A total of 366 patients (276 survivors and 90 non-survivors) were finally included in this study. Characteristics of included patients were presented in Table 1. Age and sex were equally distributed (P = 0.43, P = 0.97) in two groups. The APACHE Ⅱ score in non-survivors (30 [23–35]) was higher than that in survivors (21 [16–27], P < 0.001). The ICU length of stay and hospitalization showed no difference between the two groups (P = 0.29, P = 0.37). For primary causes of ARDS, most patients were due to intrapulmonary factors (88.0% in survivors and 90.0% in non-survivors). The number of patients without sepsis was slightly higher than those with sepsis but with no statistical significance (P = 0.18). The percentage of patients with moderate severity was highest in all severity subgroups (51.1% in survivors and 47.8% in non-survivors).
Table 1
Baseline Characteristics of the included patients
Characteristics | Survivors(n = 276) | Non-survivors(n = 90) | P value |
Age (y)* | 64 (50–75) | 66.5 (51–78) | 0.43 |
Sex, male, % | 208 (75.4%) | 68 (75.6%) | 0.97 |
Primary cause of ARDS | | | 0.61 |
Intra-pulmonary | 243 (88.0%) | 81 (90.0%) | |
Extra-pulmonary | 33 (12.0%) | 9 (10.0%) | |
ARDS Severity | | | 0.02 |
Mild | 65 (23.6%) | 12 (13.3%) | |
Moderate | 141 (51.1%) | 43 (47.8%) | |
Severe | 70 (25.3%) | 35 (38.9%) | |
Sepsis | | | 0.18 |
With | 117 (42.4%) | 31 (34.4%) | |
Without | 159 (57.6%) | 59 (65.6%) | |
APACHE Ⅱ Score* | 21 (16–27) | 30 (23–35) | < 0.001 |
Length of ICU stay* | 11 (5–21) | 12 (8–20) | 0.29 |
Length of hospitalization* | 17 (8–28) | 18 (9–32) | 0.37 |
Note. —Unless otherwise indicated, data are numbers of patients, with percentages in parentheses. *Data are medians, with interquartile ranges in parentheses. |
A total of 1218 features of each label was extracted. Seven features at the upper lung level were selected with most characteristics related to texture uniformity. At the middle lung level, twelve features were selected, most of which related to density distribution. As for the lower lung, only three features, including Interquartile Range, Mean, and the Median after wavelet transformation, were selected. Detailed information about selected radiomics features was listed in the Supplemental material Table S1. More explanation of the radiomics features can refer to the website (https://pyradiomics.readthedocs.io/en/latest/features.html).
The P/F ratio (P = 0.003), APACHE Ⅱ score (P < 0.001), Radiomics_Score_1 (P < 0.001), Radiomics_Score_2 (P < 0.001), Radiomics_Score_3 (P < 0.001), Tv/PBW (P = 0.03), RR (P = 0.03), MP (P = 0.01) and MP/PBW (P = 0.01) were associated with the 28-Day mortality in the univariate logistic regression analysis (Table 2). After the multivariate analysis, three independent risk factors including APACHE Ⅱ score (OR, 2.607, 95% CI: 1.896–3.584, P < 0.001), Radiomics_Score_2 (OR, 2.230, 95% CI: 1.387–3.583, P = 0.01), Radiomics_Score_3 (OR, 1.633, 95% CI: (1.143–2.333, P = 0.01) were selected at last.
Table 2
Factors associated with 28-Day mortality in ARDS patients with mechanical ventilation.
| Univariate analysis | Multivariate analysis |
Variables | Estimate (95% CI) | P-value | Estimate (95% CI) | P-value |
Age (y) | 0.999 (0.985–1.012) | 0.86 | NS | NS |
Sex (male) | 1.119 (0.635–1.973) | 0.70 | NS | NS |
Primary causes of ARDS | 1.029 (0.470–2.251) | 0.94 | NS | NS |
Severity of ARDS | | | | |
Moderate | 1.363 (0.668–2.783) | 0.39 | NS | NS |
Severe | 2.269 (1.077–4.780) | 0.03 | NS | NS |
With/without sepsis | 1.439 (0.856–2.419) | 0.16 | NS | NS |
P/F ratio | 0.609 (0.440–0.884) | 0.003 | NS | NS |
APACHE Ⅱ Score | 2.798 (2.058–3.805) | < 0.001 | 2.607 (1.896–3.584) | < 0.001 |
Radiomics_Score_1 | 1.762 (1.325–2.343) | < 0.001 | NS | NS |
Radiomics_Score_2 | 2.621 (1.678–4.093) | < 0.001 | 2.230 (1.387–3.583) | 0.01 |
Radiomics_Score_3 | 1.978(1.395–2.805) | < 0.001 | 1.633 (1.143–2.333) | 0.01 |
Tidal volume | 1.251 (0.985–1.590) | 0.06 | NS | NS |
Tidal volume/PBW | 1.299 (1.020–1.654) | 0.03 | NS | NS |
Respiratory Rate | 1.310 (1.032–1.664) | 0.03 | NS | NS |
Positive End-Expiratory Pressure | 1.177 (0.927–1.493) | 0.18 | NS | NS |
Peak Inspiratory Pressure | 1.080 (0.846–1.379) | 0.54 | NS | NS |
Driving_pressure | 0.982 (0.766–1.259) | 0.88 | NS | NS |
Mechanical power | 1.363 (1.083–1.715) | 0.01 | NS | NS |
Mechanical power /PBW | 1.397 (1.109–1.761) | 0.01 | NS | NS |
NS = Not significant; PBW = Predicted body weight; CI = Confidence interval. | |
A clinical model (APACHE Ⅱ score), a radiomics model (Radiomics_Score_2 + Radiomics_Score_3), and a clinical_radiomics model (APACHE Ⅱ score + Radiomics_Score_2 + Radiomics_Score_3) to predict the 28-Day mortality were constructed. The AUC of these three models in validation set were 0.758 (95% CI: 0.710 to 0.802), 0.692 (95% CI: 0.641 to 0.739), 0.813 (95% CI: 0.767 to 0.850) respectively (Fig. 3). The difference between the clinical model and the radiomics model showed no significance (P = 0.13, Delong test). When combined together, the clinical_radiomics model demonstrated higher predictive power than clinical model (P = 0.004) as well as the radiomics model (P < 0.001). Figure 3 shows the ROC curve of the three models in the training set and validation set. For the clinical_radiomics model, the sensitivity and specificity were 92.5% and 58.7%, respectively.
A total clinical_radiomics score was computed by the coefficient in the final logistic regression clinical_radiomics model. A cutoff value was defined as the corresponding score when maximum value of sensitivity plus specificity was achieved. The change of sensitivity and specificity with the clinical_radiomics score and the distribution of clinical_radiomics score in two groups are presented in supplementary material Fig. S1. All cases were divided into the lower-risk and high-risk groups with a cutoff clinical_radiomcis score of 2.3. The Kaplan-Meier survival analysis on Day 28 after ARDS onset is shown in Fig. 4A, with the log-rank test (P < 0.001).
When taking the survival status on Day 10 after ICU admission as the secondary outcome, the APACHE Ⅱ score, Radiomics_Score_2, Radiomics_Score_3 also showed statistical significance.
The odds ratios of these factors were APACHE Ⅱ score (OR, 1.108, 95% CI: 1.067 to 1.149, P < 0.001), Radiomics_Score_2 (OR, 1.906, 95% CI: 1.044 to 3.478, P = 0.03), Radiomics_Score_3 (OR, 1.665, 95% CI: 1.189 to 2.332, P = 0.003). The AUC was 0.791 (95% CI: 0.746 to 0.831), sensitivity was 76.8% and specificity was 71.0%.
With the cutoff clinical_radiomics score of 2.3, the high-risk and low-risk groups also showed significant difference by Kaplan-Meier survival analysis, shown in Fig. 4B (P < 0.001, log-rank test).
However, the correlation between these factors and the ICU length of stay or hospitalization was relatively low.
Models constructed by clinical information (APACHE Ⅱ score) and each ventilator parameter were summarized in Table 3. The Tv and Tv/PBW models were ranked the top two among all models (0.773, 95% CI: 0.726 to 0.815 and 0.770, 95% CI: 0.723 to 0.812). The AUCs of the remaining models were all less than 0.770.
Table 3
Performance of models in predicting 28-Day mortality
| AUC | Lower limit of 95% CI | Upper limit of 95% CI |
Clinical model | 0.758 | 0.710 | 0.802 |
Ventilator parameters models | | | |
Tidal Volume(ml) | 0.773 | 0.726 | 0.815 |
Tidal Volume/PBW(ml/kg) | 0.770 | 0.723 | 0.812 |
Respiratory rate (/min) | 0.765 | 0.718 | 0.807 |
Peak Inspiratory Pressure (cmH20) | 0.762 | 0.715 | 0.805 |
Positive End-Expiratory Pressure (cmH20) | 0.762 | 0.715 | 0.805 |
Driving pressure (cmH20) | 0.762 | 0.714 | 0.804 |
Mechanical power (J/min) | 0.769 | 0.722 | 0.811 |
Mechanical power /PBW (10− 3 J/min/kg) | 0.769 | 0.722 | 0.811 |
Radiomics model | 0.692 | 0.641 | 0.739 |
Clinical_Radiomics model | 0.813 | 0.767 | 0.850 |
Note. —The clinical model is based on APACHE Ⅱ score; The ventilator parameters models include the APACHE Ⅱ score and each of the individual ventilator parameters. The radiomics model comprises radiomics_score_2, radiomics_score_3. PBW = Predicted body weight; AUC = Area under the curve; CI = Confidence interval. |
When comparing the top two ventilator models with the clinical, radiomics, and clinical_radiomics models, the clinical_radiomcis model showed the best performance (P = 0.02 with Tv model and P = 0.01 with Tv/ PBW model).