Model based on the automated AI-Driven CT quantification is effective for the diagnosis of refractory Mycoplasma pneumoniae pneumonia

doi:10.21203/rs.3.rs-3329686/v1

Refractory Mycoplasma pneumoniae pneumonia（RMPP）prediction is a challenging but clinically significant challenge. A model based on AI-derived quantitative determination of lung lesions extent on initial computed tomography (CT) scan and clinical indicators has the potential to facilitate early RMPP prediction in hospitalized children. In this study, we conducted a retrospective cohort as a training set including 126 children with M. pneumoniae pneumonia (MP) admitted to Children’s Hospital of Nanjing Medical University, China from January 2019 to December 2020. We defined a constant Φ which can combine the volume and CT value of pulmonary lesions and be further used to calculate the logarithm of Φ to the base of 2 (Log₂^Φ). Finally, a clinical-imaging prediction model was developed using Log₂^Φ and clinical characteristics. The area under the receiver operating characteristic curve (ROC-AUC) was applied to performance evaluation. Then we conducted a prospective study including 54 children with MP as a test set to validate the predictive model for RMPP from January to December 2021. The clinical model yielded the AUC of 0.810 and 0.782, and the imaging model yielded the AUC of 0.764 and 0.769 in the training set and the test set, respectively. The clinical-imaging model combining Log₂^Φ, temperature(T), aspartate aminotransferase (AST), preadmission fever duration (PFD), and preadmission macrolides therapy duration (PMTD) produced the highest AUC values of 0.897 and 0.895 in the training set and the test set, respectively. Our work demonstrated that using automated quantification of lung disease at CT combined with clinical data in MPP is useful to predict RMPP.

Health sciences/Medical research

Health sciences/Risk factors

Mycoplasma pneumoniae（MP） is one of the main pathogens of community-acquired pneumonia（CAP） in children. Mycoplasma pneumoniae pneumonia (MPP) accounts for 10% - 40% of the total lower respiratory tract infection in hospitalized children, and the infection rate of pneumonia in children over 5 years old can be as high as 50% ¹. Although MPP is usually considered a type of self-limiting disease, sometimes it can cause various pulmonary and extrapulmonary complications, such as necrotizing pneumonia, bronchitis obliterans, encephalitis, arthritis, pericarditis, hemolytic anemia, and develop into serious life-threatening pneumonia^2,3. Macrolides are the first choice for the treatment of Mycoplasma pneumoniae in children. Despite treatment with macrolide antibiotics for 7 days or more, some patients still have clinical and imaging progress, which is defined as refractory Mycoplasma pneumoniae pneumonia (RMPP) ⁴.However, in recent years, with the increase of drug-resistant Mycoplasma pneumoniae strains, the incidence rate of RMPP is rising ⁵. Therefore, it is very important to improve doctors' understanding of RMPP at an early stage. In recent years, some laboratory markers have been used for early diagnosis of RMPP, such as CRP, LDH, IL-10, D-dimer and so on ^6,7. The AUC values of these factors varied from 0.7 to 0.8^8,9,showing that the single factor's diagnostic capacity needs to be improved.

Chest computed tomography (CT) is of great value in the diagnosis and disease progression of RMPP patients ^10,11.At present, the CT of RMPP is usually interpreted manually by radiologists. It is very subjective, unable to accurately and quantitatively evaluate the severity of the disease, and it is also very time-consuming and inefficient. We have realized that artificial intelligence (AI) is expected to acquire quantitative CT features and forecast the development of severe illness early and non-invasively, offering a viable prognostic indication for therapeutic therapy of COVID-19 ¹². However, to the best of our knowledge, associations between AI-derived CT features quantifying pneumonia lesions and the risk of RMPP in patients with the MP infection have not yet been reported. If AI-derived features from CT at an early stage of MPP can be used to predict RMPP, they can be particularly beneficial because of their non-invasiveness and efficiency. Therefore, we wanted to explore an artificial intelligence method to quantify chest CT in children with mycoplasma pneumonia and verify its value in the early diagnosis of RMPP.

Patients and groups. We conducted a retrospective cohort study among children admitted to the Children’s Hospital of Nanjing Medical University with MMP from January 2019 to December 2020 and a prospective cohort study from January to December 2021. The diagnosis of MP infection was based on positive serologic test results (M. pneumoniae anti-MP positive and antibody titer ≥1:160) and/or positive results for M. pneumoniae on polymerase chain reaction tests of nasopharyngeal secretions. RMPP is defined as a case with prolonged fever accompanied by deterioration of radiological findings despite appropriate management and treatment with a macrolide antibiotic for≥7 days ¹³. Finally, 126 patients were included in the retrospective cohort study as a training set and 54 patients in the prospective study as a test set. Patients with the following diseases were excluded from our study: All known congenital disorders, metabolic disease, immunodeficiency disease, blood tumor disease, bronchopulmonary dysplasia, nervous system dysplasia, and epilepsy, among others. Informed consents were obtained from all the guardians of the pediatric patients and the study was approved by the institutional ethics committee of Children’s Hospital of Nanjing Medical University.

Data collection and study variables. The demographics, clinical information, laboratory data and CT scanning were collected from inpatient electronic medical records and retrospectively analyzed. Peripheral blood samples were obtained within 24 hours since admission for the determination of the C-reactive protein (CRP), white blood cells (WBC), hemoglobin (Hb), platelets (PLT), lactate dehydrogenase (LDH), neutrophil (NEUT), erythrocyte sedimentation rate (ESR), alanine aminotransferase (ALT), aspartate aminotransferase (AST), creatine phosphokinase isoenzyme (CK-MB), creatine kinase (CK), procalcitonin (PCT), Total Protein (TP) and albumin (ALB).

In addition, the following data were also recorded: the highest temperature on the first day of admission, heart rate and respiratory rate on the day of admission, preadmission fever duration (PFD), preadmission cough duration (PCD), and preadmission macrolides therapy duration (PMTD).

CT acquisition parameters. All RMPP patients included were scanned with a 64-CT scanner (Toshiba or Philips Healthcare) within 48h after MPP was diagnosed, with exposure parameters of 110 kV scanning tube voltage, 100 mAs reference tube current, layer thickness 2.5 mm, maximum scanning layer thickness 10 mm, interval 2.5 mm, pitch 0.9, thin layer reconstruction layer thickness after scanning was 1.5 mm. The scanning range was from the thoracic entrance to the bottom of the lungs.

AI-based quantization of CT images and processing. An AI-aided diagnosis system from Shenrui (Beijing, China) was used in this study. This medical image-aided diagnosis software（model: mids-pnas, version: v1.3.0.1) can automatically identify the image characteristics of inflammatory lesions in chest CT images by learning a large number of CT image data of pneumonia patients, and carry out fully automatic lesion delineation, lesion volume measurement and quantitative analysis of volume proportion. CT chest image analysis process: (1) Import the high-resolution CT (HRCT) digital images from patients, and carry out 3D reconstruction of the whole lung, left and right lungs, lobes, and segments separately. (2) Segmentation of pulmonary lesions. (3) Count the total volume of the segmented lesions in the whole lung, left lung, and right lung separately. (4) The 3D reconstructed lung images underwent fine segmentation. 1 cubic centimeter (cm³) of lesion served as 1 minimum counting unit, the average density in this cubic centimeter was defined as ρ (ρ=HU+1024, Hounsfield Unit， HU). （5）Finally, we define a constant Φ, Φ=∑(ρ1+ρ2+ρ3.......+ ρ n-1+ρn)（Figure 1）.(6) Logarithm of Φ to the base of 2 was used to analyze the data.

Statistical analysis. Statistical analysis was performed using the R 4.04 version (R Foundation for Statistical Computing, Vienna, Austria). Chi-square test or Fisher’s exact test was used for the nominal variable. T-test was used for continuous variables¹⁴. The “rmda” package was used to construct the decision curve. The “rms” package was used to construct the nomogram and corresponding calibration curve. The variables for the model of binomial logistic regression were selected through stepwise use of Akaike’s information criterion (AIC)¹⁵. Receiver operating characteristic curves (ROC) analysis was performed using the “pROC” package. DeLong test was used for statistical comparisons of ROC curves. P values less than 0.05 were considered statistically significant.

Patient characteristics and laboratory findings. According to the inclusion criteria of this study, 126 patients were enrolled in the retrospective cohort as training set, including 41 common Mycoplasma pneumoniae pneumonia (CMPP) and 85 RMPP and 54 patients were enrolled in the prospective cohort, including 17 CMPP and 37 RMPP as test set respectively. No significant difference was found in gender (P=0.268), and age (P=0.889) (Table 1).

The RMPP and CMPP were admitted for 24h to compare clinical symptoms, biochemistry levels, and routine blood markers. PFD, PMTD, the highest temperature within 24 hours of admission, heart rate, AST, CK-MB, and ALB were significantly different in the RMPP group compared to the CMPP group (P<0.05). Respiratory rate, WBC, CRP, Hb, PLT, D-dimer, ESR, PCT, ALT, Tb, CK, and PCD was not significantly different between these two groups (P>0.05) (Table 1).

Model	Training set			Test set
	CMPP(n=41)	RMPP(n=85)	P value	CMPP(n=17)	RMPP(n=37)	P value
Sex(male/female)	21/20	44/41	0.268	12/5	22/15	0.078
Age（years）	4.81±2.79	4.88±2.79	0.889	4.81±2.79	4.64±2.51	0.210
T （℃）	37.32±1.07	38.03±1.28	<0.001	37.51±1.30	37.85±1.35	0.388
HR	116.59±17.62	123.74±18.12	0.032	117.56±14.36	123.32±15.51	0.201
R	25.61±4.94	26.78±3.84	0.079	25.62±3.93	27.03±3.94	0.243
WBC（*10⁹/L）	9.51±4.56	9.71±3.72	0.818	10.25±4.26	9.37±4.66	0.508
NEUT（*10⁹/L）	9.19±4.93	9.18±5.23	0.997	10.12±3.80	9.03±5.01	0.392
Hb （g/L）	128.05±33.63	122.11±9.86	0.257	127.13±7.18	121.32±10.97	0.028
PLT （*10⁹/L）	277.64±122.79	288.94±129.45	0.625	287.68±119.95	294.73±154.05	0.859
CRP (mg/L)	19.95±18.07	24.94±36.02	0.289	23.00±19.28	24.54±30.54	0.828
LDH (U/L)	354.16±161.08	403.07±167.87	0.108	317.37±100.80	411.19±152.01	0.011
ESR (mm/h)	31.55±15.24	30.73±18.16	0.786	32.88±12.09	29.54±18.96	0.447
MP-DNA（*10⁴）	539.98±1598.42	1949.96±10820.71	0.232	249.61±646.86	876.30±21411.18	0.112
D-dimer (ng/L)	651.84±1532.27	673.84±828.26	0.929	532.31±1098.19	713.05±912.48	0.569
PCT (ng/mL)	0.19±0.37	0.17±0.34	0.786	0.18±0.19	0.13±0.15	0.386
ALT (U/L)	15.66±12.93	22.52±28.57	0.059	15.18±11.32	20.65±17.89	0.188
AST (U/L)	29.43±10.57	35.25±18.01	0.021	25.94±8.05	34.03±14.37	0.012
TP (g/L)	68.42±4.64	67.12±5.40	0.154	69.93±4.57	67.78±4.72	0.413
ALB (g/L)	42.73±3.13	40.99±4.37	0.010	42.02±2.77	41.34±3.21	0.439
CK-MB (U/L)	73.71±39.50	39.5±47.43	0.049	75.75±50.98	65.05±59.70	0.511
CK(U/L)	86.59±71.30	67.59±63.11	0.137	22.50±8.15	40.08±34.73	0.006
PFD（days）	5.07±2.71	7.04±4.21	<0.001	5.31±2.52	7.92±4.29	0.008
PCD（days）	9.36±7.08	11.10±6.02	0.326	12.19±9.83	13.49±28.77	0.808
PMTD（days）	2.22±2.33	4.71±3.76	<0.001	2.63±3.12	4.05±3.39	0.146

Table 1. Clinical characteristics of patients in the training and test cohort. Values are presented as mean ± SD. T temperature, RMPP refractory Mycoplasma pneumoniae pneumonia, CMPP Common Mycoplasma pneumoniae pneumonia, HR heart rate, R respiratory, WBC white blood cell, NEUT neutrophil, Hb hemoglobin, PLT platelets, CRP C-reactive protein, LDH lactate dehydrogenase, ESR erythrocyte sedimentation rate, PCT procalcitonin, ALT alanine aminotransferase, AST aspartate aminotransferase, TP Total Protein, ALB albumin, CK-MB creatine phosphokinase isoenzyme, CK creatine kinase, PFD preadmission fever duration, PCD

preadmission cough duration, PMTD preadmission macrolides therapy duration.

Chest CT Quantitative Analysis. Quantitative lung features on CT are summarized in Table 2. The RMPP had a higher volume of pulmonary lesions (P<0.001) and higher extent of lesions located in the lung of the right upper lobe(p<0.001) and right lower lobe (P<0.002) compared to the CMPP. The log₂^Φ of RMPP was higher than CMPP (P<0.001).

	Training set			Test set
	CMPP(n=41)	RMPP(n=85)	P value	CMPP(n=17)	RMPP(n=37)	P value
Log₂^Φ	15.63±1.31	16.78±1.08	<0.001	15.50±1.26	16.66±1.10	0.002
Total lung volume（cc）	827.93±368.40	858.92±447.52	0.672	954.93±419.52	799.47±325.31	0.199
Volume of lung infiltration（cc）	71.72±65.43	139.68±104.62	<0.001	66.77±74.78	127.96±94.10	<0.016
Percentages of lung infiltration（%）	5.17±5.45	8.98±6.50	<0.001	3.65±5.73	8.95±6.46	0.006
right upper lobe（%）	37.13±20.78	49.99±21.10	<0.001	7.57±13.72	23.71±25.09	<0.004
middle lobe of right（%）	11.92±24.24	18.87±28.69	0.101	7.79±21.38	18.00±29.22	0.229
right lower lobe（%）	11.23±18.17	28.30±31.47	0.002	8.87±17.64	32.69±33.37	0.001
left upper lobe（%）	10.51±22.09	19.45±29.49	0.083	6.37±21.71	13.50±23.13	0.269
left lower lobe（%）	13.44±23.68	22.85±28.39	0.046	17.46±28.09	17.83±28.49	0.966

Table 2. Quantitative CT parameters of patients in the training and test set. CT Computed tomography, CMPP Common Mycoplasma pneumoniae pneumonia, RMPP refractory Mycoplasma pneumoniae pneumonia.

Logistic regression and nomogram. We used the logistic regression model through stepwise use of Akaike’s information criterion (AIC) to select the variables for the clinical model, the imaging model and the clinical-imaging model. Once the AIC value no longer decreases, the stepwise regression analysis is terminated, and the optimal regression equation was output.

Multivariate logistic regression was used to develop a clinical model, imaging model and clinical-imaging model (Table 3). We enrolled the PFD, PMTD, temperature and AST as predictors for the clinical model, Log₂^Φ as predictor for the imaging model and the clinical-imaging model enrolled the clinical makers and Log₂^Φ as predictors in a logistic regression analysis. Variables were presented in the model by using the following formula:

Clinical model:

Probability of RMPP*100= -2272+54.7T+2.6AST+14.9PFD+34.2PMTD

Imaging model:

Probability of RMPP*100= -1315+85.3Log₂^Φ

Clinical-imaging model:

Probability of RMPP*100= -4709+142.1Log₂^Φ+1.7AST+8.5PFD+57.1PMTD

	Variable	β	OR	95%CI for OR
Clinical model	T	0.547	1.728	1.203, 2.483
	AST	0.026	1.026	0.987, 1.066
	PFD	0.149	1.161	0.997, 1.351
	PMTD	0.342	1.408	1.179, 1.680
Imaging-model	Log₂^Φ	0.853	2.347	1.642, 3.532
Clinical-imaging model	Log₂^Φ	1.421	4.141	2.251, 7.621
	T	0.578	1.782	1.164, 2.728
	AST	0.017	1.018	0.972, 1.067
	PFD	0.085	1.089	0.916, 1.294
	PMTD	0.571	1.768	1.385, 2.258

Table 3. Multiple models with a combination of clinical and imaging parameters by stepwise logistic regression analysis predicting the RMPP. PFD preadmission fever duration, PMTD preadmission macrolides therapy duration, OR odds ratio.

The ROC of the training set was plotted in Fig. 2A with AUROC of 0.810, 0.764 and 0.897, and the test set was plotted in Fig. 2B with area under the receiver operating characteristic curve (AUROC) of 0.782, 0.769 and 0.895 respectively. The AUC (area under curve) value of the Clinical-imaging model (0.897, 95% CI:0.835,0.957) was significantly higher than the Clinical model (0.810, 95% CI:0.734,0.887, P=0.012) and the imaging model (0.764, 95% CI:0.675,0.854, P<0.000). Independent verification conducted in the validation cohort also showed improved performance with clinical-imaging markers being 0.895(95% CI:0.789,1.000).

A nomogram was developed by incorporating the predictors in the clinical-imaging model (Fig.3). The nomogram was created by giving each independent influencing element a weighted score. The highest score is 120 points, and the range of RMPP incidence is 0.1 to 0.9. A higher chance of occurrence is indicated by a larger score obtained by adding the distribution points of each high-risk factor in the nomogram.

Model evaluation. Fig.4A and Fig.4B present graphical representations of calibration of the training set and the test set separately, the solid line represents the performance of the clinical-imaging model, of which a closer fit to the diagonal dotted line represents a better prediction.

We subsequently compared the clinical performance of the integrated model to the clinical model and imaging model separately using decision curve analysis. Fig.5A and Fig.5B illustrate the decision curves for the three models to predict the correct diagnosis of RMPP in the training set and test set. The decision curve analysis graphically shows the clinical usefulness of each model based on a continuum of potential thresholds for diagnosis of RMPP risk (x-axis) and the net benefit of using the model to risk stratify patients (y-axis) relative to assuming CMPP. In this analysis, the clinical-imaging model provided a larger net benefit across the range of diagnosis of RMPP compared with both the clinical model.

Mycoplasma pneumoniae (MP) is a serious pathogen that causes 10% to 30% of community-acquired pneumonia (CAP) in children under the age of 18 ^16,17. Although macrolide antibiotics are useful in treating MP infection, RMPP is becoming more common as a result of macrolide usage and the emergence of antibiotic-resistant bacteria ^18,19. RMPP generally causes a long-lasting fever that can progress to severe, life-threatening pneumonia ²⁰. Therefore, the value of markers in RMPP diagnosis is becoming more important.

To our knowledge, this is the first cohort study to predict RMPP in patients with MPP using noninvasive quantitative CT measurements. We defined a constant Φ that was the summation of CT value of per unit volume of lesion quantified by AI automatically. We found that Φ combined with clinical biomarkers including temperature, AST, PMTD, and PFD could predict the risk of MPP patients progressing to RMPP. The clinical-imaging model shows good accuracy and discrimination, which indicates it may be a convenient tool to help physicians recognize RMPP earlier.

Up to now, RMPP has been predicted using a variety of clinical and laboratory data^21-23, but the more visualized characters from CT scans have hardly been studied as potential markers for RMPP. Chest radiographs and CT scans can be particularly helpful in determining regional density distribution as well as the extent and form of lung injury. Large pulmonary lesions can be seen on radiographs or CT scans in both the general and RMPP groups. The pathological abnormalities of mycoplasma pneumonia in children can be seen on a CT scan, and the symptoms can help doctors diagnose patients more accurately²⁴. It has been reported that Clinical worsening or mortality in patients with COVID-19 pneumonia can be predicted using a quantitative assessment of the radiological extent and severity of lung illness utilizing an autonomous 3D AI-based software ²⁵. However, the quantitative assessment of CT features of MPP by AI software has not been reported extensively. In this study, we analyzed the CT scans of 180 MPP patients using the AI Intelligent Assistant Analysis System. This software can precisely segment different anatomical features of the lungs in order to discover infection regions and calculate the proportion of infection in both lungs, as well as their lobes and segments.

In this study, we compared the data of CT scans based on the AI software between RMPP and CMPP. Results showed that RMPP had a higher volume of lesions and Φ. The RMPP group's radiologic findings of lobar pneumonia and large pleural effusion point to severe disease brought on by macrolide resistance, a greater MP burden, harsh host reactions, or other elements involved in the refractory response ^7,10. We identified Log₂^Φ, a constant that combined the volume and CT value of each patient's lung lesion, as a prognostic factor of the imaging model.

We identified four clinical prognostic factors, including the highest temperature on the first day of admission, AST, PMTD and PFD, in order to forecast RMPP and limit the frequency of complications and long-term lung damage. Although the etiology of RMPP is unknown, it is largely assumed that inflammatory and immunological responses have a role in the progression of the disease. Some biomarkers linked to inflammatory reactions had been used to predict RMPP by previous studies ^21,26. The peak temperature in the RMPP group was significantly higher than that in the CMPP group on the first day of admission ²⁷. These markers were higher in the RMPP group than in the CMPP group, implying that children with refractory M. pneumoniae pneumonia have more pronounced inflammatory responses. After an MP infection, hepatic impairment is common extrapulmonary damage ²¹.The study showed that the AST in the RMPP group was higher ²⁷. Macrolides have long been thought to be the best treatment for M. pneumoniae pneumonia (MP) in children ²⁸. Based on physicians ' clinical experience, most patients had already received macrolides therapy before admission, but they are still hospitalized because of poor response to macrolides. In this study, we found that the longer macrolides therapy duration before admission they took, the greater the risk of RMPP.

Finally, we integrated the Log₂^Φ and clinical predictive factors to obtain a better comprehensive imaging-clinical prediction model. The sensitivity and specificity of the training set were 0.809 and 0.864, the AUC was 0.897 and the sensitivity and specificity of the test set were 0.892 and 0.875, the AUC was 0.895. It can be seen that the performance of the integrated clinical-imaging prediction model is significantly better than that of the simple clinical or imaging label.

The present study had several limitations. The model has not been externally validated, and multi-center validation with a bigger sample size is required to achieve high-level evidence for clinical use. This method has a high level of accuracy, but it also has a lot of individual variances, takes a lot of time, and is inefficient. Finally, because the data for the study was only taken from the baseline CT and one single-center CT, more research is needed to assess the possible benefit of AI software in MPP patient care.

In conclusion, this study presented an imaging-clinical prediction model incorporates imaging and clinical risk indicators and may be utilized to quickly identify RMPP and CMPP, laying the groundwork for early clinical and accurate therapy.

Acknowledgments: We thank the patients and their families for their participation and contributions to this study. We would like to thank Zhen Zhou from Shenrui (Beijing, China) participated in this investigation by providing AI technology assistant.

Author Contributions: Conceptualization, X.H.G. and H.M.; methodology, X.C.G.; software, Y.Q.; validation, Y.T.; formal analysis, L.W.; investigation, F.L. and C.Z.; resources, S.X.; data curation, Y.Q.; writing—original draft preparation, Y.T.; writing—review and editing, All authors. visualization, X.H.G. supervision, X.C.G.; project administration, X.H.G. All authors have read and agreed to the published version of the manuscript.

Competing interests: The authors declare that they have no competing interests.

Funding: This work was supported by the Project of Nanjing Science and technology development Foundation (YKK21150) to Xuhua Ge.

Informed Consent Statement: Informed consents were obtained from all the guardians of the pediatric patients.

Ethics approval and consent to participate: The study was authorized by the Children's Hospital of Nanjing Medical University’s institutional ethics committee and was registered in the Chinese Clinical Trial Registry. All procedures were carried out in line with the Helsinki Declaration.

Consent for publication: Not applicable.

Availability of data and materials: The datasets used and/or analyzed during the current study can be obtained from the corresponding author on reasonable request.

Korppi, M., Heiskanen-Kosma, T. & Kleemola, M. Incidence of community-acquired pneumonia in children caused by Mycoplasma pneumoniae: serological results of a prospective, population-based study in primary health care. Respirology 9, 109-114. https://doi.org:10.1111/j.1440-1843.2003.00522.x (2004).
Poddighe & Dimitri. Extra-pulmonary diseases related to Mycoplasma pneumoniae in children: recent insights into the pathogenesis. Current Opinion in Rheumatology, 1. https://doi.org:10.1097/BOR.0000000000000494(2018).
Cheng, Q. et al. Clinical features and risk factors analysis of bronchitis obliterans due to refractory Mycoplasma pneumoniae pneumonia in children: a nomogram prediction model. BMC Infect Dis 21, 1085. https://doi.org:10.1186/s12879-021-06783-4 (2021).
[Guidelines for management of community acquired pneumonia in children (the revised edition of 2013) (I)]. Zhonghua er ke za zhi = Chinese journal of pediatrics. 2013;51(10):745–52.
Liu, L., Maharjan, S., Sun, J. L., Li, Y. C. & Cheng, H. J. Prevalence and clinical characteristics of septicemia in children with Mycoplasma pneumoniae pneumonia. J Int Med Res 49, 3000605211021733. https://doi.org:10.1177/03000605211021733(2021).
Yuanyuan et al. Cytokines as the good predictors of refractory Mycoplasma pneumoniae pneumonia in school-aged children. Scientific Reports 11, 6:37037. https://doi.org: 10.1038/srep37037(2016).
Zhang, Y. et al. The Clinical Characteristics and Predictors of Refractory Mycoplasma pneumoniae Pneumonia in Children. PLoS One 11, e0156465. https://doi.org:10.1371/journal.pone.0156465 (2016).
Wen, J., Su, Y., Sun, H., Zhang, H. & Li, H. The combination of initial markers to predict refractory Mycoplasma pneumoniae pneumonia in Chinese children: a case control study. Respir Res 22, 89 (2021). https://doi.org:10.1186/s12931-020-01577-9.
Huang, X. et al. Clinical significance of D-dimer levels in refractory Mycoplasma pneumoniae pneumonia. BMC Infect Dis 21, 14. https://doi.org:10.1186/s12879-020-05700-5(2021).
Choi, Y. J. et al. Clinical Characteristics of Macrolide-Refractory Mycoplasma pneumoniae Pneumonia in Korean Children: A Multicenter Retrospective Study. J Clin Med 11. https://doi.org:10.3390/jcm11020306 (2022).
Wang, M. et al. Clinical and laboratory profiles of refractory Mycoplasma pneumoniae pneumonia in children. Int J Infect Dis 29, 18-23. https://doi.org:10.1016/j.ijid.2014.07.020 (2014).
Liu, F. et al. CT quantification of pneumonia lesions in early days predicts progression to severe illness in a cohort of COVID-19 patients. Theranostics 10, 5613-5622. https://doi.org:10.7150/thno.45985(2020).
Tamura, A. et al. Methylprednisolone pulse therapy for refractory Mycoplasma pneumoniae pneumonia in children. J Infect 57, 223-228. https://doi.org:10.1016/j.jinf.2008.06.012 (2008).
Li, H., Li, T., Cai, Q., Wang, X. & Zhou, Q. Development and Validation of a Radiomics Nomogram for Differentiating Mycoplasma Pneumonia and Bacterial Pneumonia. Diagnostics 11, 1330. https://doi.org:10.3390/diagnostics11081330 (2021).
Xing, W., Sun, H., Yan, C., Zhao, C. & Ma, J. A prediction model based on DNA methylation biomarkers and radiological characteristics for identifying malignant from benign pulmonary nodules. BMC Cancer 21. https://doi.org: 10.1186/s12885-021-08002-4 (2021).
Chiang, W. C. et al. Epidemiology, clinical characteristics and antimicrobial resistance patterns of community-acquired pneumonia in 1702 hospitalized children in Singapore. Respirology 12, 254-261. https://doi.org:10.1111/j.1440-1843.2006.01036.x(2007).
Spadaro, S. et al. Biomarkers for Acute Respiratory Distress syndrome and prospects for personalised medicine. J Inflamm (Lond) 16, 1. https://doi.org:10.1186/s12950-018-0202-y(2019).
Bi, Y. et al. Development of a scale for early prediction of refractory Mycoplasma pneumoniae pneumonia in hospitalized children. Sci Rep 11, 6595. https://doi.org:10.1038/s41598-021-86086-5(2021).
Okada, T. et al. Rapid effectiveness of minocycline or doxycycline against macrolide-resistant Mycoplasma pneumoniae infection in a 2011 outbreak among Japanese children. Clin Infect Dis 55, 1642-1649. https://doi.org:10.1093/cid/cis784(2012).
Zhou, Y., Hu, M., Ye, B., Chen, Z. & Zhang, Y. Early prediction of necrotizing pneumonia from mycoplasma pneumoniae pneumonia with large pulmonary lesions in children. Sci Rep 10, 19061. https://doi.org:10.1038/s41598-020-76083-5 (2020).
Bi, Y. et al. Risk of Mycoplasma pneumoniae-related hepatitis in MP pneumonia pediatric patients: a predictive model construction and assessment. BMC Pediatr 21, 287. https://doi.org:10.1186/s12887-021-02732-x (2021).
Lu, A., Wang, C., Zhang, X., Wang, L. & Qian, L. Lactate Dehydrogenase as a Biomarker for Prediction of Refractory Mycoplasma pneumoniae Pneumonia in Children. Respiratory Care, 1469. https://doi.org:10.4187/respcare.03920 (2015).
Sauteur, Meyer, P. M., Krautter, Selina & Lilliam, et al. Improved Diagnostics Help to Identify Clinical Features and Biomarkers That Predict Mycoplasma pneumoniae Community-acquired Pneumonia in Children. Clin Infect Dis, 71,1645-1654. https://doi.org:10.1093/cid/ciz1059(2020).
Fu, X., Yang, N. & Ji, J. Application of CT images based on the optimal atlas segmentation algorithm in the clinical diagnosis of Mycoplasma Pneumoniae Pneumonia in Children. Pak J Med Sci 37, 1647-1651. https://doi.org:10.12669/pjms.37.6-WIT.4860(2021).
Chabi, M. L. et al. Automated AI-Driven CT Quantification of Lung Disease Predicts Adverse Outcomes in Patients Hospitalized for COVID-19 Pneumonia. Diagnostics (Basel) 11. https://doi.org:10.3390/diagnostics11050878(2021).
Gong, H., Sun, B., Chen, Y. & Chen, H. The risk factors of children acquiring refractory mycoplasma pneumoniae pneumonia: A meta-analysis. Medicine (Baltimore) 100, e24894. https://doi.org:10.1097/MD.0000000000024894(2021).
Xu, J. J. & Shu, L. H. [Clinical characteristics of refractory Mycoplasma pneumoniae pneumonia in children]. Zhongguo Dang Dai Er Ke Za Zhi 20, 37-42. https://doi.org:10.7499/j.issn.1008-8830.2018.01.008(2018).
Waites, K. B. & Talkington, D. F. Mycoplasma pneumoniae and Its Role as a Human Pathogen. Clinical Microbiology Reviews 17, 697-728. https://doi.org:10.1128/CMR.17.4.697-728.2004(2004).

No competing interests reported.

Model based on the automated AI-Driven CT quantification is effective for the diagnosis of refractory Mycoplasma pneumoniae pneumonia

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Methods

Results

Discussion

Declarations

References

Additional Declarations

Status:

Journal Publication

Version 1