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

Predicting mortality and risk factors of sepsis related ARDS using machine learning models

Sepsis related acute respiratory distress syndrome (ARDS) is a common and serious disease in clinic. Accurate prediction of in-hospital mortality of patients is crucial to optimize treatment and improve prognosis under the new global definition of ARDS. Our study aimed to use machine learning models to develop models that can effectively predict the in-hospital mortality of patients with sepsis related ARDS, calculate the mortality, and to identify related risk factors under the new global definition of ARDS. Based on MIMIC database, our study included 3470 first-time admission records of patients with sepsis related ARDS. After excluding 4 patients under the age of 18, 75 patients with less than 24 h stay in ICU, and 5 cases with missing indicators > 30%, finally 3386 cases were retained. The variance inflation factor (VIF) analysis was used to test the collinearity of the explanatory variables. The data were divided into the training set and the test set according to the ratio of 7:3. Six models, extreme gradient boosting (XGBoost), light gradient boosting (LightGBM), random forest (RF), classification and regression tree (CART), naive bayes (NB) and logistic regression (LR), were designed for training and testing. In the training set, XGBoost (AUROC = 0.951, 95% CI 0.942-0.961), LR (AUROC = 0.835, 95% CI 0.817-0.854), RF (AUROC = 1.0, 95% CI 1.0-1.0), LightGBM (AUROC = 1.0, 95% CI 1.0-1.0), CART (AUROC = 0.831, 95% CI 0.811-0.852), NB (AUROC = 0.793, 95% CI 0.772-0.814). In the test set, XGBoost (AUROC = 0.833, 95% CI 0.804-0.861), LR (AUROC = 0.82695% CI 0.796-0.856), RF (AUROC = 0.846, 95% CI 0.818-0.874), LightGBM (AUROC = 0.827, 95% CI 0.798-0.856), CART (AUROC = 0.753, 95% CI 0.718-0.787), NB (AUROC = 0.799, 95% CI 0.768-0.831). The RF model has the best performance on the test set. Further analyze the feature importance ranking and partial dependence plots of random forest model. Acute physiology and chronic health evaluation III (APACHE III), bicarbonate, anion gap and non-invasive blood pressure systolic were identified as the four most important risk characteristics. In this study, a variety of machine learning models have been successfully constructed to predict the in-hospital mortality of patients with sepsis related ARDS, among which the RF model performs well. Key risk factors identified include APACHE III, bicarbonate, anion gap and non-invasive blood pressure systolic. The identification of these factors helps clinicians to assess patients' conditions more accurately and develop personalized treatment plans, thereby improving the survival rate and prognosis quality of patients under the new global definition of ARDS.

Clinical Performance of Spo2/Fio2 and Pao2/Fio2 Ratio in Mechanically Ventilated Acute Respiratory Distress Syndrome Patients: A Retrospective Study

OBJECTIVES

The present study aims to evaluate the severity classification of acute respiratory distress syndrome (ARDS) in mechanically ventilated patients according to peripheral oxygen saturation by pulse oximetry (Spo2)/Fio2 ratio compared with Pao2/Fio2 ratio and the relationship between Spo2/Fio2 ratio and venous admixture.

DESIGN

Retrospective observational study.

SETTING

Medical-surgical ICU.

PATIENTS

A cohort of 258 mechanically ventilated patients with ARDS already enrolled in previous studies.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

Gas exchange, Spo2, and respiratory mechanics were measured on ICU admission and during the positive end-expiratory pressure (PEEP) trial. Radiological data from CTs were used to compute lung recruitability and to assess different lung compartments. A nonlinear association was found between Spo2/Fio2 and Pao2/Fio2. Considering the possible confounding factors of the pulse oximeter on the relationship between Spo2/Fio2 and Pao2/Fio2 ratio, arterial pH, and Paco2 had no effect. Spo2/Fio2 and Pao2/Fio2 ratio demonstrated a moderate agreement in classifying ARDS severity (intraclass correlation coefficient = 0.63). Between the correspondent classes according to Spo2/Fio2 vs. Pao2/Fio2 ratio-derived severity classifications, there was no difference in terms of respiratory mechanics, gas exchange, lung radiological characteristics and mortality in ICU, and within two levels of PEEP. A Spo2/Fio2 ratio less than 235 was able to detect 89% of patients with a venous admixture greater than 20%, similarly to a Pao2/Fio2 ratio less than 200.

CONCLUSIONS

Spo2/Fio2 ratio can detect oxygenation impairment and classify ARDS severity similarly to Pao2/Fio2 ratio in a more rapid and handy way, even during a PEEP trial. However, our results may not be applicable to different patient populations; in fact, the pulse oximeter is merely a monitoring device and the information should be personalized by the physician on the patient's characteristics and conditions.

Machine Learning Tools for Acute Respiratory Distress Syndrome Detection and Prediction

Machine learning (ML) tools for acute respiratory distress syndrome (ARDS) detection and prediction are increasingly used. Therefore, understanding risks and benefits of such algorithms is relevant at the bedside. ARDS is a complex and severe lung condition that can be challenging to define precisely due to its multifactorial nature. It often arises as a response to various underlying medical conditions, such as pneumonia, sepsis, or trauma, leading to widespread inflammation in the lungs. ML has shown promising potential in supporting the recognition of ARDS in ICU patients. By analyzing a variety of clinical data, including vital signs, laboratory results, and imaging findings, ML models can identify patterns and risk factors associated with the development of ARDS. This detection and prediction could be crucial for timely interventions, diagnosis and treatment. In summary, leveraging ML for the early prediction and detection of ARDS in ICU patients holds great potential to enhance patient care, improve outcomes, and contribute to the evolving landscape of precision medicine in critical care settings. This article is a concise definitive review on artificial intelligence and ML tools for the prediction and detection of ARDS in critically ill patients.

Quantitative Computed Tomography and Response to Pronation in COVID-19 ARDS

BACKGROUND

The use of prone position (PP) has been widespread during the COVID-19 pandemic. Whereas it has demonstrated benefits, including improved oxygenation and lung aeration, the factors influencing the response in terms of gas exchange to PP remain unclear. In particular, the association between baseline quantitative computed tomography (CT) scan results and gas exchange response to PP in invasively ventilated subjects with COVID-19 ARDS is unknown. The present study aimed to compare baseline quantitative CT results between subjects responding to PP in terms of oxygenation or CO2 clearance and those who did not.

METHODS

This was a single-center, retrospective observational study including critically ill, invasively ventilated subjects with COVID-19-related ARDS admitted to the ICUs of Niguarda Hospital between March 2020-November 2021. Blood gas samples were collected before and after PP. Subjects in whom the PaO2 /FIO2 increase was ≥ 20 mm Hg after PP were defined as oxygen responders. CO2 responders were defined when the ventilatory ratio (VR) decreased during PP. Automated quantitative CT analyses were performed to obtain tissue mass and density of the lungs.

RESULTS

One hundred twenty-five subjects were enrolled, of which 116 (93%) were O2 responders and 51 (41%) CO2 responders. No difference in quantitative CT characteristics and oxygen were observed between responders and non-responders (tissue mass 1,532 ± 396 g vs 1,654 ± 304 g, P = .28; density -544 ± 109 HU vs -562 ± 58 HU P = .42). Similar findings were observed when dividing the population according to CO2 response (tissue mass 1,551 ± 412 g vs 1,534 ± 377 g, P = .89; density -545 ± 123 HU vs -546 ± 94 HU, P = .99).

CONCLUSIONS

Most subjects with COVID-19-related ARDS improved their oxygenation at the first pronation cycle. The study suggests that baseline quantitative CT scan data were not associated with the response to PP in oxygenation or CO2 in mechanically ventilated subjects with COVID-19-related ARDS.

Monitoring lung recruitment

PURPOSE OF REVIEW

This review explores lung recruitment monitoring, covering techniques, challenges, and future perspectives.

RECENT FINDINGS

Various methodologies, including respiratory system mechanics evaluation, arterial bold gases (ABGs) analysis, lung imaging, and esophageal pressure (Pes) measurement are employed to assess lung recruitment. In support to ABGs analysis, the assessment of respiratory mechanics with hysteresis and recruitment-to-inflation ratio has the potential to evaluate lung recruitment and enhance mechanical ventilation setting. Lung imaging tools, such as computed tomography scanning, lung ultrasound, and electrical impedance tomography (EIT) confirm their utility in following lung recruitment with the advantage of radiation-free and repeatable application at the bedside for sonography and EIT. Pes enables the assessment of dorsal lung tendency to collapse through end-expiratory transpulmonary pressure. Despite their value, these methodologies may require an elevated expertise in their application and data interpretation. However, the information obtained by these methods may be conveyed to build machine learning and artificial intelligence algorithms aimed at improving the clinical decision-making process.

SUMMARY

Monitoring lung recruitment is a crucial component of managing patients with severe lung conditions, within the framework of a personalized ventilatory strategy. Although challenges persist, emerging technologies offer promise for a personalized approach to care in the future.

Lung Imaging and Artificial Intelligence in ARDS

Artificial intelligence (AI) can make intelligent decisions in a manner akin to that of the human mind. AI has the potential to improve clinical workflow, diagnosis, and prognosis, especially in radiology. Acute respiratory distress syndrome (ARDS) is a very diverse illness that is characterized by interstitial opacities, mostly in the dependent areas, decreased lung aeration with alveolar collapse, and inflammatory lung edema resulting in elevated lung weight. As a result, lung imaging is a crucial tool for evaluating the mechanical and morphological traits of ARDS patients. Compared to traditional chest radiography, sensitivity and specificity of lung computed tomography (CT) and ultrasound are higher. The state of the art in the application of AI is summarized in this narrative review which focuses on CT and ultrasound techniques in patients with ARDS. A total of eighteen items were retrieved. The primary goals of using AI for lung imaging were to evaluate the risk of developing ARDS, the measurement of alveolar recruitment, potential alternative diagnoses, and outcome. While the physician must still be present to guarantee a high standard of examination, AI could help the clinical team provide the best care possible.