Can we predict the future of respiratory failure prediction?

Critical Care volume 29, Article number: 253 (2025)

Background

Mortality in patients with acute respiratory failure remains high. Predicting progression of acute respiratory failure may be critical to improving patient outcomes. Machine learning, a subset of artificial intelligence is a rapidly expanding area, which is being integrated into several areas of clinical medicine. This manuscript will address the knowledge gap in predicting the onset and progression of respiratory failure, provide a review of existing prognostic strategies, and provide a clinical perspective on the implementation and future integration of machine learning into clinical care.

Main body

Existing strategies for predicting respiratory failure, such as prediction scores and biomarkers, offer both strengths and limitations. While these tools provide some prognostic value, machine learning presents a promising, data-driven approach to prognostication in the intensive care unit. Machine learning has already shown success in various areas of clinical medicine, although relatively few algorithms target respiratory failure prediction specifically. As machine learning grows in the context of respiratory failure, outcomes such as the need for invasive mechanical ventilation and escalation of respiratory support (e.g. non-invasive ventilation) have been identified as key targets. However, the development and implementation of machine learning models in clinical care involves complex challenges. Future success will depend on rigorous model validation, clinician collaboration, thoughtful trial design, and the application of implementation science to ensure integration into clinical care.

Conclusion

Machine learning holds promise for optimizing treatment strategies and potentially improving outcomes in respiratory failure. However, further research and development are necessary to fully realize its potential in clinical practice.

 

Flow starvation during square-flow assisted ventilation detected by supervised deep learning techniques

 

by Candelaria de Haro, Verónica Santos-Pulpón, Irene Telías, Alba Xifra-Porxas, Carles Subirà, Montserrat Batlle, Rafael Fernández, Gastón Murias, Guillermo M. Albaiceta, Sol Fernández-Gonzalo, Marta Godoy-González, Gemma Gomà, Sara Nogales, Oriol Roca, Tai Pham, Josefina López-Aguilar…

 

Critical Care volume 28, Article number: 75 (2024) Published: 14 March 2024

 

Background

Flow starvation is a type of patient-ventilator asynchrony that occurs when gas delivery does not fully meet the patients’ ventilatory demand due to an insufficient airflow and/or a high inspiratory effort, and it is usually identified by visual inspection of airway pressure waveform. Clinical diagnosis is cumbersome and prone to underdiagnosis, being an opportunity for artificial intelligence. Our objective is to develop a supervised artificial intelligence algorithm for identifying airway pressure deformation during square-flow assisted ventilation and patient-triggered breaths.

Methods

Multicenter, observational study. Adult critically ill patients under mechanical ventilation > 24 h on square-flow assisted ventilation were included. As the reference, 5 intensive care experts classified airway pressure deformation severity. Convolutional neural network and recurrent neural network models were trained and evaluated using accuracy, precision, recall and F1 score. In a subgroup of patients with esophageal pressure measurement (ΔPes), we analyzed the association between the intensity of the inspiratory effort and the airway pressure deformation.

Results

6428 breaths from 28 patients were analyzed, 42% were classified as having normal-mild, 23% moderate, and 34% severe airway pressure deformation. The accuracy of recurrent neural network algorithm and convolutional neural network were 87.9% [87.6–88.3], and 86.8% [86.6–87.4], respectively. Double triggering appeared in 8.8% of breaths, always in the presence of severe airway pressure deformation. The subgroup analysis demonstrated that 74.4% of breaths classified as severe airway pressure deformation had a ΔPes > 10 cmH2O and 37.2% a ΔPes > 15 cmH2O.

Conclusions

Recurrent neural network model appears excellent to identify airway pressure deformation due to flow starvation. It could be used as a real-time, 24-h bedside monitoring tool to minimize unrecognized periods of inappropriate patient-ventilator interaction.

Critical Care Bulletin - May 2023

 

The future of intensive care: the study of the microcirculation will help to guide our therapies

 

by J. Duranteau, D. De Backer, K. Donadello, N. I. Shapiro, S. D. Hutchings, A. Rovas, M. Legrand, A. Harrois and C. Ince 

 

Critical Care volume 27, Article number: 190 Published: 16 May 2023

 

Abstract

The goal of hemodynamic resuscitation is to optimize the microcirculation of organs to meet their oxygen and metabolic needs. Clinicians are currently blind to what is happening in the microcirculation of organs, which prevents them from achieving an additional degree of individualization of the hemodynamic resuscitation at tissue level. Indeed, clinicians never know whether optimization of the microcirculation and tissue oxygenation is actually achieved after macrovascular hemodynamic optimization. The challenge for the future is to have noninvasive, easy-to-use equipment that allows reliable assessment and immediate quantitative analysis of the microcirculation at the bedside. There are different methods for assessing the microcirculation at the bedside; all have strengths and challenges. The use of automated analysis and the future possibility of introducing artificial intelligence into analysis software could eliminate observer bias and provide guidance on microvascular-targeted treatment options. In addition, to gain caregiver confidence and support for the need to monitor the microcirculation, it is necessary to demonstrate that incorporating microcirculation analysis into the reasoning guiding hemodynamic resuscitation prevents organ dysfunction and improves the outcome of critically ill patients.

The future of AI in critical care is augmented, not artificial, intelligence

 

The future of AI in critical care is augmented, not artificial, intelligence

by Vincent X. Liu 

 

Critical Care volume 24, Article number: 673 (2020) Published: 02 December 2020

 

The field of AI—artificial intelligence—has seen tremendous success over the past decade. Today, AI touches billions of lives each day through voice and text processing, computer vision, prediction algorithms, video games, and much more. Naturally, there has also been enormous interest in applying AI to health care and, in particular, to data-rich environments like the intensive care unit. Early examples of AI in healthcare and critical care have already shown great promise [1], but also raise concerns that can be mitigated with preparation and foresight [2,3,4].

Recently, I put my own life into the hands of AI: it nearly killed me and, later, it also saved me. This harrowing experience was a potent reminder for me, an AI practitioner, that we must work to ensure this technology’s formidable capabilities are used to produce ‘augmented’, rather than just ‘artificial’, intelligence. Augmented intelligence places clinicians and ultimately patients, rather than algorithms, at its center. Where we successfully bridge the interface of clinician and machine intelligence, we have vast potential to make healthcare more effective, efficient, and sustainable. This will also ensure that health AI is safe, reliable, and equitable for all patients.