm

From: Machine learning predicts lung recruitment in acute respiratory distress syndrome using single lung CT scan

Feature group name	Data
Outcome: Δ_45-5non-aerated tissue > 15%
M5	ARDS origin, tidal volume, plateau pressure, driving pressure
M5 + M15	ARDS origin, tidal volume, plateau pressure, driving pressure, Δ_15-5 mechanical power
M5 + M15 + RPM	ARDS origin, tidal volume, plateau pressure, driving pressure, Δ_15-5 mechanical power, lung elastance
G5	ARDS origin, PaO₂/FiO₂
G5 + G15	ARDS origin, PaO₂/FiO₂, Δ_15-5 PaO₂
CT5	Age, ARDS origin, well-aerated lung tissue, non-aerated lung tissue
CT5 + G5	Age, ARDS origin, well-aerated lung tissue, non-aerated lung tissue, PaO₂/FiO₂
CT5 + M5	Age, ARDS origin, well-aerated lung tissue, non-aerated lung tissue
CT5 + G5 + G15	Age, ARDS origin, well-aerated lung tissue, non-aerated lung tissue, Δ_15-5 PaO₂
CT5 + M5 + M15 + RPM	Age, ARDS origin, well-aerated lung tissue, non-aerated lung tissue, Δ_15-5 mechanical power
CT5 + G5 + G15 + M5 + M15 + RPM	Age, ARDS origin, well-aerated lung tissue, non-aerated lung tissue, Δ_15-5 PaO₂, Δ_15-5 mechanical power
Outcome: Δ_15-5PaO2 > 24 mmHg
M5	ARDS origin, BMI, mechanical power
M5 + M15	ARDS origin, BMI, mechanical power, Δ_15-5 driving pressure
M5 + M15 + RPM	ARDS origin, BMI, mechanical power, Δ_15-5 driving pressure, chest wall elastance
G5	ARDS origin, BMI, PaO₂
CT5	ARDS origin, total lung weight, poorly aerated lung tissue, well-aerated lung tissue
CT5 + G5	ARDS origin, total lung weight, poorly aerated lung tissue, well-aerated lung tissue, PaO₂
CT5 + M5	ARDS origin, total lung weight, poorly aerated lung tissue, well-aerated lung tissue, mechanical power
CT5 + M5 + M15 + RPM	ARDS origin, total lung weight, poorly aerated lung tissue, well-aerated lung tissue, Δ_15-5 driving pressure, chest wall elastance
CT5 + G5 + M5 + M15 + RPM	ARDS origin, total lung weight, poorly aerated lung tissue, well-aerated lung tissue, Δ_15-5 driving pressure, chest wall elastance, PaO₂
M5	ARDS origin, BMI, mechanical power
CT5 + G5 + G15 + M5 + M15 + RPM	Age, ARDS origin, well-aerated lung tissue, non-aerated lung tissue, Δ_15-5 PaO₂, Δ_15-5 mechanical power

Lung recruitability was defined both as the percent change in not aerated tissue between 5 cmH₂O and 45 cmH₂O (recruiters: Δ_45-5non-aerated tissue > 15%) and as the change in PaO₂ between 5 cmH₂O and 15 cmH₂O (recruiters: Δ_15-5PaO2 > 24 mmHg). Input parameters included lung mechanics at PEEP 5 cmH₂O (M5), lung mechanics at PEEP 15 cmH₂O (M15), respiratory partitioned mechanics (RPM), gas exchange measured at PEEP 5 cmH₂O (G5), gas exchange measured at PEEP 15 cmH₂O (G15), CT imaging acquired at PEEP 5 cmH₂O (CT5)

Back to article page