Prediction of respiratory complications by quantifying lung contusion volume using chest computed tomography in patients with chest trauma

doi:10.21203/rs.3.rs-2127692/v1

Download PDF

Article

Prediction of respiratory complications by quantifying lung contusion volume using chest computed tomography in patients with chest trauma

https://doi.org/10.21203/rs.3.rs-2127692/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 19 Apr, 2023

Read the published version in Scientific Reports →

You are reading this latest preprint version

Pulmonary contusion is an important risk factor for respiratory complications in trauma patients. Hence, we aimed to determine the relationship between pulmonary contusion volume and patient outcomes and the predictability of respiratory complications. We retrospectively included 73 patients with a pulmonary contusion on chest computed tomography (CT) from 800 patients with chest trauma admitted to our facility between January 2019 and January 2020. Chest injury severity was expressed as the ratio of the normal lung volume to total lung volume by quantifying pulmonary contusion volume on chest CT. The cut-off value was 80%. Among 73 patients with pulmonary contusion (77% males, mean age: 45.3 years), 28 patients had pneumonia, and 5 had acute respiratory distress syndrome. The number of patients in the severe risk group with <80% of normal lung volume was 38, among which 23 had pneumonia. In predicting pneumonia, the area under the receiver operating characteristic curves for the ratio of the normal lung volume to total lung volume was 0.85 (95% confidence interval: 0.76–0.95, p=0.008), and the optimal threshold was 70.4%. Quantifying pulmonary contusion volume using initial CT in patients with chest trauma allows for identifying patients at high risk of delayed respiratory complications.

Pulmonary contusion is an important risk factor for respiratory complications, including pneumonia and acute respiratory distress syndrome (ARDS), in trauma patients. In addition, pulmonary contusions, which account for approximately 25% of all deaths, are an independent risk factor for mortality [1]. A pulmonary contusion can occur due to any mechanisms such as crushing and penetrating injuries and blunt trauma. These mechanisms damage alveolar capillaries, causing blood and other body fluids to accumulate in the lung tissue, thereby interfering with gas exchange and leading to hypoxia [2]. It is also related to the activation of inflammatory response via innate immunity following trauma [3]. Patients are treated with adequate oxygen supplementation, and mechanical ventilation may be required in severe cases. Pulmonary contusion usually resolves spontaneously with supportive care; however, they may progress to respiratory complications during treatment in intensive care units (ICUs).

The scoring systems for evaluating injury severity in patients with pulmonary contusions include the chest abbreviated injury scale (AIS) and injury severity score (ISS). However, because these scoring systems are obtained through anatomical abnormalities, they have limitations in representing functional impairment or predicting prognosis. Therefore, it is necessary to introduce a method for measuring the injury severity that reflects pulmonary function better. Hypoxemia is induced by ventilation-perfusion mismatch and an increase in shunt in areas with pulmonary contusion; hence, the wider the pulmonary contusion area, the more severe the degree of hypoxemia [2]. The pulmonary contusion volume may be an important risk factor in the progression of respiratory complications. Chest radiography has limitations in confirming the pulmonary contusion volume. However, it is easier to visualize pulmonary contusions and measure their area through chest computed tomography (CT). Contrast-enhanced CT is a good tool to identify bleeding and ground-glass opacity (GGO) more clearly due to parenchymal damage in chest CT. Recently, these studies have been conducted using lung segmentation technology for chest CT. Ultimately, it is important to extract the total lung and contusion volumes using CT. Automated lung segmentation algorithms are limited in handling routine data because they are developed for limited datasets of specific diseases or cases. According to a study by Hofmanninger et al. (2020), models trained with routine data were more effective in lung segmentation than public datasets in generalizability to a wide spectrum of pathologies [4]. Including or excluding pathology areas, such as effusion or pneumothorax, when using these trained models is a matter of definition and application. To determine the Hounsfield units (HU) range for segmenting of the entire lung area and the area expected to have normal lung function in trauma patients, we used a three-dimensional (3D) slicer and hand labeling data as a reference.

Therefore, this study aimed to determine the relationship between pulmonary contusion volume caused by all mechanisms of trauma and patient outcomes and whether respiratory complications can be predicted.

Patient characteristics

Among 73 patients with pulmonary contusion identified by chest CT, 56 (77%) were males, and the mean age was 45.3 ± 15.9 years. The most common mechanisms of injury were motor vehicle crashes, followed by falls and crushing injuries, with 40 (54.8%), 26 (35.6%), and five (6.8%) cases, respectively. The mean ISS was 22.3 ± 8.9 and mean chest AIS was 3.2 ± 0.7. A total of 28 patients (38.4%) were diagnosed with pneumonia, 5 (6.8%) were diagnosed with ARDS, and 2 (2.7%) died during hospitalization. The patients’ characteristics are summarized in Table 1.

The number of patients with ≥80% of normal lung volume in the moderate-risk group was 35, while the number in the severe group was 38. The mean age was 39.3 ± 16.1 years in the moderate group and 50.9 ± 13.7 years in the severe group, showing a statistically significant difference with a p-value of ≤0.005. When injury severity was compared between the two groups, the severe group had a higher ISS; nonetheless, the results were not statistically significant.

In-hospital outcomes

When the outcomes between the two groups were compared, pneumonia incidence was significantly higher in the severe group (five in the moderate group and 23 in the severe group). ARDS and in-hospital mortality were confirmed in five and two cases, respectively, in the severe group, but the difference was not statistically significant (Table 2).

Variables with an area under the curve (AUC) exceeding 0.7 in the receiver operating characteristic (ROC) curve for pneumonia prediction were the Glasgow coma scale (GCS) score, ISS, lactate level, and percentage of pulmonary contusion volume (Figure 3). The AUC was 0.77, 0.7, 0.73, and 0.85, respectively, showing the highest percentage of pulmonary contusion volume (95% confidence interval: 0.76–0.95, p=0.008, sensitivity 0.704, specificity 0.911). Additionally, the optimal threshold of the normal lung volume percentage for predicting pneumonia was 70.4%. The other optimal thresholds for GCS score, ISS, and lactate levels were 14.5, 19.5, and 2.45, respectively.

In trauma patients, lung contusions are a risk factor for respiratory complications [1-3,5,6]. Therefore, confirming the volume of pulmonary contusion on chest CT at the initial stage of injury can help predict respiratory complications [5-7]. According to previous studies, pulmonary contusion can cause respiratory complications by cytokine secretion mechanism above the threshold, including ARDS and mechanical damage to the lungs. Therefore, it is important to identify pulmonary contusion volume using radiological tests.

This study analyzed two groups by dividing the participants into moderate and severe groups according to the ratio of pulmonary contusion volumes. Miller et al. reported that the risk of ARDS according to the ratio of the pulmonary contusion volume did not increase linearly, but increased rapidly when the contusion volume was 20% of the total volume [5]. Although the cause is unclear, it is presumed that there is a threshold for ARDS caused by an intrapulmonary shunt at the injury site [5]. Although we utilized a small sample size in this study, there was no statistically significant difference in injury severity, such as the number of patients requiring transfusion, ISS, chest AIS, and GCS scores between the two groups. However, the ages of the moderate and severe groups showed statistically significant differences. As the initial pulmonary function test could not be performed in both groups, lung function could not be accurately compared. However, on chest CT, there were no significant differences in lung lesions that could impair lung function, such as pulmonary emphysema or interstitial lung disease. Depending on age, the damage to the lung itself seems to be different when impacted by trauma. As age increases, the composition of the components of each layer of the airway changes, resulting in differences in the mechanical force or pressure applied to the airway [8-11]. In addition, when lung tissue is injured in older persons, the immune and inflammatory responses are dysregulated, which is highly likely to lead to pathological conditions [12,13].

The pulmonary contusion volume we measured confirmed the GGO pattern, which may include inflammatory changes and hemorrhage caused by mechanical injury or cytokines and opacity due to aspiration at the time of injury [5]. Therefore, our center initiated antibiotic treatment with ampicillin-sulbactam to prevent aspiration pneumonia when more than half of the entire lung or more than 2/3 of the unilateral lung had a contusion. Then antibiotics were adjusted according to clinical features and sputum culture.

Pulmonary contusion is a relatively common injury in trauma patients. In addition, because a significant number of trauma patients require ventilator care in the ICU, they can be a high-risk group for respiratory complications, such as pneumonia and ARDS. Therefore, selecting these risk factors early and administering appropriate treatment, such as antibiotics, pain management, and respiratory rehabilitation, to high-risk patients is necessary. According to these data, the GCS scores, lactate levels, and pulmonary contusion volumes exceeded the AUC of 0.7 in the ROC curve for predicting pneumonia. Among these, the AUC of pulmonary contusion volume was 0.85, with the highest predictive power for future pneumonia.

This study has some limitations. First, only CT scans and history taking were performed in cases of underlying lung history. It is difficult to perform an initial pulmonary function test in trauma patients due to ICU treatment, sedation, intubation, pain, and shock. Therefore, although underlying lung history is a risk factor for future respiratory complications, it is difficult to ascertain its effect. Although the underlying lung history is identified through history taking, most patients do not know their lung function and history. Second, when chest CT is performed at the initial stage of the patient's injury, it is not performed with full inspiration. It may be difficult to view the exact volume as it is performed in various breathing phases for each patient. However, as the lung is a dynamic organ that moves continuously, we deduce that there will be difficulties overcoming this limitation.

In conclusion, quantifying pulmonary contusion volume using initial CT in patients with chest trauma allows early identification of patients at high risk of delayed respiratory complications such as pneumonia. Moreover, it can improve the treatment effect in patients with chest trauma by preventing complications.

Informed consent was obtained from all participants or their legal guardians. All methods were carried out in accordance with relevant guidelines and regulations. This study was approved by the ethics committee of Pusan National University Hospital. IRB No. 2011-017-097, the date of IRB approval: November 12, 2020

Study design and patients

Among 800 patients with chest trauma admitted to our level I trauma center between January 2019 and January 2020, we retrospectively analyzed 73 patients diagnosed with pulmonary contusion based on chest CT. The inclusion criteria were as follows: (1) Patients injured because of trauma, (2) patients diagnosed with pulmonary contusion by chest CT, and (3) patients with chest AIS score of one or higher. In addition, the following cases were excluded: (1) absence of chest CT at the time of admission and (2) death within the first 48 hours of admission (Figure 1).

Data collection

Data collected included patient demographics, mechanism of injury, injury severity, mechanical ventilator application period, length of hospital stay, length of ICU stay, pneumonia, and ARDS, and was obtained through medical records. Injury severity included ISS, chest AIS, head and neck AIS, initial GCS score, transfusion requirement in the emergency room, and initial lactate level. In addition, the partial pressure of oxygen in arterial blood (PaO₂) and inspiratory fraction of oxygen (FiO₂) were obtained from daily arterial blood gas analysis and ventilator settings to calculate the PaO2/FiO2 ratio.

Lung area segmentation algorithm on chest CT

Two methods of lung segmentation that complement each other were used. First, the total and normal lung volumes were obtained using a 3D slicer program. The 3D slicer is a free, open-source software for medical image computing. After loading the CT Digital Imaging and Communications in Medicine images of the patients into the 3D slicer, we identified the HU range for the most appropriate lung segmentation. After adjusting the HU range, we compensated by hand labeling if a part with pulmonary contusion was missed (Figure 2). With reference to this method, we found the HU range, -950 to -450, including the total and normal lung volumes other than pulmonary contusion in the lung segmentation model, the U-net (R231) algorithm introduced in Hofmanninger’s study [4]. This algorithm utilized the U-net (R231CovidWeb) model, which additionally learns the lungs of coronavirus disease patients. We segmented the total, normal lung, and pulmonary contusion volume by contrasting these two methods. When the lungs were first segmented using this model, the normal lung volume was smaller than expected. This was because the pneumothorax was excluded from normal lung volume. Traumatic pneumothorax is often caused by laceration of the visceral pleura caused by rib fracture, the alveolar rupture caused by increased alveolar pressure due to chest compression, and visceral or mediastinal pleura rupture, rather than due to underlying lung disease [14]. The collapsed lung was expected to expand after thoracostomy; hence, the pneumothorax part was included as the normal lung volume, not the pulmonary contusion.

Three-dimensional (3D) filtering algorithm for pneumothorax detection

As previously described, pneumothorax can be considered to be a normal lung. This is because the HU of air in pneumothorax has a different HU range from alveoli; hence, additional search algorithms are needed to detect pneumothorax. Consequently, the search method was supplemented to recognize only the pneumothorax, excluding the alveoli part using the denoising technique through 3D filtering.

Calculation of the ratio of the normal lung volume to the total lung volume

The ratio of the normal lung volume to the total lung volume was calculated by including the pneumothorax volume in the normal lung volume, using the results obtained from the search method. According to the results of previous studies, ARDS incidence sharply increases when the pulmonary contusion volume exceeds 20% of the total lung volume. Hence, when the normal lung volume was ≥80% of the total lung volume, it was defined as the moderate group, and when it was < 80%, it was defined as the severe group (Figure 1) [1, 5-7].

Primary outcome and definition of ARDS

Outcomes were obtained, including pneumonia, ARDS, in-hospital mortality, hospital stay, ICU stay, and ventilator application period. ARDS was defined as the absence of evidence of congestive heart failure with a PaO2/FiO2 ratio of ≤200 and bilateral diffuse infiltration on chest radiography [15].

Statistical analysis

When appropriate, summary statistics are reported as median and interquartile range or mean with standard deviation. Categorical variables are expressed as numbers and percentages. The chi-square test and Fisher’s exact test were performed to compare the frequencies of categorical variables between the groups. The Mann–Whitney U-test and Wilcoxon rank-sum test were performed to compare the mean values of continuous variables. We used the ROC curve and the AUC to evaluate predictive factors for pneumonia. Statistical significance was set at a p-value of <0.05. SPSS (version 22.0; IBM Corp., Armonk, NY, USA) was used to analyze the data.

Data availability

The datasets used and analyzed during the current study available from the corresponding author on reasonable request.

Acknowledgments

This study was supported by Biomedical Research Institute Grant (20200306), Pusan National University Hospital

Author contributions

N.H. Lee: Investigation, Writing – Original, Writing – Rewiew & Editing, *S.H. Kim: Conceptualization, Formal analysis, Project administration, S. Seo: Methodology, Formal analysis, B.J. Kim: Software, Data Curation, C.S. Lee: Software, Formal analysis, G.H. Kim: Validation, S.J. Park: Investigation, S.H. Kim: Resources, D.Y. Ryu: Resources, H.H. Kim: Data Curation, S.B. Lee: Visualization, C.I. Park: Resources, J.H. Kim: Supervision

Competing Interests

The authors declare no competing interests.

Becher, R.D. et al. An innovative approach to predict the development of adult respiratory distress syndrome in patients with blunt trauma. J. Trauma Acute Care Surg. 73, 1229–1235 (2002).
Ganie, F. A. et al. Lung contusion: A clinico-pathological entity with unpredictable clinical course. Bull. Emerg. Trauma. 1, 7–16 (2013).
Daurat, A. et al. Thoracic trauma severity score on admission allows to determine the risk of delayed ARDS in trauma patients with pulmonary contusion. Injury. 47, 147–153 (2016).
Hofmanninger, J. et al. Automatic lung segmentation in routine imaging is primarily a data diversity problem, not a methodology problem. Eur. Rad. Exp. 4, 1–3 (2020).
Miller, P.R. et al. ARDS after pulmonary contusion: Accurate measurement of contusion volume identifies high-risk patients. J. Trauma. 51, 223–228; discussion 229 (2001).
Mahmood, I. et al. Clinical significance and prognostic implications of quantifying pulmonary contusion volume in patients with blunt chest trauma. Med. Sci. Monit. 23, 3641–3648 (2017).
Wang, S., Ruan, Z., Zhang, J. & Jin, W. The value of pulmonary contusion volume measurement with three-dimensional computed tomography in predicting acute respiratory distress syndrome development. Ann. Thorac. Surg. 92, 1977–1983 (2011).
Kim, J., Heise, R. L., Reynolds, A. M. & Pidaparti, R. M. Aging effects on airflow dynamics and lung function in human bronchioles. PLOS ONE. 12, e0183654 (2017).
Sokolov, I., Iyer, S. & Woodworth, C. D. Recovery of elasticity of aged human epithelial cells in vitro. Nanomedicine. 2, 31–36 (2006).
Suki, B. & Bartolák-Suki, E. Biomechanics of the aging lung parenchyma in. Engineering Materials and Processes. 95–133 (Springer, 2015).
Kim, J., Heise, R. L., Reynolds, A. M. & Pidaparti, R. M. Quantification of age-related lung tissue mechanics under mechanical ventilation. Med. Sci. (Basel). 5, 21 (2017).
Setzer, F. et al. Susceptibility to ventilator induced lung injury is increased in senescent rats. Crit. Care. 17, R99 (2013).
Paxson, J. A. et al. Age-dependent decline in mouse lung regeneration with loss of lung fibroblast clonogenicity and increased myofibroblastic differentiation. PLOS ONE. 6, e23232 (2011).
Sharma, A. & Jindal, P. Principles of diagnosis and management of traumatic pneumothorax. J. Emerg. Trauma Shock. 1, 34–41 (2008).
Gayzik, F. S., Hoth, J. J., Daly, M., Meredith, J. W. & Stitzel, J. D. A finite element-based injury metric for pulmonary contusion: Investigation of candidate metrics through correlation with computed tomography. Stapp. Car Crash J. 51, 189–209 (2007).

Table 1. Patient characteristics

Value (n = 73)

Demographics

Age (mean ± SD)

Male (%)

45.3 ± 15.9

56 (76.7%)

Mechanism of injury (%)

Motor vehicle crash

Fall down

Crush injury

Stab wound

Assault

40 (54.8%)

26 (35.6%)

5 (6.8%)

1 (1.4%)

Injury severity (mean ± SD)

ISS

Chest AIS

Head and neck AIS

GCS at ER

Patients requiring blood transfusion

22.3 ± 8.9

3.2 ± 0.7

1.0 ± 1.5

13.4 ± 3.1

Physiologic parameters

Lactate

SBP

3.8 ± 3.0

112.9 ± 31.6

Pulmonary status

Pneumonia

ARDS

Ventilator days

28 (38.4%)

5 (6.8%)

7.4 ± 14.8

Patient outcomes

ICU stay

Hospital stay

Mortality

9.4 ± 16.2

31.9 ± 56.4

2 (2.7%)

Abbreviations: ARDS, acute respiratory disease syndrome; AIS, abbreviated injury scale; ER, emergency room; GCS, Glasgow coma scale; ICU, intensive care unit; ISS, injury severity score; SBP, systolic blood pressure; SD, standard deviation

Table 2. Comparison between moderate and severe groups

(Normal lung volume)/(Total lung volume)

≥80%

(Moderate group)

<80%

(Severe group)

p-value

Demographics

Age (mean ± SD)

Male

39.3 ± 16.1

50.9 ± 13.7

<0.005

Injury severity (mean ± SD)

ISS

Chest AIS

GCS at ER

Patients requiring blood transfusion

21.2 ± 7.7

2.8 ± 0.6

14.1 ± 2.5

23.4 ± 9.8

3.4 ± 0.7

12.8 ± 3.5

0.29

0.07

Outcome

Pneumonia

ARDS

Mortality

Hospital stay

ICU stay

Ventilator days

35.2 ± 80.3

6.5 ± 19.1

4.6 ± 17.3

28.9 ± 20.2

12.0 ± 12.9

9.9 ± 11.9

<0.005

0.090

0.54

0.66

0.17

0.14

No competing interests reported.

Download PDF

Journal Publication

published 19 Apr, 2023

Read the published version in Scientific Reports →

Editorial decision: Major revision
07 Feb, 2023
Reviews received at journal
05 Feb, 2023
Reviewers agreed at journal
26 Jan, 2023
Reviews received at journal
08 Jan, 2023
Reviewers agreed at journal
27 Dec, 2022
Reviewers invited by journal
27 Dec, 2022
Editor assigned by journal
27 Dec, 2022
Editor invited by journal
05 Oct, 2022
Submission checks completed at journal
05 Oct, 2022
First submitted to journal
03 Oct, 2022

You are reading this latest preprint version

Prediction of respiratory complications by quantifying lung contusion volume using chest computed tomography in patients with chest trauma

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Results

Discussion

Methods

Declarations

References

Tables

Additional Declarations

Status:

Journal Publication

Version 1