Quantitative analysis of pulmonary vascular alterations in children with refractory Mycoplasma pneumoniae pneumonia

Introduction

Community-acquired pneumonia (CAP) is the leading cause of pediatric hospitalization, with Mycoplasma pneumoniae (MP) infection accounting for 20–40% of pediatric CAP cases (1-3). Although MP pneumonia (MPP) is generally self-limiting, 6.83–40.84% of cases can progress to refractory MPP (RMPP), significantly increasing the risk of pulmonary and extrapulmonary complications, including pulmonary necrosis and systemic embolic events (4-8). RMPP is characterized by persistent fever and worsening clinical symptoms and radiological findings even after ≥7 days of appropriate macrolide therapy, often requiring immune modulators or second-line antibiotics (9,10). To prevent disease progression and minimize associated complications, early recognition and diagnosis of RMPP are essential.

Moreover, patients with MPP are known to exhibit a hypercoagulable state (11). Recent studies have found that patients with RMPP exhibit higher levels of C-reactive protein (CRP), lactate dehydrogenase (LDH), neutrophil ratio (NR), and D-dimer as compared to those with non-RMPP, suggesting a heightened inflammatory and hypercoagulable state (12-14). Indeed, pulmonary embolisms have been reported in patients with RMPP, involving lobar, segmental, and even subsegmental or more distal arteries (15-17). Pathological findings from a mouse model of MP infection further indicated the narrowing of pulmonary arterioles (<500 µm diameter) (18). However, conventional computed tomography (CT), which primarily captures macroscopic parenchymal changes, lacks the resolution to visualize these microvessels (19,20). Consequently, it remains unclear whether the pronounced hypercoagulability and inflammation observed in patients with RMPP also impact the pulmonary microvasculature in affected children.

Recent advancements in quantitative vascular imaging have enhanced the understanding of pulmonary microvascular abnormalities in pulmonary diseases (21,22). In coronavirus disease 2019 (COVID-19), CT-derived pulmonary blood volume (PBV) analysis, combined with deep learning-based vascular segmentation models, has uncovered significant microvascular changes, including reduced perfusion in small vessels and compensatory dilation of larger vessels. These findings have demonstrated strong prognostic value and reflected pathological microvascular changes associated with endothelial injury and a hypercoagulable state in patients with COVID-19 (22,23). Importantly, this approach uses routine CT scans, avoiding additional radiation exposure, which is especially important for pediatric patients. Given the shared mechanisms of endothelial inflammation and hypercoagulability between COVID-19 and RMPP (24-26), we hypothesize that similar microvascular abnormalities may also occur in patients with RMPP. Specifically, we propose that CT-derived quantitative PBV parameters—particularly the proportion of PBV contained in small
(<5 mm²) vs. large (>10 mm²) vessels—may differ between RMPP and non-RMPP patients and reflect microvascular and hemodynamic alterations.

Therefore, this study conducted quantitative pulmonary vascular analysis to further investigate the pulmonary microvascular changes in patients with RMPP and to evaluate the value of PBV parameters in predicting RMPP occurrence. We present this article in accordance with the STROBE reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1008/rc).

Methods

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments and was approved by the Institutional Ethics Board of Children’s Hospital, Zhejiang University School of Medicine (approval No. 2024-IRB-0189-P-01), and the Institutional Ethics Board of Rizhao Hospital of Traditional Chinese Medicine (approval No. 2024-IRB-039). The requirement for informed consent was waived due to the retrospective nature of the analysis.

Study design and patient selection

This retrospective study enrolled consecutive pediatric patients diagnosed with MPP from two medical institutions, forming a cross-validation cohort and an external testing cohort. The cross-validation cohort included patients admitted to the general inpatient units of a tertiary pediatric specialty hospital from July 2019 to December 2023 (Figure 1A). The external testing cohort comprised patients hospitalized in the pediatric department of a tertiary general hospital between January 2023 and April 2024 (Figure 1B).

Figure 1 Flowchart of study enrollment. (A) The data collection process used for the cross-validation cohort. (B) The data collection process used for the external testing cohort. CT, computed tomography.

The inclusion criteria were as follows: (I) pediatric patients aged >28 days to ≤18 years; (II) confirmed clinical diagnosis of MPP; and (III) availability of baseline chest CT scans during hospitalization for all cohorts, with follow-up CT scans within 2 months after discharge required only for the cross-validation cohort. The exclusion criteria were as follows: (I) inadequate quality of baseline or follow-up CT images for quantitative analysis; (II) complete absence of laboratory examination data; (III) the presence of immunodeficiency or documented prior use of immunosuppressive treatment; and (IV) the need for mechanical ventilation support during hospitalization. Ultimately, the cross-validation and external testing cohorts comprised 512 and 124 patients, respectively. The diagnostic criteria for MPP can be found in Appendix 1.

Data collection and chest CT acquisition

All patients included in this study had a confirmed discharge diagnosis of MPP. Clinical, etiological, and laboratory data were retrospectively collected from the digital hospital information system. Clinical variables included demographic characteristics, preexisting comorbidities, duration of fever and cough prior to admission, preadmission treatments, initial vital signs, and the requirement for oxygen support upon admission. Disease severity was classified according to the 2023 Chinese guidelines for pediatric MPP (27), as detailed in Appendix 2. Pathogen detection results, including those for viral, bacterial, and atypical pathogens, were documented, as detailed in Appendix 3. The laboratory variables obtained on admission included complete blood count, liver and renal function tests, inflammatory markers, coagulation tests, and electrolytes.

Data on clinical outcomes, complications, and in-hospital treatments (e.g., glucocorticoids, intravenous immunoglobulin, and bronchoscopy) were extracted from discharge records. Regarding outcome, RMPP was defined as persistent fever, worsening clinical symptoms, progression on pulmonary imaging, or extrapulmonary complications after ≥7 days of macrolide therapy (9), and patients were categorized into RMPP and non-RMPP groups.

The clinical indications for performing chest CT during hospitalization in patients with MPP are detailed in Appendix 4. Thin-section (≤1.5 mm), non-contrast chest CT images were obtained from the picture archiving and communication system for analysis. The earliest chest CT obtained during hospitalization was selected as the baseline CT. The median interval from symptom onset to baseline CT was 7.44 days [interquartile range (IQR), 5.50–9.42 days] for the cross-validation cohort and 9.00 days (IQR, 7.00–12.00 days) for the external testing cohort. For the cross-validation cohort, both baseline and follow-up chest CT scans within 2 months after discharge were retrospectively collected. For the external testing cohort, only baseline chest CT scans were acquired.

Quantitative pulmonary vascular analysis

PBVs were automatically segmented via a previously validated UV-Net-based segmentation model (28), with the detailed methodological procedures provided in Appendix 5. The segmented pulmonary vessels were subsequently classified into three categories according to their cross-sectional areas: vessels with cross-sectional areas <5 mm² (BV5), 5–10 mm² (BV5–10), and >10 mm² (BV10) (Figure 2). The blood volume within each category was calculated and expressed as a percentage of the total PBV, with BV5, BV5–10, and BV10 denoted as BV5%, BV5–10%, and BV10%, respectively.

Figure 2 Volume-rendered chest CT images of a 5-year-old girl with RMPP and atelectasis at baseline (A) and follow-up (B), with color-coded segmentation indicating pulmonary vascular cross-sectional areas. (A) The baseline CT scan showed BV5% of 57.53%, whereas (B) the follow-up CT revealed an increase in BV5% to 71.16%. The color-coded representation differentiates blood volume in vessels with cross-sectional areas of <5, 5–10, and >10 mm². BV5%, the percentage of PBV contained in vessels with a cross-sectional area less than

5 mm²; CT, computed tomography; PBV, pulmonary blood volume; RMPP, refractory Mycoplasma pneumoniae pneumonia.

Development and validation of RMPP predictive models

Predictive models for RMPP were developed with clinical variables and PBV parameters. Initially, candidate predictors collected at admission—including demographics, clinical symptoms, prior treatments, vital signs, laboratory tests, and PBV parameters—were screened, while those with >40% missing data in either the cross-validation cohort or the external testing cohort were excluded. Least absolute shrinkage and selection operator (LASSO) regression was subsequently applied in the cross-validation cohort, and 19 predictors were selected (16 clinical and 3 PBV parameters).

The extreme gradient boosting (XGBoost) algorithm was selected for model development. Additional details on the selection of algorithms can be found in Appendix 6 and Table S1. Based on the selected predictors, two models were constructed: (I) a clinical-only model comprising the 16 selected clinical variables; and (II) a clinical-blood volume (clinical-BV) model combining all 19 variables. For model development, model hyperparameters were optimized via five-fold cross-validation and retrained on the full cross-validation cohort via the identified optimal hyperparameters.

Model performance was evaluated internally and externally via receiver operating characteristic (ROC) curve analysis, with the area under the ROC curve (AUC), accuracy, precision, sensitivity, specificity, and F1 score being calculated for both cohorts. Differences in the AUC values between the clinical and clinical-BV models were compared via the DeLong test. SHapley Additive exPlanations (SHAP) values were computed to quantify feature importance in the clinical-BV model (29).

Statistical analysis

Statistical analyses were performed via R software version 4.3.1 (The R Foundation for Statistical Computing, Vienna, Austria). Categorical variables were compared via the Chi-squared test, whereas continuous variables were compared via the Student t-test (normally distributed data) or the Wilcoxon rank-sum test (nonnormally distributed data). Multiple comparisons were corrected via the Holm method. Missing data were imputed via multiple imputation methods via R’s “mice” package.

Associations between PBV parameters and RMPP were conducted within the cross-validation cohort. Numeric variables were standardized through Z-score normalization, and univariate logistic regression (LR) was performed to preliminarily examine the associations. Clinically relevant variables with <40% missing data were selected via LASSO regression (“glmnet” R package) and included in the multivariable models to independently evaluate PBV parameters. Additionally, based on the age distribution in the cross-validation cohort (Figure S1), age-related immune differences, and higher MPP prevalence in children over 5 years old (30), subgroup analysis was performed for the 5- to 10-year age group. Pearson correlation further assessed relationships between significant PBV parameters and key laboratory indices (D-dimer, CRP, and LDH), as well as duration of fever.

XGBoost models were developed using the “caret” and “xgboost” R packages, SHAP analysis was completed via the “SHAPforxgboost” R package, and ROC analysis was conducted via the “pROC” R package. P values less than 0.05 were considered statistically significant.

Results

Comparison of clinical and laboratory characteristics between RMPP and non-RMPP

As presented in Table 1, the cross-validation cohort comprised 512 patients (256 males and 256 females; median age 6.42 years; IQR, 4.33–8.25 years), with 265 classified as non-RMPP and 247 classified as RMPP. Patients in the RMPP group demonstrated significantly longer durations of fever and cough before admission, along with a higher incidence of severe/critical pneumonia. Moreover, these patients were more likely to require oxygen support and exhibited elevated body temperature, pulse rate, and respiratory rate (all P values ≤0.01). In terms of clinical outcomes (Table S2), the RMPP group experienced significantly longer hospital stays, extended total fever duration, and a higher rate of pulmonary complications than did the non-RMPP group (all P values ≤0.03).

In terms of etiology (Table S3), no statistically significant differences were observed between the groups (all P values >0.05). Laboratory findings (Table S4) indicated that patients with RMPP had significantly higher CRP levels, erythrocyte sedimentation rates (ESRs), neutrophil counts, LDH levels, and D-dimer levels (all P values ≤0.03).

Analysis of PBV parameters at baseline and follow-up

As depicted in Table 1 and Figure 3A, the BV5% was markedly lower in the RMPP group than in the non-RMPP group (58.50% vs. 60.63%, P=0.007), whereas the proportion of the BV5–10% and BV10% were significantly higher in the RMPP group than in the non-RMPP group (median BV5–10%: 15.74% vs. 15.13%, P=0.03; median BV10%: 20.90% vs. 19.39%, P=0.004).

Figure 3 Pulmonary vascular volume changes between patients with RMPP and those with non-RMPP. (A) A bar plot comparing the proportion of blood volume in three vessel-size categories on baseline CT between the RMPP and non-RMPP groups. The three categories include BV5%, BV5–10%, and BV10%. (B) A box plot illustrating the BV10% distribution in the non-RMPP and RMPP groups at both the baseline and follow-up CT scans. *, P<0.05; **, P<0.01; ***, P<0.001. BV5%, the percentage of PBV contained in vessels with a cross-sectional area less than 5 mm²; BV5–10%, the percentage of PBV contained in vessels with a cross-sectional area between 5 and 10 mm²; BV10%, the percentage of PBV contained in vessels with a cross-sectional area greater than 10 mm²; CT, computed tomography; PBV, pulmonary blood volume; RMPP, refractory Mycoplasma pneumoniae pneumonia.

Follow-up CT was performed with a median interval of 16.00 days (IQR, 13.00–23.00 days). As illustrated in Table S5 and Figure 3B, follow-up CT scans revealed a significant increase in BV5% relative to baseline (P<0.001). In contrast, both the BV5–10% and BV10% values were significantly lower on follow-up CT than at baseline (all P values <0.01), suggesting pulmonary blood distribution may recover over time. Notably, during follow-up, the RMPP group, as compared to the non-RMPP group, still had a lower BV5% (median: 62.42% vs. 63.55%, P=0.03) and a greater BV10% (median: 19.08% vs. 18.31%, P=0.003) (Table 2).

Correlation between RMPP and quantitative PBV features and their associations with RMPP-related risk factors

As shown in Table 3, in the cross-validation cohort, the univariate analyses revealed that each PBV parameter was significantly associated with the occurrence of RMPP
(all P values ≤0.03). In subsequent multivariate LR analyses, after the clinical covariates were adjusted for, BV5% remained an independent protective factor against RMPP [odds ratio (OR) =0.70; 95% confidence interval (CI): 0.54–0.89; P=0.005], whereas BV10% emerged as an independent risk factor (OR =1.49; 95% CI: 1.17–1.92; P=0.002). Similarly, among patients aged 5–10 years, after adjustments were made for the relevant covariates, BV5% remained an independent protective factor against RMPP (OR =0.68; 95% CI: 0.49–0.83; P=0.018), whereas BV10% remained a significant risk factor (OR =1.58; 95% CI: 1.15–2.19; P=0.005).

Moreover, as depicted in Figure 4, correlation analyses revealed that BV5% was significantly inversely correlated with key inflammatory and coagulation biomarkers, including NR (r=−0.12), CRP (r=−0.13), and D-dimer (r=−0.19), respectively (all P values <0.01); conversely, BV10% was significantly positively correlated with NR (r=0.14), CRP (r=0.14), and D-dimer (r=0.21) (all P values <0.01).

Figure 4 Pearson correlation between quantitative pulmonary vascular parameters and established risk factors for RMPP. The scatter plots illustrate the associations of BV5% and BV10% with key clinical indicators, including the NR, CRP level, D-dimer, LDH level, and fever duration before admission. The data points are color-coded for BV5% (red) and BV10% (blue), with linear regression trendlines fitted for each parameter. All variables were standardized via Z-score normalization to ensure comparability. BV5%, the percentage of PBV contained in vessels with a cross-sectional area less than 5 mm²; BV10%, the percentage of PBV contained in vessels with a cross-sectional area greater than 10 mm²; CRP, C-reactive protein; LDH, lactate dehydrogenase; NR, neutrophil ratio; PBV, pulmonary blood volume; RMPP, refractory Mycoplasma pneumoniae pneumonia.

XGBoost-based RMPP prediction enhanced by quantitative PBV parameters

To improve the predictive performance for RMPP, we incorporated both clinical variables and quantitative PBV features into XGBoost-based models. Feature selection via LASSO regression identified 19 variables, including 16 clinical and 3 PBV-derived features (Figure 5A).

Figure 5 Model interpretation and performance evaluation. (A) Ranking of the 19 selected features by the mean absolute SHAPs value in the XGBoost model. The leftmost numbers indicate the mean absolute SHAP values, which represent each feature’s relative contribution to the prediction of RMPP. Additionally, each dot corresponds to an individual sample, with the X-axis denoting the SHAP value. (B) ROC curves for the cross-validation cohort and external testing cohort comparing the predictive performance of the clinical-BV model, which incorporates 19 variables (16 clinical variables and 3 blood volume-derived quantitative features), with that of the clinical-only model, which uses only the 16 clinical variables. ADA, adenosine deaminase; AUC, area under the ROC curve; BV5%, the percentage of PBV contained in vessels with a cross-sectional area less than 5 mm²; BV5–10%, the percentage of PBV contained in vessels with a cross-sectional area between 5 and 10 mm²; BV10%, the percentage of PBV contained in vessels with a cross-sectional area greater than 10 mm²; CK, creatine kinase; clinical-BV, clinical-blood volume; CRP, C-reactive protein; GGT, gamma-glutamyl transferase; HGB, hemoglobin; PBV, pulmonary blood volume; PDW, platelet distribution width; PLT, platelet count; RMPP, refractory Mycoplasma pneumoniae pneumonia; ROC, receiver operating characteristic; SHAPs, SHapley Additive exPlanations; XGBoost, extreme gradient boosting.

As shown in Table 4 and Figure 5B, in the cross-validation cohort, the clinical-BV model demonstrated a significantly higher AUC (AUC =0.91) than did the clinical-only model (AUC =0.88) (P<0.001) and better accuracy, precision, sensitivity, specificity, and F1 score.

SHAP analysis of the clinical-BV model (Figure 5A) revealed that the five most influential features in the cross-validation cohort were duration of azithromycin therapy before admission, body temperature, presence of severe pneumonia, CRP level, and BV10%, highlighting the importance of both inflammation markers and PBV features.

To assess the generalizability of the model, we evaluated its performance in an independent external testing cohort (Table S6), which included 124 patients (median age 7.00 years, IQR, 5.00–9.00 years; 52% male). In this cohort, although the difference was not statistically significant (P=0.09), the clinical-BV model had a higher AUC (AUC =0.84) than did the clinical-only model (AUC =0.81), with superior performance in accuracy, precision, specificity, and F1 score.

Discussion

In this study, patients with RMPP exhibited consistent microvascular remodeling patterns on CT scans, characterized by decreased BV5% (micro vessels <5 mm²) and increased BV10% (macro vessels >10 mm²) at both baseline and follow-up. These changes indicate altered distribution of pulmonary blood vessel volumes and were observed in patients with acute-phase pulmonary microcirculatory disturbances during the COVID-19 pandemic (22). Furthermore, multivariate analysis further revealed that BV5%, which was negatively correlated with inflammatory and coagulation biomarkers, was as an independent protective predictor for RMPP. In contrast, BV10% emerged as an independent risk factor and was positively correlated with these same biomarkers. When integrated into XGBoost-based models, these quantitative PBV parameters substantially improved RMPP prediction in the cross-validation cohort as compared to models using only clinical variables. Although the difference in AUC in the external testing cohort was not statistically significant, the clinical-BV model nonetheless surpassed the clinical-only model in accuracy, precision, specificity, and F1 score. These findings suggest that incorporating PBV parameters can enhance the predictive performance for RMPP and maintain clinical applicability in independent cohorts.

As BV5% represents the proportion of vessels with cross-sectional areas less than 5 mm² relative to total PBV, the significant decrease in BV5% observed in our study indicates a reduction in pulmonary microvascular volume in patients with RMPP. Consistent with these CT findings, histopathological analyses in murine models of repeated MP infections suggest the presence of pulmonary arteriole wall thickening and luminal narrowing, which can lead to increased pulmonary arterial pressure (18). Microscopically, pulmonary arterioles are defined as vessels measuring less than 500 µm in diameter, corresponding to a subset of the vasculature categorized as BV5% on CT (31). However, given that our measurement of BV5% was based solely on morphological criteria without differentiation between arteries and veins, it remains unclear whether the observed reduction represents a loss of microarterioles or microvenules. Nonetheless, the aforementioned pathological findings strongly suggest that the microvascular changes observed in patients with RMPP may arise from alterations in the pulmonary arteriolar. However, confirmation of this conclusion requires further analysis that differentiates between arteries and veins.

Given that the pathogenesis of MPP involves direct toxicity, immune-mediated injury, and vascular inflammation/thrombosis (25,26,32), we hypothesize that the observed reduction in BV5%—reflecting decreased microvascular volume in patients with RMPP—may be associated with the microvascular endothelial inflammation or thrombosis triggered by enhanced coagulation and inflammatory states. Meanwhile, the observed increase in BV10% could reflect the compensatory dilation of larger vessels secondary to reduced microvascular volume. This hypothesis is supported by three lines of indirect evidence. First, patients with RMPP in the cross-validation cohort had significantly elevated D-dimer levels, CRP levels, ESRs, and neutrophil counts. Second, correlation analyses further reinforced these findings by revealing a negative association between BV5% and markers of inflammation or coagulation. Finally, all recorded vascular thrombosis events occurred in the RMPP group. Additionally, SHAP value analysis of the XGBoost model revealed that BV10% held a higher importance than did BV5%. This observation suggests that large-vessel dilation is an active compensatory response, not merely a passive consequence. However, further studies are needed to confirm these proposed mechanisms and establish definitive causality.

Compared with traditional volumetric analyses of pneumonia, such as consolidation volume, quantitative pulmonary vascular analysis can reveal more subtle pathophysiological changes. Conventional volumetric analyses primarily reflect parenchymal tissue injury but offer limited insights into the underlying pathogenic mechanisms (33). In contrast, pulmonary vascular analysis, particularly through reductions in BV5%, effectively captures the decrease in microvascular volume among patients with RMPP. This reduction highlights the specific pathophysiological processes in RMPP, in which heightened coagulation and inflammation states may affect pulmonary microvasculature, resulting in diminished microvascular volume (25). Moreover, incorporating PBV parameters into clinical predictive models improves early RMPP identification. The clinical-BV model demonstrated superior performance as compared to the clinical-only model (AUC: 0.91 vs. 0.88). This model may allow for the timely use of immunomodulators or second-line antibiotics and potentially prevent complications such as necrotizing pneumonia. Moreover, these quantitative vascular measurements can be obtained from routine CT scans without extra examinations or additional radiation exposure.

Patients with RMPP continued to exhibit relative reductions in BV5% at follow-up (median duration 16 days) as compared to non-RMPP patients, mirroring the vascular recovery patterns observed in post-COVID-19 studies, where residual vascular anomalies persisted in 87.4% of patients at 6 months (34). Notably, post-COVID-19 research also indicates that even after complete radiological resolution at 12–16 weeks after discharge, some patients experience dyspnea along with reductions in capillary blood volume and diffusing capacity of the lungs for carbon monoxide (DL_CO) (35). This further underscores the necessity of ongoing vascular and pulmonary functional monitoring beyond symptomatic and radiological recovery.

Our study involved several limitations that should be acknowledged. First, the segmentation software used does not differentiate between arteries and veins, but these may exhibit distinct pathological responses to MP infection. This also limited our ability to examine the specific pathophysiological mechanisms related to microvascular alteration. In future studies, pulmonary vessel segmentation with artery–vein differentiation will allow for a more precise analysis. Second, the absence of a normal pediatric control group limited our ability to establish whether MPP induces alterations in PBV distribution relative to healthy children. Finally, clinical symptoms and pulmonary function data were not systematically collected, which limited our ability to comprehensively assess the association between changes in PBV parameters, clinical disease progression, and long-term pulmonary function. Future studies should incorporate longitudinal imaging, comprehensive clinical evaluations, and pulmonary function tests—such as the ratio of forced expiratory volume in 1 second to forced vital capacity and DL_CO—to better clarify the physiological impact of pulmonary microvascular alterations.

Conclusions

This study demonstrated that pediatric RMPP is associated with significant pulmonary microvascular alterations, characterized by reduced BV5% and elevated BV10%. These quantitative PBV parameters could serve as independent RMPP predictors, enhance the performance of XGBoost-based predictive models, and can be easily integrated into routine CT workflows. Assessing pulmonary vascular changes through routine CT imaging provides a noninvasive approach for identifying high-risk patients and guiding early risk stratification and intervention.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1008/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1008/dss

Funding: This work was supported by the National Natural Science Foundation of China (NSFC) (No. 82171934) to L.S., the Medical Health Science and Technology Project of Zhejiang Provincial Health Commission (No. 2023KY832) to G.Y., and Peking Union Medical College Hospital Talent Cultivation Program Category C (No. UBJ10148) to L.S.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1008/coif). Z.Z. and W.T. are currently employed by Beijing Deepwise and League of PhD Technology Co., Ltd. The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the institutional ethics board of Children’s Hospital, Zhejiang University School of Medicine (approval No. 2024-IRB-0189-P-01) and the institutional ethics board of Rizhao Hospital of Traditional Chinese Medicine (approval No. 2024-IRB-039). Individual consent for this retrospective analysis was waived.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Principi N, Esposito S. Management of severe community-acquired pneumonia of children in developing and developed countries. Thorax 2011;66:815-22. [Crossref] [PubMed]
2. Wang YS, Zhou YL, Bai GN, Li SX, Xu D, Chen LN, et al. Expert consensus on the diagnosis and treatment of macrolide-resistant Mycoplasma pneumoniae pneumonia in children. World J Pediatr 2024;20:901-14. [Crossref] [PubMed]
3. Lee GE, Lorch SA, Sheffler-Collins S, Kronman MP, Shah SS. National hospitalization trends for pediatric pneumonia and associated complications. Pediatrics 2010;126:204-13. [Crossref] [PubMed]
4. Lemaître C, Angoulvant F, Gabor F, Makhoul J, Bonacorsi S, Naudin J, Alison M, Faye A, Bingen E, Lorrot M. Necrotizing pneumonia in children: report of 41 cases between 2006 and 2011 in a French tertiary care center. Pediatr Infect Dis J 2013;32:1146-9. [Crossref] [PubMed]
5. Zheng HQ, Ma YC, Chen YQ, Xu YY, Pang YL, Liu L. Clinical Analysis and Risk Factors of Bronchiolitis Obliterans After Mycoplasma Pneumoniae Pneumonia. Infect Drug Resist 2022;15:4101-8. [Crossref] [PubMed]
6. Ling Y, Ning J, Xu Y. Explore the Predictive Value of Peripheral Blood Cell Parameters in Refractory Mycoplasma pneumoniae Pneumonia in Children Over 6 Years Old. Front Pediatr 2021;9:659677. [Crossref] [PubMed]
7. Yang B, Zhang W, Gu W, Zhang X, Wang M, Huang L, Zhu C, Yan Y, Ji W, Ni H, Chen Z. Differences of clinical features and prognosis between Mycoplasma pneumoniae necrotizing pneumonia and non-Mycoplasma pneumoniae necrotizing pneumonia in children. BMC Infect Dis 2021;21:797. [Crossref] [PubMed]
8. Zhang XX, Gu WJ, Chen ZR, He YY, Hao CL, Yan YD, Zhu CH, Wang YQ, Huang L, Ji W. Epidemiological analysis of refractory mycoplasma Pneumoniae pneumonia in children in Suzhou from 2011 to 2015. Journal of Pediatric Pharmacy 2019;25:7-10.
9. Tamura A, Matsubara K, Tanaka T, Nigami H, Yura K, Fukaya T. Methylprednisolone pulse therapy for refractory Mycoplasma pneumoniae pneumonia in children. J Infect 2008;57:223-8. [Crossref] [PubMed]
10. Tong L, Huang S, Zheng C, Zhang Y, Chen Z. Refractory Mycoplasma pneumoniae Pneumonia in Children: Early Recognition and Management. J Clin Med 2022;11:2824. [Crossref] [PubMed]
11. Li T, Yu H, Hou W, Li Z, Han C, Wang L. Evaluation of variation in coagulation among children with Mycoplasma pneumoniae pneumonia: a case-control study. J Int Med Res 2017;45:2110-8. [Crossref] [PubMed]
12. Li M, Wei X, Zhang SS, Li S, Chen SH, Shi SJ, Zhou SH, Sun DQ, Zhao QY, Xu Y. Recognition of refractory Mycoplasma pneumoniae pneumonia among Myocoplasma pneumoniae pneumonia in hospitalized children: development and validation of a predictive nomogram model. BMC Pulm Med 2023;23:383. [Crossref] [PubMed]
13. Huang W, Xu X, Zhao W, Cheng Q. Refractory Mycoplasma Pneumonia in Children: A Systematic Review and Meta-analysis of Laboratory Features and Predictors. J Immunol Res 2022;2022:9227838. [Crossref] [PubMed]
14. Gong H, Sun B, Chen Y, Chen H. The risk factors of children acquiring refractory mycoplasma pneumoniae pneumonia: A meta-analysis. Medicine (Baltimore) 2021;100:e24894. [Crossref] [PubMed]
15. Fu Y, Zhang TQ, Dong CJ, Xu YS, Dong HQ, Ning J. Clinical characteristics of 14 pediatric mycoplasma pneumoniae pneumonia associated thrombosis: a retrospective study. BMC Cardiovasc Disord 2023;23:1. [Crossref] [PubMed]
16. Liu J, He R, Wu R, Wang B, Xu H, Zhang Y, Li H, Zhao S. Mycoplasma pneumoniae pneumonia associated thrombosis at Beijing Children's hospital. BMC Infect Dis 2020;20:51. [Crossref] [PubMed]
17. Chen S, Ke S, Vinturache A, Dong X, Ding G. Pulmonary embolism associated with Mycoplasma pneumoniae pneumonia in children. Pediatr Pulmonol 2023;58:3605-8. [Crossref] [PubMed]
18. Liu DX, Peng DX, Chen R, Lei HT, Che DY, Zhao SY. Chronic pulmonary infection caused by Mycoplasma pneumoniae leading to pulmonary arteriole remodeling and pulmonary hypertension in rats. J Tongji Med Univ 1995;15:223-6. [Crossref] [PubMed]
19. Estépar RS, Kinney GL, Black-Shinn JL, Bowler RP, Kindlmann GL, Ross JC, et al. Computed tomographic measures of pulmonary vascular morphology in smokers and their clinical implications. Am J Respir Crit Care Med 2013;188:231-9. [Crossref] [PubMed]
20. Yoon SH, Lee JH, Kim BN, Chest CT. Findings in Hospitalized Patients with SARS-CoV-2: Delta versus Omicron Variants. Radiology 2023;306:252-60. [Crossref] [PubMed]
21. Washko GR, Nardelli P, Ash SY, Vegas Sanchez-Ferrero G, Rahaghi FN, Come CE, et al. Arterial Vascular Pruning, Right Ventricular Size, and Clinical Outcomes in Chronic Obstructive Pulmonary Disease. A Longitudinal Observational Study. Am J Respir Crit Care Med 2019;200:454-61. [Crossref] [PubMed]
22. Poletti J, Bach M, Yang S, Sexauer R, Stieltjes B, Rotzinger DC, Bremerich J, Walter Sauter A, Weikert T. Automated lung vessel segmentation reveals blood vessel volume redistribution in viral pneumonia. Eur J Radiol 2022;150:110259. [Crossref] [PubMed]
23. Morris MF, Pershad Y, Kang P, Ridenour L, Lavon B, Lanclus M, Godon R, De Backer J, Glassberg MK. Altered pulmonary blood volume distribution as a biomarker for predicting outcomes in COVID-19 disease. Eur Respir J 2021;58:2004133. [Crossref] [PubMed]
24. Ackermann M, Verleden SE, Kuehnel M, Haverich A, Welte T, Laenger F, Vanstapel A, Werlein C, Stark H, Tzankov A, Li WW, Li VW, Mentzer SJ, Jonigk D. Pulmonary Vascular Endothelialitis, Thrombosis, and Angiogenesis in Covid-19. N Engl J Med 2020;383:120-8. [Crossref] [PubMed]
25. Georgakopoulou VE, Lempesis IG, Sklapani P, Trakas N, Spandidos DA. Exploring the pathogenetic mechanisms of Mycoplasmapneumoniae (Review). Exp Ther Med 2024;28:271. [Crossref] [PubMed]
26. Narita M. Pathogenesis of extrapulmonary manifestations of Mycoplasma pneumoniae infection with special reference to pneumonia. J Infect Chemother 2010;16:162-9. [Crossref] [PubMed]
27. National Health Commission of the People’s Republic of China. Guidelines for the diagnosis and treatment of Mycoplasma pneumoniae pneumonia in children (2023 edition). International Journal of Epidemiology and Infectious Disease 2023;50:79-85.
28. Bai Y, Wang X, Zhou Z, Wu Z, Feng Q, Qi J. Pulmonary segments segmentation with hierarchical weak labels. In: 2023 IEEE 20th International Symposium on Biomedical Imaging (ISBI). IEEE; 2023:1-5.
29. *Lundberg SM, Lee SI. A unified approach to interpreting model predictions. In: Advances in Neural Information Processing Systems 30 (NIPS 2017). 2017:4768-77.
30. Yan C, Xue GH, Zhao HQ, Feng YL, Cui JH, Yuan J. Current status of Mycoplasma pneumoniae infection in China. World J Pediatr 2024;20:1-4. [Crossref] [PubMed]
31. Rahaghi FN, Argemí G, Nardelli P, Domínguez-Fandos D, Arguis P, Peinado VI, Ross JC, Ash SY, de La Bruere I, Come CE, Diaz AA, Sánchez M, Washko GR, Barberà JA, San José Estépar R. Pulmonary vascular density: comparison of findings on computed tomography imaging with histology. Eur Respir J 2019;54:1900370. [Crossref] [PubMed]
32. Hu J, Ye Y, Chen X, Xiong L, Xie W, Liu P. Insight into the Pathogenic Mechanism of Mycoplasma pneumoniae. Curr Microbiol 2022;80:14. [Crossref] [PubMed]
33. Qian Y, Tao Y, Wu L, Zhou C, Liu F, Xu S, Miao H, Gao X, Ge X. Model based on the automated AI-driven CT quantification is effective for the diagnosis of refractory Mycoplasma pneumoniae pneumonia. Sci Rep 2024;14:16172. [Crossref] [PubMed]
34. Mohamed I, de Broucker V, Duhamel A, Giordano J, Ego A, Fonne N, Chenivesse C, Remy J, Remy-Jardin M. Pulmonary circulation abnormalities in post-acute COVID-19 syndrome: dual-energy CT angiographic findings in 79 patients. Eur Radiol 2023;33:4700-12. [Crossref] [PubMed]
35. Dal Negro RW, Turco P, Povero M. Long-lasting dyspnoea in patients otherwise clinically and radiologically recovered from COVID pneumonia: a probe for checking persisting disorders in capillary lung volume as a cause. Multidiscip Respir Med 2022;17:875. [Crossref] [PubMed]