Construction and validation of neonatal asphyxia risk prediction model based on machine learning

Construction and validation of neonatal asphyxia risk prediction model based on machine learning | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Construction and validation of neonatal asphyxia risk prediction model based on machine learning Sisi Yi, Qiufang LI, Xieli SHI, Yanping TENG, xinfen xu This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7660175/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 4 You are reading this latest preprint version Abstract Background: Neonatal asphyxia is a leading cause of neonatal mortality and long-term neurodevelopmental impairment, particularly in low-resource settings, underscoring the need for improved predictive tools. Methods: This study developed a machine learning (ML) model using multi-source clinical data from 250 neonates (including mild/moderate asphyxia cases) at Zhejiang University Women’s Hospital. After meta-analysis identified 11 key risk factors, data preprocessing involved Random Forest imputation, standardization, and SMOTE for class balancing. Six ML algorithms (XGBoost, RF, Bagging, SVM, MLP, TabPFN) were evaluated, with SHAP analysis for interpretability. Results: XGBoost demonstrated superior performance (recall=0.82, precision=0.82, F1-score=0.82), with nuchal cord, assisted delivery, and prolonged labor emerging as top predictors. Ensemble methods (RF, Bagging) followed, while traditional models (SVM, MLP) and TabPFN showed lower efficacy. Conclusions: This study presents a validated ML framework for neonatal asphyxia prediction, offering clinical utility for early risk stratification and informed decision-making, particularly in resource-constrained environments. Neonatal asphyxia Machine learning (ML) Risk prediction Multi-source data integration Meta-analysis Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 1. Introduction Neonatal asphyxia remains one of the principal causes of neonatal mortality and long-term neurodevelopmental disabilities worldwide. According to World Health Organization (WHO) statistics, approximately 1 million newborns die from asphyxia annually, with millions more facing long-term health complications [ 1 ]. Developing countries bear a disproportionately higher incidence and mortality rate, imposing sub-stantial burdens on public health systems. Studies reveal that over one-third of neonatal deaths in urban India are attributable to asphyxia [ 2 ]. In China, this condition ranks as the third leading cause of mortality in children under five years old, following prematurity and neonatal pneumonia, accounting for 30%-35% of neonatal deaths. Annually, approximately 160,000 children under five succumb to this condition [ 3 ]. Fundamentally characterized as a hypoxic disorder, neonatal asphyxia arises from multiple antenatal, intrapartum, and postnatal factors [ 4 ]. It manifests as absent or depressed respiration at birth, subsequently leading to multi-organ dysfunction. With high fatality and disability rates, it constitutes a major contributor to neonatal mortal-ity, cerebral palsy, and intellectual disabilities [ 5 ]. This condition not only severely im-pacts infant health but also imposes heavy socioeconomic burdens on families and so-ciety, requiring substantial resource allocation for long-term medical care, rehabilita-tion, and special education [ 6 ]. The Apgar score, as a widely adopted clinical neonatal assessment tool, holds significant clinical value. Through its simple and rapid compo-site indicators, it enables preliminary evaluation of neonatal physiological status within minutes after birth, providing crucial guidance for clinical decision-making [ 7 ]. However, Apgar scoring outcomes are influenced by multiple variables including ges-tational age, birth weight, maternal medication use, pharmacological/anaesthetic interventions, and congenital anomalies [ 8 ]. Furthermore, certain components of the scoring system involve subjective interpretation, potentially leading to inter-rater variability. Consequently, the Apgar score is primarily utilized to assess infant respon-siveness to resuscitation measures rather than predicting long-term health out-comes [ 9 ]. Particularly, the 1-minute score demonstrates limited indicative value for long-term clinical prognosis, highlighting the need for a reliable real-time method to derive this assessment [ 10 ]. Although research on risk factors for neonatal asphyxia exists domestically and internationally, clinically effective risk prediction models remain scarce. Developing such models could help hospitals rationally allocate nursing resources, enable caregiv-ers to adjust interventions promptly, reduce the incidence of neonatal asphyxia, and safeguard infant survival. Accurate disease assessment and asphyxia risk prediction would facilitate personalized patient management, optimize human resource alloca-tion, enhance nursing efficiency, and ultimately save more lives. Given the critical and rapidly evolving nature of neonatal conditions, an ideal assessment tool would assist healthcare providers in promptly verifying treatment efficacy and modifying care strategies. Therefore, neonatal asphyxia risk prediction holds vital significance for im-proving patient prognoses. Current neonatal asphyxia risk prediction models predominantly employ traditional statistical methods (such as Logistic regression and Cox proportional hazards regression). While these approaches offer simplicity and interpretability for binary classification tasks, they exhibit notable limitations. These models inherently assume linear relationships between independent and dependent variables, failing to effec-tively address complex nonlinear interactions. Furthermore, their performance is sen-sitive to data distribution characteristics, with model stability and predictive capabil-ity significantly compromised by multicollinearity or outliers [ 11 ]. In contrast, ML models demonstrate substantial optimization potential through nonlinear transformations, feature engineering refinement, and sophisticated algorithmic architectures. These capabilities enable superior capture of complex data patterns and nonlinear re-lationships, thereby markedly enhancing prediction accuracy and generalization ca-pacity. Such advancements provide more efficient and precise solutions for intricate classification challenges [ 12 ]. In summary, improving neonatal asphyxia outcomes represents a critical priority for hospital administrators. Developing accessible and accurate risk assessment tools holds significant clinical value by assisting nursing staff in clinical evaluations, informed decision-making, treatment efficacy verification, and rational allocation of medical resources, ultimately reducing neonatal asphyxia inci-dence. The pursuit of clinically applicable and precise risk prediction tools is para-mount for enhancing caregivers predictive capabilities, enabling early implementation of targeted interventions, and improving neonatal outcomes. This study focuses on collecting relevant datasets and developing a model to predict neonatal asphyxia risk, while validating its performance in accuracy, reliability, and practical effectiveness. Key innovations include multi-source data integration, in-corporating clinical data, biomarkers, and socioeconomic factors to construct a more comprehensive risk prediction framework. Data is sourced from a tertiary maternal and child health hospital, with empirical validation conducted to assess the model’s applicability and precision within specific clinical settings and populations. Addition-ally, the model incorporates automated prediction capabilities to enhance efficiency. The outcomes of this research will provide healthcare professionals with a practical tool to support more precise diagnostic and intervention decisions during prenatal and intrapartum stages. By enabling early identification of high-risk neonates, the model aims to optimize medical resource allocation and improve the efficiency and effec-tiveness of healthcare systems. Furthermore, this study will contribute a scientific foundation for shaping public health policies, advancing perinatal care services, and refining preventive strategies. Through analyzing the relative contributions of diverse factors to neonatal asphyxia risk, policymakers will be empowered to design targeted preventive measures and intervention plans, ultimately safeguarding neonatal health outcomes. 2. Related works The Apgar score serves as a physiological assessment tool based on five clinical indicators—skin color, heart rate, respiratory rhythm, muscle tone, and reflex irritability (each scored 0–2 points, totaling 10 points) [ 13 ]. This scoring system provides peri-natal medical teams with objective criteria for evaluating intrauterine adaptation sta-tus (1-minute score) and dynamically monitoring resuscitation effectiveness (5/10-minute scores) through rapid, standardized multi-dimensional analysis. While its timeliness and operational practicality hold significant clinical value, it is critical to acknowledge its limitations as a screening tool for asphyxia. The Chinese diagnostic consensus [ 14 ] for neonatal asphyxia emphasizes that low Apgar scores must be inter-preted alongside auxiliary methods such as arterial blood gas analysis, neurological assessments, and multi-organ injury evaluations to ensure comprehensive diagnosis and mitigate misjudgment risks arising from over reliance on a single scoring metric. In recent years, researchers have explored risk prediction models for neonatal asphyxia through diverse methodologies. A study by Zhang Youjun [ 15 ] employed binary logistic regression to identify nine clinical risk factors associated with neonatal as-phyxia, including preterm birth, abnormal fetal position, intrapartum fever, umbilical cord anomalies, placental abnormalities, fetal distress, regular prenatal checkups, pro-longed labor, and amniotic fluid abnormalities. Internal validation via 1,000 iterations of Bootstrap resampling yielded a final nomogram prediction model with a concord-ance index (C-index) of 0.810, indicating strong predictive performance. Calibration curves generated using R Studio further demonstrated robust alignment between pre-dicted and observed outcomes. Additionally, an Iranian cross-sectional study [ 16 ] ap-plied multiple ML algorithms—logistic regression, decision tree (DT) classifier, random forest (RF) classifier, XGBoost (Extreme Gradient Boosting) classifier, permutation classifier, feedforward deep learning, LightGBM (Light Gradient Boosting Machine), and support vector machine (SVM)—to predict birth asphyxia. The findings high-lighted RF classification as the most accurate algorithm for this purpose. These studies collectively advance novel approaches for early warning and intervention in neonatal asphyxia. 3. Methods One primary objective of this study is to predict the incidence of neonatal asphyxia using practical ML methods and tools. To achieve this goal, six ML algorithms were employed: TabPFN, Bootstrap Aggregating (Bagging), Multilayer Perceptron (MLP), SVM, XGBoost, and RF. The research workflow comprised four main phases: preprocessing, standardization, evaluation, and modeling. For clarity, Fig. 1 illustrates the complete sequence of stages involved in this investigation. All ML models and techniques were developed and implemented using Python 3.8. Computational tasks were executed on a Windows operating system with an Intel(R) Core(TM) i7-10750H CPU @ 2.60 GHz (2592 MHz, 6 cores, 12 logical processors). 3.1. Data acquisition The data for this study were extracted from the neonatal electronic medical record system of Women's Hospital, School of Medicine, Zhejiang University. Structured Query Language (SQL) was utilized to retrieve complete case data meeting research criteria from January to December 2024. Adopting a retrospective cohort design, we constructed an analytical dataset comprising 40 clinical features—including gesta-tional age, delivery mode, and umbilical cord abnormalities—after feature engineering (feature selection, categorical variable transformation, and missing value imputation). A total of 250 neonates with birth asphyxia were enrolled, consisting of 187 mild asphyxia cases (74.8%) and 63 moderate asphyxia cases (25.2%). The cohort included 146 males (58.4%) and 103 females (41.2%), with a median birth weight of 3150g (IQR: 2850–3450g) and a median gestational age of 38 + 3 weeks (range: 32 + 1–41 + 5 weeks). All cases met the following inclusion criteria: (1) singleton live birth at the study hospital; (2) complete perinatal maternal-neonatal clinical records; (3) definitive Apgar scores and asphyxia diagnosis. Exclusion criteria comprised: (1) intrauterine death; (2) major congenital anomalies (e.g., congenital 3heart disease); (3) multiple gestation; (4) miss-ing critical clinical data. Data collection strictly adhered to clinical practice guidelines, with Apgar scoring and diagnostic assessments performed by attending neonatologists or higher-ranking clinicians. Study variables spanned three dimensions: neonatal bio-logical characteristics (sex, birth weight, gestational age), delivery-related parameters (labor duration, amniotic fluid characteristics, umbilical cord abnormalities), and ma-ternal pregnancy features (gestational complications, parity, delivery mode). As sys-tematically detailed in Table 1 , obstetric-related parameters constituted 62.5% (25/40) of variables, neonatal assessment metrics accounted for 27.5% (11/40), and maternal baseline characteristics comprised 10.0% (4/40). Table 1 Comprehensive overview of the features. SN Attribute name Description Type 1 Maternal Age < 20 = 0, 6months gestation, of either a live birth or stillbirth Numerical 4 Mother's job If exists 1, else 0 Nominal 5 residence Nominal 6 BMI2 Body mass index in 2nd trimester18.5-24.9 = 1, 25-29.9 = 2, >30 = 3 Nominal 7 Conception method natural conception = 1, else 0 Nominal 8 Sex Gender of the fetus, male = 1, female = 0 Nominal 9 singleton If exists 1, else 0 Nominal 10 adverse pregnancy history If exists 1, else 0 Nominal 11 Miscarriage If exists 1, else 0 Nominal 12 Gestational Age 28 ~ 31 + 6w = 0, 32 ~ 33 + 6w = 1, 34 ~ 36 + 6w = 2, 37 ~ 40 + = 3 Nominal 13 Birth Weight 4000g = 4 Nominal 14 Type of Delivery Natural childbirth = 1, C-section = 2, Forceps-assisted delivery = 3 Nominal 15 Abnormal fetal heart rate If exists 1, else 0 Nominal 16 Labor Process prolonged labor = 2, precipitate labor = 3 Nominal 17 Cord around the neck (CAN) Umbilical Cord Entanglement If exists 1, else 0 Nominal 18 Abnormal umbilical artery Doppler S/D ratio If exists 1, else 0 Nominal 19 Fetal growth restriction (FGR) If exists 1, else 0 Nominal 20 Abnormal Fetal Position If exists 1, else 0 Nominal 21 Placenta previa If exists 1, else 0 Nominal 22 Placental adhesion If exists 1, else 0 Nominal 23 Placental implantation If exists 1, else 0 Nominal 24 Placental abruption If exists 1, else 0 Nominal 25 Fetal Hypoxia If exists 1,else 0 Nominal 26 Premature rupture of membranes (PROM) If exists 1, else 0 Nominal 27 Amniotic Fluid Contamination If exists 1, else 0 Nominal 28 Amniotic fluid index (AFI) oligohydramnios = 1, normal amniotic fluid volume = 2, polyhydramnios = 3 Nominal 29 Intrauterine infection If exists 1, else 0 Nominal 30 Uterine rupture If exists 1, else 0 Nominal 31 Pelvic/abdominal adhesions If exists 1, else 0 Nominal 32 Gestational hypertension If exists 1, else 0 Nominal 33 Gestational diabetes If exists 1, else 0 Nominal 34 Preeclampsia If exists 1,else 0 Nominal 35 Moderate preeclampsia If exists 1, else 0 Nominal 36 Pregnancy complicated by thyroid disease If exists 1, else 0 Nominal 37 edication taken in early pregnancy If exists 1, else 0 Nominal 38 Anemia in Pregnancy If exists 1, else 0 Nominal 39 Co-existing infectious disease If exists 1, else 0 Nominal 40 Apgar Score 4,5,6,7 Nominal 3.2. Feature extraction Neonatal asphyxia is a severe and common perinatal complication with a complex pathogenesis involving multiple potential risk factors. These factors may encom-pass maternal health during pregnancy, intrapartum conditions, and neonatal physio-logical characteristics. To deeply understand their impact on neonatal asphyxia, the extraction of critical feature variables is essential. Feature extraction enables the iden-tification of core variables strongly associated with neonatal asphyxia from vast clini-cal datasets, thereby enabling more precise identification of high-risk populations and providing actionable insights for targeted interventions. The meta-analytic approach, as a systematic research methodology, offers distinct advantages. First, it integrates data from numerous studies, amplifies sample sizes, and enhances statistical power, allowing more accurate estimation of the association strength between risk factors and neonatal asphyxia. Second, meta-analysis synthesizes and compares findings across studies, assesses heterogeneity among them, and reveals the stability of potential risk factors across diverse populations, regions, and study conditions. Furthermore, it compensates for limitations inherent in individual studies—such as small sample sizes or restricted designs—by delivering comprehensive and objective conclusions through pooled analysis. By employing meta-analysis, this study aims to clarify key determi-nants of neonatal asphyxia. Such insights will empower clinicians to proactively iden-tify high-risk neonatesduring perinatal management, implement tailored preventive strategies, and optimize perinatal care protocols. Ultimately, this approach seeks to reduce the incidence of neonatal asphyxia and improve neonatal health outcomes. 3.2.1. Retrieval strategy A computer-based search was conducted in the following databases: China National Knowledge Infrastructure (CNKI), Wanfang, VIP, Chinese Biomedical Literature Database (CBM), PubMed, Embase, Web of Science, and Cochrane Library. The search was restricted to Chinese and English languages, with a timeframe spanning from the inception of each database to June 18, 2024. Search terms combined controlled vocab-ulary (subject headings) and free-text keywords. Chinese search words: "neonatal as-phyxia/asphyxia/perinatal asphyxia"; "Risk factors/risk factors/influencing fac-tors/related factors/predictors/causes/investigations". English search terms: “risk fac-tor /relevant factors/predictor/associate factors/influence”;“neonatal asphyxia / as-phyxia of the newborn / a.neonatorum / apnea neonatorum / neonate asphyxia / neo-nates asphyxia / newborns asphyxia/'apnoea neonatorum/ newborn suffocation/ newborn hypoxia/'newborn hypoxia”. The retrieval strategy is adjusted appropri-ately according to the requirements of each database. 3.2.2. Literature screening and data extraction In this study, researchers independently conducted literature screening and data extraction, followed by cross-verification to ensure the quality and relevance of selected literature. For resolving discrepancies in contentious cases, a third re-searcher was introduced to arbitrate, ensuring impartiality and objectivity in the re-view process. Additionally, Endnote software was employed to systematically manage retrieved literature, facilitating subsequent analysis and integration. The process of literature screening is divided into two important steps: preliminary screening and full-text reading. In the initial screening stage, researchers focused on reviewing the ti-tles and abstracts of the articles to determine whether they met the inclusion criteria. The next steps involved a thorough full-text reading of the eligible literature, thus fur-ther confirming its suitability. The final extracted literature information covered mul-tiple dimensions, including author, publication year, study country, study object, study type, diagnostic criteria, sample size and related risk factors. 3.2.3. Literature quality evaluation This study was independently conducted by two professionally trained researchers to ensure scientific rigor throughout the research process. For literature quality as-sessment, we employed risk-of-bias evaluation tools to systematically analyze and as-sess the reliability and validity of included studies. Cohort and case-control studies were evaluated using the Newcastle-Ottawa Scale (NOS), a validated quality assess-ment instrument that provides detailed scoring across three domains. The NOS is di-vided into two main parts, respectively for the evaluation of cohort and case-control studies, and the overall is composed of three parts. Specifically, the object selection consisted of four items, each of which was awarded one point; The inter-group com-parability component consists of one item, with a maximum score of two points; The evaluation section for the outcome or exposure factor contains three items, again scoring one point for each item. The scale has a total score of nine points, and the scor-ing criteria is: the research literature is divided into different quality levels, a score of 0 to 3 is classified as low quality research, a score of 4 to 6 is medium quality research, while a score of 7 to 9 is regarded as high quality research. In addition, the literature Quality evaluation of cross-sectional studies was carried out according to the evalua-tion criteria recommended by the Agency for Healthcare Research and Quality (AHRQ). The criterion contains 11 assessment items with response options of "yes", "no" or "not clear" on an 11-point scale. Under this scoring system, a score of 0 to 3, a score of 4 to 7, and a score of 8 and above correspond to low, medium and high quality literature, respectively. Figure 2 illustrates the Screening process, which is divided into three main stages: Identification, screening, and Included. 3.2.4. Statistical methods Meta-analysis was conducted using Stata 18.0 software. For continuous variables, weighted mean difference (WMD) and its 95% confidence interval (95% CI) were calculated, while odds ratios (ORs) with 95% CIs were applied for dichotomous variables. Heterogeneity was assessed using the I² statistic and Cochran’s Q-test. A fixed-effect model was employed when heterogeneity was deemed acceptable (P > 0.10 and I² < 50%), whereas a random-effects model was adopted for significant heterogeneity (P ≤ 0.10 or I² ≥ 50%). Descriptive analysis was performed for studies where effect sizes could not be pooled. In this study, the number of articles included and meta-merged was small, and less than 10 articles were analyzed for a single risk factor, so funnel plot analysis was not performed. In addition, this study set that when P < 0.05, it was considered statistically significant. A total of 8 studies involving 5,593 cases were in-cluded, analyzing 24 risk factors. The results of meta-analysis showed that 11 risk fac-tors were statistically significant. Among them, low birth weight infants (OR = 3.826), maternal age less than 20 years old (OR = 1.946), premature infants (OR = 2.609), assisted delivery (OR = 4.238), prolonged labor (OR = 3.880), non-cephalic presentation (OR = 4.861), intrauterine distress (OR = 5.412), amniotic fluid contamination (OR = 8) 53), primiparity (OR = 3.539), umbilical cord around the neck (OR = 3.830), and anemia dur-ing pregnancy (OR = 1.667) were risk factors for neonatal asphyxia. Sensitivity analysis showed that the combined results of fixed effect model and random effect model were basically stable. Table 2 presents the results of the heterogeneity analysis of different risk factors for neonatal asphyxia in the meta-analysis. Table 2 Standardized Mean Differences (95% Confidence Intervals) and Heterogeneity (P-value, I²) of the Association Between 14 Risk Factors and Neonatal Asphyxia from Random-Effects Me-ta-Analyses Heterogeneity Analysis Studies, No. β (95%CI) P value I2, % Low birth weight infant 4 3.826(2.765, 5.296) <0.001 36.0 Jumbo baby 2 0.700(0.338, 1.450) 0.337 0 The maternal age was less than 20 years 2 1.946(1.305, 2.902) 0.001 0 Premature infant 4 2.609(1.815, 3.752) <0.001 0 Caesarean section 4 0.784(0.485, 1.266) 0.319 0 assisted delivery 5 4.238(2.907, 6.180) <0.001 35.6 Prolonged labor 4 3.880(1.072, 14.051) 0.039 83.9 Abnormal fetal position 3 4.861(3.288, 7.188) <0.001 0 Intrauterine distress 3 5.412(3.578, 8.187) <0.001 0 Umbilical cord around neck 2 3.830(2.228, 6.583) <0.001 0 Amniotic fluid contamination 5 8.532(3.687, 19.741) <0.001 84.9 primiparity 3 3.539(2.202, 5.688) <0.001 4.9 Pregnancy-induced hypertension syndrome 2 2.155(0.709, 6.547) 0.176 0 Anemia of pregnancy 4 1.667(1.174, 2.368) 0.004 53.8 3.3. Distribution of Extracted Features This study analyzed clinical data from 250 neonates, encompassing 12 core clinical indicators, including gestational age, Apgar scores, and birth weight. Data analysis showed that only 2 cases of "Amniotic Fluid Contamination" had missing values, and 1 case was missing for each of the other 11 indicators. The total missing values of the whole data set accounted for 0.43% (13/3000), and the data completeness was 99.57%. This indicates that the dataset has reliable statistical integrity with applicability to clinical studies. The distribution of missing data and treatment effects are detailed in Table 3 . Figure 3 shows that the clinical core index, neonatal Apgar score, is concen-trated in the 4–7 region (accounting for 100% of the total sample), which meets the di-agnostic criteria of mild to moderate neonatal asphyxia defined by the International Academy of Pediatrics (AAP) (Apgar score 4–6 is moderate asphyxia, and 7 is the crit-ical value of mild asphyxia) [ 17 ]. The distribution characteristics of the score suggest that the study cohort accurately covers the risk group of perinatal mild to moderate asphyxation, and provides a clinically relevant research sample for exploring the pathophysiological characteristics and prognostic factors of neonates in this subgroup. It should be noted in particular that severe asphyxia (Apgar ≤ 3) cases were not includ-ed in this dataset. This design not only meets the ethical requirements of the study, but also ensures the homogeneity of the pathological degree of the analyzed subjects. Fig-ure 4 is a Correlation Heatmap that shows the correlation between different variables in a dataset. The feature correlation matrix shows that most of the features have weak linear correlations (absolute correlation coefficient < 0.3), which indicates that the fea-tures in the dataset are relatively independent and may reduce the effect of multicol-linearity on the model. Kernel density estimation (KDE) analysis of Fig. 5 revealed patterns of association between different perinatal characteristics and neonatal as-phyxia severity (stratified by Apgar score). Specifically, newborns with intrauterine distress, contaminated amniotic fluid, umbilical cord around the neck, and anemia during pregnancy are more likely to suffer from asphyxation. In addition, premature infants and newborns with abnormal fetal position have a higher proportion of as-phyxia. Table 3 The number of Missing Values in Dataset. Attributes No. of missing values Maternal Age 1 Apgar Score 1 Birth weight 1 Gestational Age 1 Type of Delivery 1 Labor Process 1 Abnormal Fetal Position 1 Fetal Hypoxia 1 Amniotic Fluid Contamination 2 Umbilical Cord Entanglement 1 Parity 1 Anemia in Pregnancy 1 4. Experimental Setup 4.1. Data preprocessing Data preprocessing serves as a critical phase in data analysis and ML, encompassing data cleaning, transformation, and integration to enhance analytical quality. Multiple techniques were applied, including outlier detection, missing value imputa-tion, categorical variable encoding, and data standardization, to optimize overall data integrity. For handling missing data, the RF algorithm was employed to impute missing values. To improve model effectiveness, normalization was performed using the StandardScaler function from the Sklearn library, which scales features to a standard-ized range (0–1). This ensures consistent scaling across variables, thereby mitigating bias from disparate measurement units and enhancing model performance. To address class imbalance, this study implemented the Synthetic Minority Over-sampling Technique (SMOTE). SMOTE strategically expands the decision boundary of the minority class by generating synthetic samples between minority class instances and their k-nearest neighbors within the feature space. 4.2. Model Introduction This study employed random sampling to partition the dataset into a training set (70%) and test set (30%), ensuring robust model evaluation accuracy and validity. ML algorithms were applied to both the original dataset and resampled datasets (post-class balancing). To enhance result robustness and reliability, each algorithm was validated across multiple randomly partitioned training-test splits. A consistent ran-dom seed was used for data shuffling to guarantee reproducibility. Through meticu-lous feature selection, the study aimed to identify the optimal feature subset that maximizes predictive performance, thereby deepening insights into inherent data pat-terns. ML classification techniques, as a core component of AI, enable precise data classification and prediction. In this research, we utilized diverse ML classifiers, including TabPFN, Bagging, Multilayer Perceptron (MLP), SVM, XGBoost, and RF. These classi-fiers were applied to 11 feature variables encompassing critical clinical indicators, with hyperparameter optimization tailored to each ML model. The subsequent subsec-tions elaborate on the fundamental principles of each ML classification method. 4.2.1. TabPFN TabPFN is a Transformer-based model specifically designed for classification tasks on small-scale tabular datasets. Trained offline by learning posterior predictive distributions over synthetic datasets, it defines a hypothesis space that captures relationships between input features and output l Following single-epoch training, TabPFN enables rapid prediction without requiring hyperparameter tuning. It pro-cesses training and testing samples simultaneously, generating predictions through a single forward pass, thereby achieving notable computational efficiency [ 18 ]. 4.2.2. SVM SVM is a widely-used ML algorithm primarily employed for classification and re-gression tasks. It operates by identifying an optimal hyperplane that maximizes the margin between classes in the feature space. To address nonlinear separability, SVM leverages the kernel trick, implicitly mapping data into higher-dimensional spaces. Additionally, it employs soft margins to accommodate noisy or overlapping data points. While SVM excels in applications such as image recognition and text classifica-tion, its computational inefficiency for large datasets and sensitivity to parameter se-lection (e.g., kernel type, regularization) pose practical limitations. These characteris-tics necessitate careful tuning and scalability considerations in real-world implemen-tations [ 19 ]. 4.2.3. XGBoost XGBoost is a highly efficient gradient-boosted tree algorithm that optimizes the tree-building process to enhance model performance and computational speed. It in-corporates regularization terms to control model complexity and mitigate overfitting, supports diverse loss functions for enhanced flexibility across tasks, and leverages parallelized processing to accelerate training. These features collectively enable XGBoost to demonstrate exceptional performance in various ML applications, partic-ularly in handling structured data and large-scale datasets [ 20 ]. 4.2.4. RF RF is an ensemble learning algorithm that enhances model accuracy and robustness by constructing multiple decision trees. In RF, each tree is trained on randomly selected subsets of features and samples (via bootstrapping). The final prediction is de-rived by aggregating outputs through majority voting (for classification) or averaging (for regression), effectively reducing model variance and improving resistance to over-fitting. As a versatile supervised learning algorithm, RF excels in both classification and regression tasks, efficiently handles high-dimensional data, and provides robust estimates of feature importance through metrics like Gini impurity reduction or mean decrease in accuracy [ 21 ]. 4.2.5. Bagging Bagging is a ML technique that enhances prediction accuracy by integrating outputs from multiple models. Its core principle involves constructing diverse models on distinct data subsets and consolidating their predictions to form a robust learner. As an ensemble method, Bagging employs random sampling with replacement to create mul-tiple data subsets, trains models independently on these subsets, and aggregates their predictions (e.g., majority voting for classification, averaging for regression) to achieve more stable outcomes. A key advantage lies in its ability to reduce model variance by averaging prediction errors across individual models while simultaneously simplifying computational complexity through parallelizable training processes [ 22 ]. 4.2.6. MLP MLP is a widely used artificial neural network in ML. As a supervised learning algorithm, it employs backpropagation to adjust network weights based on errors encountered during training. An MLP architecture comprises multiple layers: an input layer, one or more hidden layers, and an output layer. Neurons within each layer are interconnected via weighted connections, enabling information flow and hierarchical computation across the network [ 23 ]. 4.3. Parameter setting and evaluation metrics Table 4 presents the selected parameters and fully optimized hyperparameter configurations of the best-performing ML models across categories. To ensure experi-mental fairness and reproducibility, all models were evaluated under identical train-ing/testing splits and computational constraints. Establishing a robust evaluation framework is critical when assessing classifier performance. To enhance model inter-pretability, this study employed SHAP value analysis. SHAP values, rooted in coopera-tive game theory, quantify the contribution of each feature to model predictions by calculating its marginal contribution to the prediction outcome. This method eluci-dates the relationship between features and model outputs by decomposing predic-tions into additive contributions from individual features. The sum of all fea-ture-specific Shapley values equals the difference between the model’s predicted value and a baseline value (typically the dataset’s mean prediction) [ 24 ]. Table 5 presents a confusion matrix, where True Positives (TP) are the correctly identified positive instances, and True Negatives (TN) are the correctly identified negative instances. These values form the basis for calculating critical metrics such as accuracy, recall, precision, and F1-score. The confusion matrix serves as a cornerstone tool for evaluating classifier performance. It provides granular data on true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN), forming the basis for calculating critical metrics such as accuracy, recall, precision, and F1-score. These metrics are indispensable for assessing classification model effectiveness, as they comprehensively characterize classifier performance while revealing strengths and weaknesses to determine suitability for specific tasks [ 25 ]. Equations 1 to 4 show the core metric formulation used to evaluate the performance of the classification model. These metrics are the key evalua-tion metrics used in ML to evaluate the performance of classification models and in-clude Accuracy, Precision, Recall, and F1-score. Table 4 Hyperparameter results for top-performing models in each category for asphyxia admis-sion prediction across the complete case. SVM Value MLP Value RF Value C 1.0 alpha 1e-4 criterion gini kernel rbf hidden_layer_sizes 100 splitter best degree 3 activation relu n_estimators 100 gamma scale solver adam m auto class_weight balanced - - max_depth 10 tol 1e-3 - - min_samples_split 2 Bagging Value TabPFN Value XGBoost Value max_ features 1.0 N_ensemble_configs 32 criterion reg: squarederror n_ estimators 10 device cpu Max _depth 6 bootstrap True Inference _precision float32 Num _boost _round 100 - - Normalize _y True learning_rate 0.3 - - seed 0 subsample 1.0 - - device cpu colsample_bytree 1.0 5. Results and analysis Table 6 systematically compares the predictive performance metrics of six machine learning models, including core evaluation parameters such as recall, precision, and F1-score. The experimental results demonstrate significant performance differences among the models in neonatal asphyxia risk prediction. The XGBoost model exhibited optimal predictive performance (recall: 0.82, precision: 0.83, F1-score: 0.82), significantly outperforming ensemble methods (Bagging: F1 = 0.79; Random Forest: F1 = 0.75) and traditional models (SVM: F1 = 0.67; MLP: F1 = 0.66; TabPFN: F1 = 0.69). The improved recall of Random Forest (0.75 vs. 0.71 in previous Iranian studies) may stem from the inclusion of comprehensive placental pathology indicators (e.g., adhesion, implantation), while TabPFN's suboptimal performance (F1 = 0.69) confirms the limitations of Transformer architectures in small-scale clinical datasets (n = 250). The performance differences further validate XGBoost's superiority in handling nonlinear feature interactions (e.g., umbilical cord abnormalities combined with prolonged labor), whereas traditional statistical models (e.g., logistic regression) underperform due to their linearity assumptions. These findings provide a more precise risk stratification tool for clinical practice, particularly for optimizing medical resource allocation in resource-constrained settings through early identification of high-risk cases. Future research should address sample representativeness limitations through multicenter validation and develop more interpretable clinical decision support systems to facilitate model translation. This study systematically evaluated the predictive value of clinical features for neonatal asphyxia risk using SHAP analysis. Figure 6 (SHAP bar plot) identified umbilical cord entanglement (mean SHAP: +1.86) and amniotic fluid contamination (+ 1.50) as the strongest positive predictors, while Fig. 7 (multi-model scatter plot) revealed algorithmic interpretation differences, particularly XGBoost's superior predictive capability for cord entanglement (SHAP range: +2.0 to + 3.0). Key findings include: 1) Umbilical cord entanglement (OR = 3.830, 95%CI: 2.228–6.583), assisted delivery (OR = 4.238, 95%CI: 2.907–6.180), prolonged labor (OR = 3.880, 95%CI: 1.072–14.051), and amniotic fluid contamination (OR = 8.532, 95%CI: 3.687–19.741) emerged as critical risk predictors, with severe contamination (Grade III) reaching SHAP values > + 2.5; 2) Delivery mode consistently ranked as the most stable high-weight predictor (top-2 in all models), its bimodal SHAP distribution reflecting risk variations across delivery methods; 3) Apgar score showed limited predictive value (SHAP < 0.3); 4) Maternal age exhibited nonlinear effects - younger mothers (< 25 years) carried higher risk (SHAP: +0.8), while advanced maternal age (≥ 35) demonstrated neutral/protective effects (SHAP: -0.15 to 0), potentially reflecting differential prenatal monitoring intensity. These results suggest clinicians should prioritize high-risk feature combinations (e.g., cord entanglement with amniotic contamination) and enhance monitoring for primiparas (OR = 3.539, 95%CI: 2.202–5.688). The weaker explanatory power of TabPFN (SHAP range: ±1.5) further confirms Transformer's limitations in small clinical datasets. Table 5 Confusion Matrix. Predicted Negative Positive Actual Negative True Negative (TN) False Positive (FP) Positive False Negative (FN) True Positive (TP) $$\:Accuracy=\frac{TP+TN}{TP+TN+FP+FN}$$ 1 $$\:Precision=\frac{TP}{TP+FP}$$ 2 $$\:Recall=\frac{TP}{TP+FN}$$ 3 Table 6 The result of the six ML models with different categories and complete case. Classifier Model Recall Score Precision Score Accuracy Score F1 Score SVM 0.66 0.71 0.66 0.67 MLP 0.66 0.72 0.66 0.66 RF 0.75 0.76 0.75 0.75 Bagging 0.78 0.8 0.78 0.79 TabPFN 0.69 0.72 0.69 0.69 XGBoost 0.82 0.83 0.82 0.82 6. Discussion Early prediction of neonatal asphyxia plays a pivotal role in perinatal management, offering benefits such as optimized healthcare resource allocation, reduced demand for intensive care, and provision of critical risk information to families for psychological and logistical preparedness. This study established a predictive framework comprising two core components: data preprocessing and model evaluation. During the data preprocessing phase, data quality was enhanced through systematic steps, missing value imputation (using the RF algorithm), categorical variable encoding, and data standardization. Additionally, the SMOTE was applied to address class imbalance, thereby improving model accuracy and laying a robust foundation for subsequent training. In the model evaluation phase, six advanced classifiers—TabPFN, Bagging, MLP, SVM, XGBoost, and RF—were rigorously assessed for their predictive efficacy in neonatal asphyxia. The study integrated 11 evidence-based independent risk factors spanning critical prenatal to intrapartum variables, such as low birth weight, primi-parity, abnormal fetal position, labor dynamics, fetal monitoring indices, umbilical cord abnormalities, and maternal comorbidities. A randomized sampling strategy en-sured clinically representative distribution of subgroups in training and validation sets. During the model validation phase, while traditional ML models (e.g., XGBoost) demonstrated satisfactory predictive accuracy and reliability, their clinical application faces two primary challenges. The first is the limited interpretability of the model, alt-hough SHAP value analysis was employed to quantify feature contributions to predic-tions, discrepancies persist between certain SHAP values and clinical knowledge, which may undermine clinicians’ trust in the model. The second is the "black box" problem of deep learning methods, despite achieving high recall rates in our center’s test dataset with advanced deep learning methods (e.g., enhanced Transformer archi-tectures), their opaque feature extraction mechanisms fail to meet transparency re-quirements for clinical decision-making. Future research must address two critical translational medicine challenges: 1. Enhancing Model Interpretability: Through advanced SHAP value analysis or alterna-tive explainability tools to enable visualization and clinical validation of feature con-tributions, thereby strengthening clinical trust in model outputs. 2. Conducting Prospective Clinical Trials: To validate the model’s real-world impact on clinical end-points, ensuring its effectiveness and reliability in practical healthcare settings. As this study represents a single-center investigation with geographically concentrated data, future work requires multicenter validation to confirm the model’s generalizability and stability across diverse populations and care environments. Only by bridging these technological gaps can artificial intelligence truly evolve into an effective decision support tool in perinatal medicine, offering scientifically grounded, precision strategies for preventing and managing neonatal asphyxia to improve infant health outcomes. 7. Conclusions and future work In this study, we successfully developed and validated a ML-based predictive model for neonatal asphyxia risk assessment. The model integrates multi-source data, including clinical information, biomarkers, and socioeconomic factors, to enhance pre-dictive comprehensiveness and accuracy. By employing advanced data preprocessing techniques, such as outlier detection, missing value imputation, categorical variable encoding, and data standardization, we improved data quality, thereby bolstering model robustness. Six ML algorithms were rigorously evaluated: TabPFN, Bagging, Multilayer Perceptron (MLP), SVM, XGBoost, and RF, with exhaustive hyperparameter tuning. Results demonstrated that the XGBoost model achieved superior performance in predictive accuracy and reliability, while Bagging and RF also exhibited strong effi-cacy. SHAP value analysis provided interpretable insights into model predictions, elu-cidating the influence of key features. Despite these achievements, we acknowledge limitations in data standardization, algorithm optimization, and clinical translation. Future efforts will focus on establishing regional medical data consortiums to enhance data diversity; adopting distributed learning technologies to address data silos; conducting diagnostic-therapeutic hybrid trials to evaluate real-world clinical impact and improving clinical data standardization and healthcare interoperability. This study offers novel perspectives and tools for neonatal asphyxia risk prediction, with significant clinical implications for improving infant outcomes and optimizing healthcare resource allocation. We anticipate these findings will serve as a foundation for further research and practical applications in this field. In conclusion, this work provides innovative methodologies and directions for related research, while charting pathways for future advancements. Through continued research and technological innovation, we aim to refine predictive models further, delivering stronger decision-support tools for clinical practice. Abbreviations Declarations Author Contributions: Conceptualization, Sisi Yi and Xinfen XU; methodology, Xieli SHI; software, Xieli SHI; validation, Qiufang LI and Xinfen XU; formal analysis, Xieli SHI; investigation, Sisi Yi; resources, Yanping TENG; data curation, Sisi Yi; writing---original draft preparation, Sisi Yi; writing---review and editing, Qiufang LI; supervision, Qiufang LI and Xinfen XU. All authors have read and agreed to the published version of the manuscript. Funding: This research received no external funding. Ethical Approval and Consent to Participate declarations: This study was approved by the Institutional Review Board (IRB) of the Women's Hospital, School of Medicine, Zhejiang University (ethics number: PRO2023-3395). Informed consent was waived due to the patients have been discharged and are difficult to contact. This study will not bring additional adverse effects to the patients. The researchers will strictly adhere to the confidentiality principle. Relevant research information is only accessible to the investigators or the Ethics Committee. Consent to Publish declaration: The corresponding author confirms that all authors have read and agreed to the publication of this manuscript. The manuscript does not contain any personal data that could lead to the identification of any individual. Data Availability Statement: The corresponding author will share the data upon request due to legal and ethical reasons.. Conflicts of Interest: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgements: Not applicable References Organization WH. Maternal mortality measurement: guidance to improve national reporting. 2022. Fottrell E, Osrin D, et al. Cause-specific neonatal mortality: analysis of 3772 neonatal deaths in Nepal, Bangladesh, Malawi and India. Archives of Disease in Childhood: Fetal and Neonatal Edition. 2015;100(5):F439-F47. China Maternal and Child Health Association. Report on the development of maternal and child health in China (2019) (Part 1) (in Chinese). Chinese Journal of Women and Child Health. 2019; (5): 1-8. Chen ZL, Liu J. Interpretation of "Recommended standards for diagnosis and classification of neonatal asphyxia" (in Chinese). Chinese Journal of Contemporary Pediatrics. 2013. Douglas-Escobar M, Weiss MD. Hypoxic-Ischemic Encephalopathy: A Review for the Clinician. JAMA Pediatrics. 2015; 169(4). Techane MA, Alemu TG, Wubneh CA, et al. The effect of gestational age, low birth weight and parity on birth asphyxia among neonates in sub-Saharan Africa: systematic review and meta-analysis. Italian Journal of Pediatrics. 2021; 2022. Wojcieszek AM, Portela A. WHO recommendations on maternal and newborn care for a positive postnatal experience: strengthening the maternal and newborn care continuum. Global Health. 2023. Pahnabi A. APGAR Scores in Cesarean Deliveries: Effects of General and Spinal Anesthesia: A Systematic Review. Journal of Pediatrics. 2025. Jepson HA, Tichy AM. The Apgar score: evolution, limitations, and scoring guidelines. Birth: Issues in Perinatal Care. 1991. American Academy of Pediatrics. The Apgar score. Pediatrics. 2015. Harrell FE, Harrell J. Cox proportional hazards regression model. In: Regression Modeling Strategies. Springer; 2015. Barnova K, Kahankova R. Artificial intelligence and machine learning in electronic fetal monitoring. Methods in Engineering. 2024. Apgar V. A proposal for a new method of evaluation of the newborn infant. Anesthesia & Analgesia. 1953. Chen JL, Liu J. Interpretation of the Experts' Consensus on the criteria for the diagnosis and grading of neonatal asphyxia in China. Translational Pediatrics. 2013. Zhang YJ. Analysis of clinical risk factors and construction of a prediction model for neonatal asphyxia (in Chinese) [Master’s thesis, Anhui Medical University]. 2023. Darsareh F, Ranjbar A, Farashah MV, Mehrnoush V, Shekari M, Jahromi MS. Application of machine learning to identify risk factors of birth asphyxia. BMC Pregnancy and Childbirth. 2023; 23(1): 156. Weiner GM. Updates for the neonatal resuscitation program and resuscitation guidelines. NeoReviews. 2022. Hollmann N, Eggensperger K. TabPFN: A transformer that solves small tabular classification problems in a second. arXiv preprint arXiv:2022. Cortes C, Vapnik V. Support-vector networks. Machine Learning. 1995; 20(3): 273-297. Chen T, Guestrin C. XGBoost: A scalable tree boosting system. In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM; 2016. Breiman L. Random forests. Machine Learning. 2001; 45(1): 5-32. Breiman L. Bagging predictors. Machine Learning. 1996; 24(2): 123-140. Bisong E. The multilayer perceptron (MLP). In: Building Machine Learning and Deep Learning Models on Google Cloud Platform. Packt Publishing; 2019. Lundberg SM, Lee SI. A Unified Approach to Interpreting Model Predictions. In: Advances in Neural Information Processing Systems. 2017. Fawcett T. An introduction to ROC analysis. Pattern Recognition Letters. 2005; 27(8): 861-874. Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 09 Mar, 2026 Reviewers invited by journal 21 Oct, 2025 Editor assigned by journal 15 Oct, 2025 First submitted to journal 11 Oct, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7660175","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":532966748,"identity":"73ac3a68-a1ad-424e-9554-3a1c13679f13","order_by":0,"name":"Sisi Yi","email":"","orcid":"","institution":"Women's Hospital School of Medicine Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"Sisi","middleName":"","lastName":"Yi","suffix":""},{"id":532966749,"identity":"b9da29f7-f284-45fe-a953-04d5ed9d6301","order_by":1,"name":"Qiufang LI","email":"","orcid":"","institution":"Women's Hospital School of Medicine Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"Qiufang","middleName":"","lastName":"LI","suffix":""},{"id":532966750,"identity":"c08a14f1-ec3b-4ca6-a2a4-4715ed39d225","order_by":2,"name":"Xieli SHI","email":"","orcid":"","institution":"Women's Hospital School of Medicine Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"Xieli","middleName":"","lastName":"SHI","suffix":""},{"id":532966751,"identity":"e622f1b1-aa50-4755-86a2-af054cf652b3","order_by":3,"name":"Yanping TENG","email":"","orcid":"","institution":"Women's Hospital School of Medicine Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"Yanping","middleName":"","lastName":"TENG","suffix":""},{"id":532966752,"identity":"833139b6-b4d6-45b7-b8f2-4745c7bd1b98","order_by":4,"name":"xinfen xu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAx0lEQVRIiWNgGAWjYJCCAww2NgyMDUAWD/Fa0tJI1MLAkHYYQhOlxeBGjuHhgoTz9swzEhgfvG1jkDcnpEVyRlrC4RkJt5kZZyQwG85tYzDc2UBAC79E8oHDvD9uswG1sEnztjEkGBwgoIVNIrHhME/COR6gFvbfRGkB28KTcEACZAszUVoke54lALUkGzD2PGyWnHNOwnADIS0Gx3OMP/Mk2Nkbticf/PCmzEaeoC1wYNgAjkwJYtUDgTwJakfBKBgFo2CEAQA74zwDDUgBLQAAAABJRU5ErkJggg==","orcid":"","institution":"Zhejiang University School of Medicine Women's Hospital","correspondingAuthor":true,"prefix":"","firstName":"xinfen","middleName":"","lastName":"xu","suffix":""}],"badges":[],"createdAt":"2025-09-19 16:07:38","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7660175/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7660175/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":95222974,"identity":"e33aa59d-4c9d-4ec5-9036-c2a24b0b6685","added_by":"auto","created_at":"2025-11-05 16:21:27","extension":"xml","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7377,"visible":true,"origin":"","legend":"","description":"","filename":"itjpITJPD2501086.xml","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/4a712e66c93b5eee2b45b2fb.xml"},{"id":95063604,"identity":"6eae6c39-ab0f-4bd1-bf07-1d8cc481360b","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":919,"visible":true,"origin":"","legend":"","description":"","filename":"ITJPD250108625300.go.xml","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/7759f29f917095ad1a6686eb.xml"},{"id":95063607,"identity":"d2fdee59-780c-4129-af8d-ae832681ac93","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":889,"visible":true,"origin":"","legend":"","description":"","filename":"ITJPD2501086Import.xml","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/c343afa72730dda904511404.xml"},{"id":95063613,"identity":"0c4f29d1-68d7-42cb-8ae4-c133e2d4dff6","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"xml","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":112321,"visible":true,"origin":"","legend":"","description":"","filename":"ITJPD25010860enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/1c3e110dfed0ef3f253a8e60.xml"},{"id":95063608,"identity":"c9ac1ba3-ab78-4f61-ac8d-3666178fcc88","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"emf","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":133100,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.emf","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/999d91a847517a6280c7f13c.emf"},{"id":95222807,"identity":"dfb8a302-f270-4a7b-beb8-1dbf829592b2","added_by":"auto","created_at":"2025-11-05 16:21:10","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":121278,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/3f64d9e5bc62ad8c2b156c76.png"},{"id":95222668,"identity":"6860c1ad-b805-46f2-aafb-b0b9995ef76f","added_by":"auto","created_at":"2025-11-05 16:20:57","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":67924,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/5442f4e30e792620d13172e7.png"},{"id":95063619,"identity":"3560b1b4-c8fc-4c43-b2c1-f99af3e281cd","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":36867,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/17d1da5cddb675073c6654ff.png"},{"id":95063614,"identity":"d2f896a6-02d8-432c-ab6f-b175beead021","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":84318,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/2bbacdeac2b5a50ba6aadae5.png"},{"id":95063618,"identity":"4e87895d-8fd6-4b0d-99bd-3bb18578c43e","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":202280,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/30f9eebc82e5416297cd9872.png"},{"id":95222880,"identity":"f0360138-15ee-4ccc-8e63-7b5f33b3423d","added_by":"auto","created_at":"2025-11-05 16:21:17","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":422492,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/3bff9fc8456a63dd25e865e9.png"},{"id":95223026,"identity":"350dba66-1857-4a90-8dc6-5f53234614fc","added_by":"auto","created_at":"2025-11-05 16:21:32","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":151017,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/2a1c134cf5e9d54b7e3ac85b.png"},{"id":95063621,"identity":"586f6280-dc8a-4ee3-a599-bff5fb2316fc","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":145802,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/b828238d9398d66a988d5dcb.png"},{"id":95224524,"identity":"7a3530cb-b1a8-4b6d-8e78-0ccc3b3ee0e1","added_by":"auto","created_at":"2025-11-05 16:23:51","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":38976,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/9468002660214371d84374fd.png"},{"id":95063622,"identity":"cf5cf522-3c75-496c-a763-e55fdcd46446","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":28551,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/d724c36438bb6429aed34007.png"},{"id":95222858,"identity":"3fb0cbfd-7980-434f-bdd5-5c86dc47848f","added_by":"auto","created_at":"2025-11-05 16:21:16","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13298,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/1a14557d80280ab5752f9bbb.png"},{"id":95063623,"identity":"64c6f2d6-4823-4cb1-aa54-77ee730db50d","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":25911,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/1d366fb06cbda1aabaf4aa1e.png"},{"id":95063628,"identity":"79c70095-a02e-4fdc-8160-55b40e3e5132","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":53278,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/588026c41196f02aab739033.png"},{"id":95063610,"identity":"97d4885e-0e20-4a2c-abc7-343a2892696d","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":98667,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/5324a60ae932c63792cba0d5.png"},{"id":95063627,"identity":"7bac41b4-e037-449d-abe1-d6f39a32298b","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":44903,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/0fb7de3af22af4093d34dc35.png"},{"id":95063624,"identity":"5a35e816-d116-4e07-8266-8446a396dbb2","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"xml","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":112375,"visible":true,"origin":"","legend":"","description":"","filename":"ITJPD25010860structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/e7813cade12dd9a44dacc78f.xml"},{"id":95063625,"identity":"a07dedb1-ff05-4fac-9681-0baf97c7f4a1","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"html","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":117766,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/6090013e3b8b75b2dfde7ad3.html"},{"id":95063601,"identity":"746a8131-5cb6-4404-a76f-70a08cb2e964","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":75300,"visible":true,"origin":"","legend":"\u003cp\u003eFlow chart of the research methodology: research phases\u003c/p\u003e","description":"","filename":"1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/cdf35c70a291275070b61b02.jpg"},{"id":95063600,"identity":"6a7ffd5c-3990-4c67-8df7-9191fccf535d","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":64225,"visible":true,"origin":"","legend":"\u003cp\u003eLiterature Screening Flowchart\u003c/p\u003e","description":"","filename":"2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/39d9dbdbf9ac0e889f7a5757.jpg"},{"id":95063603,"identity":"cea6ef30-0bb1-4770-b798-895c8565ee6f","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":23375,"visible":true,"origin":"","legend":"\u003cp\u003eAn illustration showing the imbalanced distribution among four dataset groups: Apgar scores of 4, 5, 6, and 7.\u003c/p\u003e","description":"","filename":"3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/210e552203f4c0b08d005f90.jpg"},{"id":95063606,"identity":"b2903e98-672d-4483-9890-c8039c472550","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":38854,"visible":true,"origin":"","legend":"\u003cp\u003eThe heatmap illustrates the correlation between the dataset features and neonatal as-phyxia, where darker cell colors indicate a higher correlation.\u003c/p\u003e","description":"","filename":"4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/6ec9d3b7837abf13a10929eb.jpg"},{"id":95063605,"identity":"9704623a-2562-46b4-872d-16f24939d58e","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":73034,"visible":true,"origin":"","legend":"\u003cp\u003eKernel Density Estimation (KDE) plot showing the distribution of the 11 features associ-ated with neonatal asphyxia\u003c/p\u003e","description":"","filename":"5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/4a35c9271f98db43d6a78c67.jpg"},{"id":95063609,"identity":"4b8e9d84-ea7d-4116-916b-06e68e59e74a","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"jpg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":126010,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP values predicted by different models\u003c/p\u003e","description":"","filename":"6.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/df5b01dd92e3c2d7a3d32fc0.jpg"},{"id":95063616,"identity":"c2ff9e5b-0f16-480f-b7ca-da9abce9c976","added_by":"auto","created_at":"2025-11-04 01:18:20","extension":"jpg","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":102387,"visible":true,"origin":"","legend":"\u003cp\u003e(Feature Importance Plot) Comparative feature importance analysis among several models (SVM, MLP, RF, Bagging, TabPFN, XGBoost) - solo models allowing extraction of feature importance\u003c/p\u003e","description":"","filename":"7.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/912a6ebcb7885e93069d3df9.jpg"},{"id":95230249,"identity":"25df9044-ebfe-42cc-b87f-b232f98f956a","added_by":"auto","created_at":"2025-11-05 16:37:04","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1530224,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7660175/v1/5c519e10-4a40-4690-ae80-24e3b8c82832.pdf"}],"financialInterests":"","formattedTitle":"Construction and validation of neonatal asphyxia risk prediction model based on machine learning","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eNeonatal asphyxia remains one of the principal causes of neonatal mortality and long-term neurodevelopmental disabilities worldwide. According to World Health Organization (WHO) statistics, approximately 1\u0026nbsp;million newborns die from asphyxia annually, with millions more facing long-term health complications [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Developing countries bear a disproportionately higher incidence and mortality rate, imposing sub-stantial burdens on public health systems. Studies reveal that over one-third of neonatal deaths in urban India are attributable to asphyxia [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. In China, this condition ranks as the third leading cause of mortality in children under five years old, following prematurity and neonatal pneumonia, accounting for 30%-35% of neonatal deaths. Annually, approximately 160,000 children under five succumb to this condition [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Fundamentally characterized as a hypoxic disorder, neonatal asphyxia arises from multiple antenatal, intrapartum, and postnatal factors [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. It manifests as absent or depressed respiration at birth, subsequently leading to multi-organ dysfunction. With high fatality and disability rates, it constitutes a major contributor to neonatal mortal-ity, cerebral palsy, and intellectual disabilities [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. This condition not only severely im-pacts infant health but also imposes heavy socioeconomic burdens on families and so-ciety, requiring substantial resource allocation for long-term medical care, rehabilita-tion, and special education [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. The Apgar score, as a widely adopted clinical neonatal assessment tool, holds significant clinical value. Through its simple and rapid compo-site indicators, it enables preliminary evaluation of neonatal physiological status within minutes after birth, providing crucial guidance for clinical decision-making [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. However, Apgar scoring outcomes are influenced by multiple variables including ges-tational age, birth weight, maternal medication use, pharmacological/anaesthetic interventions, and congenital anomalies [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Furthermore, certain components of the scoring system involve subjective interpretation, potentially leading to inter-rater variability. Consequently, the Apgar score is primarily utilized to assess infant respon-siveness to resuscitation measures rather than predicting long-term health out-comes [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Particularly, the 1-minute score demonstrates limited indicative value for long-term clinical prognosis, highlighting the need for a reliable real-time method to derive this assessment [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eAlthough research on risk factors for neonatal asphyxia exists domestically and internationally, clinically effective risk prediction models remain scarce. Developing such models could help hospitals rationally allocate nursing resources, enable caregiv-ers to adjust interventions promptly, reduce the incidence of neonatal asphyxia, and safeguard infant survival. Accurate disease assessment and asphyxia risk prediction would facilitate personalized patient management, optimize human resource alloca-tion, enhance nursing efficiency, and ultimately save more lives. Given the critical and rapidly evolving nature of neonatal conditions, an ideal assessment tool would assist healthcare providers in promptly verifying treatment efficacy and modifying care strategies. Therefore, neonatal asphyxia risk prediction holds vital significance for im-proving patient prognoses.\u003c/p\u003e\u003cp\u003eCurrent neonatal asphyxia risk prediction models predominantly employ traditional statistical methods (such as Logistic regression and Cox proportional hazards regression). While these approaches offer simplicity and interpretability for binary classification tasks, they exhibit notable limitations. These models inherently assume linear relationships between independent and dependent variables, failing to effec-tively address complex nonlinear interactions. Furthermore, their performance is sen-sitive to data distribution characteristics, with model stability and predictive capabil-ity significantly compromised by multicollinearity or outliers [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. In contrast, ML models demonstrate substantial optimization potential through nonlinear transformations, feature engineering refinement, and sophisticated algorithmic architectures. These capabilities enable superior capture of complex data patterns and nonlinear re-lationships, thereby markedly enhancing prediction accuracy and generalization ca-pacity. Such advancements provide more efficient and precise solutions for intricate classification challenges [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. In summary, improving neonatal asphyxia outcomes represents a critical priority for hospital administrators. Developing accessible and accurate risk assessment tools holds significant clinical value by assisting nursing staff in clinical evaluations, informed decision-making, treatment efficacy verification, and rational allocation of medical resources, ultimately reducing neonatal asphyxia inci-dence. The pursuit of clinically applicable and precise risk prediction tools is para-mount for enhancing caregivers predictive capabilities, enabling early implementation of targeted interventions, and improving neonatal outcomes.\u003c/p\u003e\u003cp\u003eThis study focuses on collecting relevant datasets and developing a model to predict neonatal asphyxia risk, while validating its performance in accuracy, reliability, and practical effectiveness. Key innovations include multi-source data integration, in-corporating clinical data, biomarkers, and socioeconomic factors to construct a more comprehensive risk prediction framework. Data is sourced from a tertiary maternal and child health hospital, with empirical validation conducted to assess the model\u0026rsquo;s applicability and precision within specific clinical settings and populations. Addition-ally, the model incorporates automated prediction capabilities to enhance efficiency. The outcomes of this research will provide healthcare professionals with a practical tool to support more precise diagnostic and intervention decisions during prenatal and intrapartum stages. By enabling early identification of high-risk neonates, the model aims to optimize medical resource allocation and improve the efficiency and effec-tiveness of healthcare systems. Furthermore, this study will contribute a scientific foundation for shaping public health policies, advancing perinatal care services, and refining preventive strategies. Through analyzing the relative contributions of diverse factors to neonatal asphyxia risk, policymakers will be empowered to design targeted preventive measures and intervention plans, ultimately safeguarding neonatal health outcomes.\u003c/p\u003e"},{"header":"2. Related works","content":"\u003cp\u003eThe Apgar score serves as a physiological assessment tool based on five clinical indicators\u0026mdash;skin color, heart rate, respiratory rhythm, muscle tone, and reflex irritability (each scored 0\u0026ndash;2 points, totaling 10 points) [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. This scoring system provides peri-natal medical teams with objective criteria for evaluating intrauterine adaptation sta-tus (1-minute score) and dynamically monitoring resuscitation effectiveness (5/10-minute scores) through rapid, standardized multi-dimensional analysis. While its timeliness and operational practicality hold significant clinical value, it is critical to acknowledge its limitations as a screening tool for asphyxia. The Chinese diagnostic consensus [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e] for neonatal asphyxia emphasizes that low Apgar scores must be inter-preted alongside auxiliary methods such as arterial blood gas analysis, neurological assessments, and multi-organ injury evaluations to ensure comprehensive diagnosis and mitigate misjudgment risks arising from over reliance on a single scoring metric.\u003c/p\u003e\u003cp\u003eIn recent years, researchers have explored risk prediction models for neonatal asphyxia through diverse methodologies. A study by Zhang Youjun [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e] employed binary logistic regression to identify nine clinical risk factors associated with neonatal as-phyxia, including preterm birth, abnormal fetal position, intrapartum fever, umbilical cord anomalies, placental abnormalities, fetal distress, regular prenatal checkups, pro-longed labor, and amniotic fluid abnormalities. Internal validation via 1,000 iterations of Bootstrap resampling yielded a final nomogram prediction model with a concord-ance index (C-index) of 0.810, indicating strong predictive performance. Calibration curves generated using R Studio further demonstrated robust alignment between pre-dicted and observed outcomes. Additionally, an Iranian cross-sectional study [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e] ap-plied multiple ML algorithms\u0026mdash;logistic regression, decision tree (DT) classifier, random forest (RF) classifier, XGBoost (Extreme Gradient Boosting) classifier, permutation classifier, feedforward deep learning, LightGBM (Light Gradient Boosting Machine), and support vector machine (SVM)\u0026mdash;to predict birth asphyxia. The findings high-lighted RF classification as the most accurate algorithm for this purpose. These studies collectively advance novel approaches for early warning and intervention in neonatal asphyxia.\u003c/p\u003e"},{"header":"3. Methods","content":"\u003cp\u003eOne primary objective of this study is to predict the incidence of neonatal asphyxia using practical ML methods and tools. To achieve this goal, six ML algorithms were employed: TabPFN, Bootstrap Aggregating (Bagging), Multilayer Perceptron (MLP), SVM, XGBoost, and RF. The research workflow comprised four main phases: preprocessing, standardization, evaluation, and modeling. For clarity, Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e illustrates the complete sequence of stages involved in this investigation. All ML models and techniques were developed and implemented using Python 3.8. Computational tasks were executed on a Windows operating system with an Intel(R) Core(TM) i7-10750H CPU @ 2.60 GHz (2592 MHz, 6 cores, 12 logical processors).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e3.1. Data acquisition\u003c/h2\u003e\u003cp\u003eThe data for this study were extracted from the neonatal electronic medical record system of Women's Hospital, School of Medicine, Zhejiang University. Structured Query Language (SQL) was utilized to retrieve complete case data meeting research criteria from January to December 2024. Adopting a retrospective cohort design, we constructed an analytical dataset comprising 40 clinical features\u0026mdash;including gesta-tional age, delivery mode, and umbilical cord abnormalities\u0026mdash;after feature engineering (feature selection, categorical variable transformation, and missing value imputation). A total of 250 neonates with birth asphyxia were enrolled, consisting of 187 mild asphyxia cases (74.8%) and 63 moderate asphyxia cases (25.2%). The cohort included 146 males (58.4%) and 103 females (41.2%), with a median birth weight of 3150g (IQR: 2850\u0026ndash;3450g) and a median gestational age of 38\u0026thinsp;+\u0026thinsp;3 weeks (range: 32\u0026thinsp;+\u0026thinsp;1\u0026ndash;41\u0026thinsp;+\u0026thinsp;5 weeks). All cases met the following inclusion criteria: (1) singleton live birth at the study hospital; (2) complete perinatal maternal-neonatal clinical records; (3) definitive Apgar scores and asphyxia diagnosis. Exclusion criteria comprised: (1) intrauterine death; (2) major congenital anomalies (e.g., congenital 3heart disease); (3) multiple gestation; (4) miss-ing critical clinical data. Data collection strictly adhered to clinical practice guidelines, with Apgar scoring and diagnostic assessments performed by attending neonatologists or higher-ranking clinicians. Study variables spanned three dimensions: neonatal bio-logical characteristics (sex, birth weight, gestational age), delivery-related parameters (labor duration, amniotic fluid characteristics, umbilical cord abnormalities), and ma-ternal pregnancy features (gestational complications, parity, delivery mode). As sys-tematically detailed in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, obstetric-related parameters constituted 62.5% (25/40) of variables, neonatal assessment metrics accounted for 27.5% (11/40), and maternal baseline characteristics comprised 10.0% (4/40).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eComprehensive overview of the features.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSN\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAttribute name\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eDescription\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eType\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMaternal Age\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;20\u0026thinsp;=\u0026thinsp;0,\u0026lt;35\u0026thinsp;=\u0026thinsp;1, \u0026ge;\u0026thinsp;35\u0026thinsp;=\u0026thinsp;2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGravid\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eThe number of pregnancies\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNumerical\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eParity\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003ePregnancies that resulted in the delivery at \u0026gt;\u0026thinsp;6months gestation, of either a live birth or stillbirth\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNumerical\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMother's job\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eresidence\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eBMI2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eBody mass index in 2nd trimester18.5-24.9\u0026thinsp;=\u0026thinsp;1, 25-29.9\u0026thinsp;=\u0026thinsp;2, \u0026gt;30\u0026thinsp;=\u0026thinsp;3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eConception method\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003enatural conception\u0026thinsp;=\u0026thinsp;1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e8\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSex\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eGender of the fetus, male\u0026thinsp;=\u0026thinsp;1, female\u0026thinsp;=\u0026thinsp;0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e9\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003esingleton\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eadverse pregnancy history\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMiscarriage\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGestational Age\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e28\u0026thinsp;~\u0026thinsp;31\u0026thinsp;+\u0026thinsp;6w\u0026thinsp;=\u0026thinsp;0, 32\u0026thinsp;~\u0026thinsp;33\u0026thinsp;+\u0026thinsp;6w\u0026thinsp;=\u0026thinsp;1, 34\u0026thinsp;~\u0026thinsp;36\u0026thinsp;+\u0026thinsp;6w\u0026thinsp;=\u0026thinsp;2, 37\u0026thinsp;~\u0026thinsp;40\u0026thinsp;+\u0026thinsp;=\u0026thinsp;3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e13\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eBirth Weight\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u0026lt;1500g\u0026thinsp;=\u0026thinsp;1, 1500\u0026thinsp;~\u0026thinsp;2500g\u0026thinsp;=\u0026thinsp;2, \u0026ge;\u0026thinsp;2500g\u0026thinsp;=\u0026thinsp;3, \u0026gt;4000g\u0026thinsp;=\u0026thinsp;4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e14\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eType of Delivery\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eNatural childbirth\u0026thinsp;=\u0026thinsp;1, C-section\u0026thinsp;=\u0026thinsp;2, Forceps-assisted delivery\u0026thinsp;=\u0026thinsp;3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e15\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAbnormal fetal heart rate\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLabor Process\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eprolonged labor\u0026thinsp;=\u0026thinsp;2, precipitate labor\u0026thinsp;=\u0026thinsp;3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e17\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCord around the neck (CAN) Umbilical Cord Entanglement\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e18\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAbnormal umbilical artery Doppler S/D ratio\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e19\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFetal growth restriction (FGR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e20\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAbnormal Fetal Position\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e21\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePlacenta previa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e22\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePlacental adhesion\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e23\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePlacental implantation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e24\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePlacental abruption\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFetal Hypoxia\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1,else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e26\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePremature rupture of membranes (PROM)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e27\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAmniotic Fluid Contamination\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e28\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAmniotic fluid index (AFI)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eoligohydramnios\u0026thinsp;=\u0026thinsp;1, normal amniotic fluid volume\u0026thinsp;=\u0026thinsp;2, polyhydramnios\u0026thinsp;=\u0026thinsp;3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e29\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIntrauterine infection\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e30\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eUterine rupture\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e31\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePelvic/abdominal adhesions\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e32\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGestational hypertension\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e33\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGestational diabetes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e34\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePreeclampsia\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1,else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e35\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eModerate preeclampsia\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e36\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePregnancy complicated by thyroid disease\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e37\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eedication taken in early pregnancy\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e38\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAnemia in Pregnancy\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e39\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCo-existing infectious disease\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIf exists 1, else 0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eApgar Score\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4,5,6,7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNominal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e3.2. Feature extraction\u003c/h2\u003e\u003cp\u003eNeonatal asphyxia is a severe and common perinatal complication with a complex pathogenesis involving multiple potential risk factors. These factors may encom-pass maternal health during pregnancy, intrapartum conditions, and neonatal physio-logical characteristics. To deeply understand their impact on neonatal asphyxia, the extraction of critical feature variables is essential. Feature extraction enables the iden-tification of core variables strongly associated with neonatal asphyxia from vast clini-cal datasets, thereby enabling more precise identification of high-risk populations and providing actionable insights for targeted interventions. The meta-analytic approach, as a systematic research methodology, offers distinct advantages. First, it integrates data from numerous studies, amplifies sample sizes, and enhances statistical power, allowing more accurate estimation of the association strength between risk factors and neonatal asphyxia. Second, meta-analysis synthesizes and compares findings across studies, assesses heterogeneity among them, and reveals the stability of potential risk factors across diverse populations, regions, and study conditions. Furthermore, it compensates for limitations inherent in individual studies\u0026mdash;such as small sample sizes or restricted designs\u0026mdash;by delivering comprehensive and objective conclusions through pooled analysis. By employing meta-analysis, this study aims to clarify key determi-nants of neonatal asphyxia. Such insights will empower clinicians to proactively iden-tify high-risk neonatesduring perinatal management, implement tailored preventive strategies, and optimize perinatal care protocols. Ultimately, this approach seeks to reduce the incidence of neonatal asphyxia and improve neonatal health outcomes.\u003c/p\u003e\u003cdiv id=\"Sec6\" class=\"Section3\"\u003e\u003ch2\u003e3.2.1. Retrieval strategy\u003c/h2\u003e\u003cp\u003eA computer-based search was conducted in the following databases: China National Knowledge Infrastructure (CNKI), Wanfang, VIP, Chinese Biomedical Literature Database (CBM), PubMed, Embase, Web of Science, and Cochrane Library. The search was restricted to Chinese and English languages, with a timeframe spanning from the inception of each database to June 18, 2024. Search terms combined controlled vocab-ulary (subject headings) and free-text keywords. Chinese search words: \"neonatal as-phyxia/asphyxia/perinatal asphyxia\"; \"Risk factors/risk factors/influencing fac-tors/related factors/predictors/causes/investigations\". English search terms: \u0026ldquo;risk fac-tor /relevant factors/predictor/associate factors/influence\u0026rdquo;;\u0026ldquo;neonatal asphyxia / as-phyxia of the newborn / a.neonatorum / apnea neonatorum / neonate asphyxia / neo-nates asphyxia / newborns asphyxia/'apnoea neonatorum/ newborn suffocation/ newborn hypoxia/'newborn hypoxia\u0026rdquo;. The retrieval strategy is adjusted appropri-ately according to the requirements of each database.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section3\"\u003e\u003ch2\u003e3.2.2. Literature screening and data extraction\u003c/h2\u003e\u003cp\u003eIn this study, researchers independently conducted literature screening and data extraction, followed by cross-verification to ensure the quality and relevance of selected literature. For resolving discrepancies in contentious cases, a third re-searcher was introduced to arbitrate, ensuring impartiality and objectivity in the re-view process. Additionally, Endnote software was employed to systematically manage retrieved literature, facilitating subsequent analysis and integration. The process of literature screening is divided into two important steps: preliminary screening and full-text reading. In the initial screening stage, researchers focused on reviewing the ti-tles and abstracts of the articles to determine whether they met the inclusion criteria. The next steps involved a thorough full-text reading of the eligible literature, thus fur-ther confirming its suitability. The final extracted literature information covered mul-tiple dimensions, including author, publication year, study country, study object, study type, diagnostic criteria, sample size and related risk factors.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section3\"\u003e\u003ch2\u003e3.2.3. Literature quality evaluation\u003c/h2\u003e\u003cp\u003eThis study was independently conducted by two professionally trained researchers to ensure scientific rigor throughout the research process. For literature quality as-sessment, we employed risk-of-bias evaluation tools to systematically analyze and as-sess the reliability and validity of included studies. Cohort and case-control studies were evaluated using the Newcastle-Ottawa Scale (NOS), a validated quality assess-ment instrument that provides detailed scoring across three domains. The NOS is di-vided into two main parts, respectively for the evaluation of cohort and case-control studies, and the overall is composed of three parts. Specifically, the object selection consisted of four items, each of which was awarded one point; The inter-group com-parability component consists of one item, with a maximum score of two points; The evaluation section for the outcome or exposure factor contains three items, again scoring one point for each item. The scale has a total score of nine points, and the scor-ing criteria is: the research literature is divided into different quality levels, a score of 0 to 3 is classified as low quality research, a score of 4 to 6 is medium quality research, while a score of 7 to 9 is regarded as high quality research. In addition, the literature Quality evaluation of cross-sectional studies was carried out according to the evalua-tion criteria recommended by the Agency for Healthcare Research and Quality (AHRQ). The criterion contains 11 assessment items with response options of \"yes\", \"no\" or \"not clear\" on an 11-point scale. Under this scoring system, a score of 0 to 3, a score of 4 to 7, and a score of 8 and above correspond to low, medium and high quality literature, respectively. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e illustrates the Screening process, which is divided into three main stages: Identification, screening, and Included.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section3\"\u003e\u003ch2\u003e3.2.4. Statistical methods\u003c/h2\u003e\u003cp\u003eMeta-analysis was conducted using Stata 18.0 software. For continuous variables, weighted mean difference (WMD) and its 95% confidence interval (95% CI) were calculated, while odds ratios (ORs) with 95% CIs were applied for dichotomous variables. Heterogeneity was assessed using the I\u0026sup2; statistic and Cochran\u0026rsquo;s Q-test. A fixed-effect model was employed when heterogeneity was deemed acceptable (P\u0026thinsp;\u0026gt;\u0026thinsp;0.10 and I\u0026sup2; \u0026lt; 50%), whereas a random-effects model was adopted for significant heterogeneity (P\u0026thinsp;\u0026le;\u0026thinsp;0.10 or I\u0026sup2; \u0026ge; 50%). Descriptive analysis was performed for studies where effect sizes could not be pooled. In this study, the number of articles included and meta-merged was small, and less than 10 articles were analyzed for a single risk factor, so funnel plot analysis was not performed. In addition, this study set that when P\u0026thinsp;\u0026lt;\u0026thinsp;0.05, it was considered statistically significant. A total of 8 studies involving 5,593 cases were in-cluded, analyzing 24 risk factors. The results of meta-analysis showed that 11 risk fac-tors were statistically significant. Among them, low birth weight infants (OR\u0026thinsp;=\u0026thinsp;3.826), maternal age less than 20 years old (OR\u0026thinsp;=\u0026thinsp;1.946), premature infants (OR\u0026thinsp;=\u0026thinsp;2.609), assisted delivery (OR\u0026thinsp;=\u0026thinsp;4.238), prolonged labor (OR\u0026thinsp;=\u0026thinsp;3.880), non-cephalic presentation (OR\u0026thinsp;=\u0026thinsp;4.861), intrauterine distress (OR\u0026thinsp;=\u0026thinsp;5.412), amniotic fluid contamination (OR\u0026thinsp;=\u0026thinsp;8) 53), primiparity (OR\u0026thinsp;=\u0026thinsp;3.539), umbilical cord around the neck (OR\u0026thinsp;=\u0026thinsp;3.830), and anemia dur-ing pregnancy (OR\u0026thinsp;=\u0026thinsp;1.667) were risk factors for neonatal asphyxia. Sensitivity analysis showed that the combined results of fixed effect model and random effect model were basically stable. Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e presents the results of the heterogeneity analysis of different risk factors for neonatal asphyxia in the meta-analysis.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eStandardized Mean Differences (95% Confidence Intervals) and Heterogeneity (P-value, I\u0026sup2;) of the Association Between 14 Risk Factors and Neonatal Asphyxia from Random-Effects Me-ta-Analyses\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003eHeterogeneity\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAnalysis\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eStudies, No.\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eβ (95%CI)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eP value\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eI2, %\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLow birth weight infant\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3.826(2.765, 5.296)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e36.0\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eJumbo baby\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.700(0.338, 1.450)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.337\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eThe maternal age was less than 20 years\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.946(1.305, 2.902)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.001\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePremature infant\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.609(1.815, 3.752)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCaesarean section\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.784(0.485, 1.266)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.319\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eassisted delivery\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4.238(2.907, 6.180)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e35.6\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eProlonged labor\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3.880(1.072, 14.051)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.039\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e83.9\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAbnormal fetal position\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4.861(3.288, 7.188)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIntrauterine distress\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5.412(3.578, 8.187)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eUmbilical cord around neck\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3.830(2.228, 6.583)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAmniotic fluid contamination\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8.532(3.687, 19.741)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e84.9\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eprimiparity\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3.539(2.202, 5.688)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e4.9\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePregnancy-induced hypertension syndrome\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.155(0.709, 6.547)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.176\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAnemia of pregnancy\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.667(1.174, 2.368)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.004\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e53.8\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e3.3. Distribution of Extracted Features\u003c/h2\u003e\u003cp\u003eThis study analyzed clinical data from 250 neonates, encompassing 12 core clinical indicators, including gestational age, Apgar scores, and birth weight. Data analysis showed that only 2 cases of \"Amniotic Fluid Contamination\" had missing values, and 1 case was missing for each of the other 11 indicators. The total missing values of the whole data set accounted for 0.43% (13/3000), and the data completeness was 99.57%. This indicates that the dataset has reliable statistical integrity with applicability to clinical studies. The distribution of missing data and treatment effects are detailed in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e. Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e shows that the clinical core index, neonatal Apgar score, is concen-trated in the 4\u0026ndash;7 region (accounting for 100% of the total sample), which meets the di-agnostic criteria of mild to moderate neonatal asphyxia defined by the International Academy of Pediatrics (AAP) (Apgar score 4\u0026ndash;6 is moderate asphyxia, and 7 is the crit-ical value of mild asphyxia) [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. The distribution characteristics of the score suggest that the study cohort accurately covers the risk group of perinatal mild to moderate asphyxation, and provides a clinically relevant research sample for exploring the pathophysiological characteristics and prognostic factors of neonates in this subgroup. It should be noted in particular that severe asphyxia (Apgar\u0026thinsp;\u0026le;\u0026thinsp;3) cases were not includ-ed in this dataset. This design not only meets the ethical requirements of the study, but also ensures the homogeneity of the pathological degree of the analyzed subjects. Fig-ure 4 is a Correlation Heatmap that shows the correlation between different variables in a dataset. The feature correlation matrix shows that most of the features have weak linear correlations (absolute correlation coefficient\u0026thinsp;\u0026lt;\u0026thinsp;0.3), which indicates that the fea-tures in the dataset are relatively independent and may reduce the effect of multicol-linearity on the model. Kernel density estimation (KDE) analysis of Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e revealed patterns of association between different perinatal characteristics and neonatal as-phyxia severity (stratified by Apgar score). Specifically, newborns with intrauterine distress, contaminated amniotic fluid, umbilical cord around the neck, and anemia during pregnancy are more likely to suffer from asphyxation. In addition, premature infants and newborns with abnormal fetal position have a higher proportion of as-phyxia.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eThe number of Missing Values in Dataset.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"2\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAttributes\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo. of missing values\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMaternal Age\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eApgar Score\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBirth weight\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGestational Age\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eType of Delivery\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLabor Process\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAbnormal Fetal Position\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFetal Hypoxia\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAmniotic Fluid Contamination\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eUmbilical Cord Entanglement\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eParity\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAnemia in Pregnancy\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"4. Experimental Setup","content":"\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003e4.1. Data preprocessing\u003c/h2\u003e\u003cp\u003eData preprocessing serves as a critical phase in data analysis and ML, encompassing data cleaning, transformation, and integration to enhance analytical quality. Multiple techniques were applied, including outlier detection, missing value imputa-tion, categorical variable encoding, and data standardization, to optimize overall data integrity. For handling missing data, the RF algorithm was employed to impute missing values. To improve model effectiveness, normalization was performed using the StandardScaler function from the Sklearn library, which scales features to a standard-ized range (0\u0026ndash;1). This ensures consistent scaling across variables, thereby mitigating bias from disparate measurement units and enhancing model performance. To address class imbalance, this study implemented the Synthetic Minority Over-sampling Technique (SMOTE). SMOTE strategically expands the decision boundary of the minority class by generating synthetic samples between minority class instances and their k-nearest neighbors within the feature space.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003e4.2. Model Introduction\u003c/h2\u003e\u003cp\u003eThis study employed random sampling to partition the dataset into a training set (70%) and test set (30%), ensuring robust model evaluation accuracy and validity. ML algorithms were applied to both the original dataset and resampled datasets (post-class balancing). To enhance result robustness and reliability, each algorithm was validated across multiple randomly partitioned training-test splits. A consistent ran-dom seed was used for data shuffling to guarantee reproducibility. Through meticu-lous feature selection, the study aimed to identify the optimal feature subset that maximizes predictive performance, thereby deepening insights into inherent data pat-terns.\u003c/p\u003e\u003cp\u003eML classification techniques, as a core component of AI, enable precise data classification and prediction. In this research, we utilized diverse ML classifiers, including TabPFN, Bagging, Multilayer Perceptron (MLP), SVM, XGBoost, and RF. These classi-fiers were applied to 11 feature variables encompassing critical clinical indicators, with hyperparameter optimization tailored to each ML model. The subsequent subsec-tions elaborate on the fundamental principles of each ML classification method.\u003c/p\u003e\u003cdiv id=\"Sec14\" class=\"Section3\"\u003e\u003ch2\u003e4.2.1. TabPFN\u003c/h2\u003e\u003cp\u003eTabPFN is a Transformer-based model specifically designed for classification tasks on small-scale tabular datasets. Trained offline by learning posterior predictive distributions over synthetic datasets, it defines a hypothesis space that captures relationships between input features and output l Following single-epoch training, TabPFN enables rapid prediction without requiring hyperparameter tuning. It pro-cesses training and testing samples simultaneously, generating predictions through a single forward pass, thereby achieving notable computational efficiency [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section3\"\u003e\u003ch2\u003e4.2.2. SVM\u003c/h2\u003e\u003cp\u003eSVM is a widely-used ML algorithm primarily employed for classification and re-gression tasks. It operates by identifying an optimal hyperplane that maximizes the margin between classes in the feature space. To address nonlinear separability, SVM leverages the kernel trick, implicitly mapping data into higher-dimensional spaces. Additionally, it employs soft margins to accommodate noisy or overlapping data points. While SVM excels in applications such as image recognition and text classifica-tion, its computational inefficiency for large datasets and sensitivity to parameter se-lection (e.g., kernel type, regularization) pose practical limitations. These characteris-tics necessitate careful tuning and scalability considerations in real-world implemen-tations [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section3\"\u003e\u003ch2\u003e4.2.3. XGBoost\u003c/h2\u003e\u003cp\u003eXGBoost is a highly efficient gradient-boosted tree algorithm that optimizes the tree-building process to enhance model performance and computational speed. It in-corporates regularization terms to control model complexity and mitigate overfitting, supports diverse loss functions for enhanced flexibility across tasks, and leverages parallelized processing to accelerate training. These features collectively enable XGBoost to demonstrate exceptional performance in various ML applications, partic-ularly in handling structured data and large-scale datasets [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section3\"\u003e\u003ch2\u003e4.2.4. RF\u003c/h2\u003e\u003cp\u003eRF is an ensemble learning algorithm that enhances model accuracy and robustness by constructing multiple decision trees. In RF, each tree is trained on randomly selected subsets of features and samples (via bootstrapping). The final prediction is de-rived by aggregating outputs through majority voting (for classification) or averaging (for regression), effectively reducing model variance and improving resistance to over-fitting. As a versatile supervised learning algorithm, RF excels in both classification and regression tasks, efficiently handles high-dimensional data, and provides robust estimates of feature importance through metrics like Gini impurity reduction or mean decrease in accuracy [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec18\" class=\"Section3\"\u003e\u003ch2\u003e4.2.5. Bagging\u003c/h2\u003e\u003cp\u003eBagging is a ML technique that enhances prediction accuracy by integrating outputs from multiple models. Its core principle involves constructing diverse models on distinct data subsets and consolidating their predictions to form a robust learner. As an ensemble method, Bagging employs random sampling with replacement to create mul-tiple data subsets, trains models independently on these subsets, and aggregates their predictions (e.g., majority voting for classification, averaging for regression) to achieve more stable outcomes. A key advantage lies in its ability to reduce model variance by averaging prediction errors across individual models while simultaneously simplifying computational complexity through parallelizable training processes [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec19\" class=\"Section3\"\u003e\u003ch2\u003e4.2.6. MLP\u003c/h2\u003e\u003cp\u003eMLP is a widely used artificial neural network in ML. As a supervised learning algorithm, it employs backpropagation to adjust network weights based on errors encountered during training. An MLP architecture comprises multiple layers: an input layer, one or more hidden layers, and an output layer. Neurons within each layer are interconnected via weighted connections, enabling information flow and hierarchical computation across the network [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec20\" class=\"Section2\"\u003e\u003ch2\u003e4.3. Parameter setting and evaluation metrics\u003c/h2\u003e\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e presents the selected parameters and fully optimized hyperparameter configurations of the best-performing ML models across categories. To ensure experi-mental fairness and reproducibility, all models were evaluated under identical train-ing/testing splits and computational constraints. Establishing a robust evaluation framework is critical when assessing classifier performance. To enhance model inter-pretability, this study employed SHAP value analysis. SHAP values, rooted in coopera-tive game theory, quantify the contribution of each feature to model predictions by calculating its marginal contribution to the prediction outcome. This method eluci-dates the relationship between features and model outputs by decomposing predic-tions into additive contributions from individual features. The sum of all fea-ture-specific Shapley values equals the difference between the model\u0026rsquo;s predicted value and a baseline value (typically the dataset\u0026rsquo;s mean prediction) [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e presents a confusion matrix, where True Positives (TP) are the correctly identified positive instances, and True Negatives (TN) are the correctly identified negative instances. These values form the basis for calculating critical metrics such as accuracy, recall, precision, and F1-score. The confusion matrix serves as a cornerstone tool for evaluating classifier performance. It provides granular data on true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN), forming the basis for calculating critical metrics such as accuracy, recall, precision, and F1-score. These metrics are indispensable for assessing classification model effectiveness, as they comprehensively characterize classifier performance while revealing strengths and weaknesses to determine suitability for specific tasks [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. Equations\u0026nbsp;\u003cspan refid=\"Equ1\" class=\"InternalRef\"\u003e1\u003c/span\u003e to 4 show the core metric formulation used to evaluate the performance of the classification model. These metrics are the key evalua-tion metrics used in ML to evaluate the performance of classification models and in-clude Accuracy, Precision, Recall, and F1-score.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eHyperparameter results for top-performing models in each category for asphyxia admis-sion prediction across the complete case.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSVM\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eValue\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eMLP\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eValue\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eRF\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eValue\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eC\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003ealpha\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1e-4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003ecriterion\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003egini\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ekernel\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003erbf\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003ehidden_layer_sizes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e100\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003esplitter\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003ebest\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003edegree\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eactivation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003erelu\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003en_estimators\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003egamma\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003escale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003esolver\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eadam\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003em\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eauto\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eclass_weight\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ebalanced\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003emax_depth\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e10\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003etol\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1e-3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003emin_samples_split\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBagging\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eValue\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eTabPFN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eValue\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eXGBoost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eValue\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003emax_\u003c/p\u003e\u003cp\u003efeatures\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eN_ensemble_configs\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e32\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003ecriterion\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003ereg:\u003c/p\u003e\u003cp\u003esquarederror\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003en_\u003c/p\u003e\u003cp\u003eestimators\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003edevice\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003ecpu\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eMax\u003c/p\u003e\u003cp\u003e_depth\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e6\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ebootstrap\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrue\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eInference\u003c/p\u003e\u003cp\u003e_precision\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003efloat32\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eNum\u003c/p\u003e\u003cp\u003e_boost\u003c/p\u003e\u003cp\u003e_round\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eNormalize\u003c/p\u003e\u003cp\u003e_y\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTrue\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003elearning_rate\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.3\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eseed\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003esubsample\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1.0\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003edevice\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003ecpu\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003ecolsample_bytree\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1.0\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"5. Results and analysis","content":"\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e6\u003c/span\u003e systematically compares the predictive performance metrics of six machine learning models, including core evaluation parameters such as recall, precision, and F1-score. The experimental results demonstrate significant performance differences among the models in neonatal asphyxia risk prediction. The XGBoost model exhibited optimal predictive performance (recall: 0.82, precision: 0.83, F1-score: 0.82), significantly outperforming ensemble methods (Bagging: F1\u0026thinsp;=\u0026thinsp;0.79; Random Forest: F1\u0026thinsp;=\u0026thinsp;0.75) and traditional models (SVM: F1\u0026thinsp;=\u0026thinsp;0.67; MLP: F1\u0026thinsp;=\u0026thinsp;0.66; TabPFN: F1\u0026thinsp;=\u0026thinsp;0.69). The improved recall of Random Forest (0.75 vs. 0.71 in previous Iranian studies) may stem from the inclusion of comprehensive placental pathology indicators (e.g., adhesion, implantation), while TabPFN's suboptimal performance (F1\u0026thinsp;=\u0026thinsp;0.69) confirms the limitations of Transformer architectures in small-scale clinical datasets (n\u0026thinsp;=\u0026thinsp;250). The performance differences further validate XGBoost's superiority in handling nonlinear feature interactions (e.g., umbilical cord abnormalities combined with prolonged labor), whereas traditional statistical models (e.g., logistic regression) underperform due to their linearity assumptions. These findings provide a more precise risk stratification tool for clinical practice, particularly for optimizing medical resource allocation in resource-constrained settings through early identification of high-risk cases. Future research should address sample representativeness limitations through multicenter validation and develop more interpretable clinical decision support systems to facilitate model translation.\u003c/p\u003e\u003cp\u003eThis study systematically evaluated the predictive value of clinical features for neonatal asphyxia risk using SHAP analysis. Figure\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e (SHAP bar plot) identified umbilical cord entanglement (mean SHAP: +1.86) and amniotic fluid contamination (+\u0026thinsp;1.50) as the strongest positive predictors, while Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e (multi-model scatter plot) revealed algorithmic interpretation differences, particularly XGBoost's superior predictive capability for cord entanglement (SHAP range: +2.0 to +\u0026thinsp;3.0). Key findings include: 1) Umbilical cord entanglement (OR\u0026thinsp;=\u0026thinsp;3.830, 95%CI: 2.228\u0026ndash;6.583), assisted delivery (OR\u0026thinsp;=\u0026thinsp;4.238, 95%CI: 2.907\u0026ndash;6.180), prolonged labor (OR\u0026thinsp;=\u0026thinsp;3.880, 95%CI: 1.072\u0026ndash;14.051), and amniotic fluid contamination (OR\u0026thinsp;=\u0026thinsp;8.532, 95%CI: 3.687\u0026ndash;19.741) emerged as critical risk predictors, with severe contamination (Grade III) reaching SHAP values\u0026thinsp;\u0026gt;\u0026thinsp;+\u0026thinsp;2.5; 2) Delivery mode consistently ranked as the most stable high-weight predictor (top-2 in all models), its bimodal SHAP distribution reflecting risk variations across delivery methods; 3) Apgar score showed limited predictive value (SHAP\u0026thinsp;\u0026lt;\u0026thinsp;0.3); 4) Maternal age exhibited nonlinear effects - younger mothers (\u0026lt;\u0026thinsp;25 years) carried higher risk (SHAP: +0.8), while advanced maternal age (\u0026ge;\u0026thinsp;35) demonstrated neutral/protective effects (SHAP: -0.15 to 0), potentially reflecting differential prenatal monitoring intensity. These results suggest clinicians should prioritize high-risk feature combinations (e.g., cord entanglement with amniotic contamination) and enhance monitoring for primiparas (OR\u0026thinsp;=\u0026thinsp;3.539, 95%CI: 2.202\u0026ndash;5.688). The weaker explanatory power of TabPFN (SHAP range: \u0026plusmn;1.5) further confirms Transformer's limitations in small clinical datasets.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eConfusion Matrix.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003ePredicted\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eNegative\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003ePositive\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eActual\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNegative\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eTrue Negative (TN)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eFalse Positive (FP)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePositive\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eFalse Negative (FN)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTrue Positive (TP)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e\n$$\\:Accuracy=\\frac{TP+TN}{TP+TN+FP+FN}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ2\" name=\"EquationSource\"\u003e\n$$\\:Precision=\\frac{TP}{TP+FP}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e2\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equ3\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ3\" name=\"EquationSource\"\u003e\n$$\\:Recall=\\frac{TP}{TP+FN}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e3\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab6\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 6\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eThe result of the six ML models with different categories and complete case.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eClassifier Model\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eRecall Score\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003ePrecision Score\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eAccuracy Score\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eF1 Score\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSVM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.66\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.71\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.66\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.67\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMLP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.66\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.72\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.66\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.66\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRF\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.75\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.76\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.75\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.75\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBagging\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.78\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.8\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.78\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.79\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTabPFN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.69\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.72\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.69\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.69\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eXGBoost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.82\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.83\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.82\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.82\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e"},{"header":"6. Discussion","content":"\u003cp\u003eEarly prediction of neonatal asphyxia plays a pivotal role in perinatal management, offering benefits such as optimized healthcare resource allocation, reduced demand for intensive care, and provision of critical risk information to families for psychological and logistical preparedness. This study established a predictive framework comprising two core components: data preprocessing and model evaluation. During the data preprocessing phase, data quality was enhanced through systematic steps, missing value imputation (using the RF algorithm), categorical variable encoding, and data standardization. Additionally, the SMOTE was applied to address class imbalance, thereby improving model accuracy and laying a robust foundation for subsequent training. In the model evaluation phase, six advanced classifiers\u0026mdash;TabPFN, Bagging, MLP, SVM, XGBoost, and RF\u0026mdash;were rigorously assessed for their predictive efficacy in neonatal asphyxia. The study integrated 11 evidence-based independent risk factors spanning critical prenatal to intrapartum variables, such as low birth weight, primi-parity, abnormal fetal position, labor dynamics, fetal monitoring indices, umbilical cord abnormalities, and maternal comorbidities. A randomized sampling strategy en-sured clinically representative distribution of subgroups in training and validation sets.\u003c/p\u003e\u003cp\u003eDuring the model validation phase, while traditional ML models (e.g., XGBoost) demonstrated satisfactory predictive accuracy and reliability, their clinical application faces two primary challenges. The first is the limited interpretability of the model, alt-hough SHAP value analysis was employed to quantify feature contributions to predic-tions, discrepancies persist between certain SHAP values and clinical knowledge, which may undermine clinicians\u0026rsquo; trust in the model. The second is the \"black box\" problem of deep learning methods, despite achieving high recall rates in our center\u0026rsquo;s test dataset with advanced deep learning methods (e.g., enhanced Transformer archi-tectures), their opaque feature extraction mechanisms fail to meet transparency re-quirements for clinical decision-making.\u003c/p\u003e\u003cp\u003eFuture research must address two critical translational medicine challenges: 1. Enhancing Model Interpretability: Through advanced SHAP value analysis or alterna-tive explainability tools to enable visualization and clinical validation of feature con-tributions, thereby strengthening clinical trust in model outputs. 2. Conducting Prospective Clinical Trials: To validate the model\u0026rsquo;s real-world impact on clinical end-points, ensuring its effectiveness and reliability in practical healthcare settings. As this study represents a single-center investigation with geographically concentrated data, future work requires multicenter validation to confirm the model\u0026rsquo;s generalizability and stability across diverse populations and care environments. Only by bridging these technological gaps can artificial intelligence truly evolve into an effective decision support tool in perinatal medicine, offering scientifically grounded, precision strategies for preventing and managing neonatal asphyxia to improve infant health outcomes.\u003c/p\u003e"},{"header":"7. Conclusions and future work","content":"\u003cp\u003eIn this study, we successfully developed and validated a ML-based predictive model for neonatal asphyxia risk assessment. The model integrates multi-source data, including clinical information, biomarkers, and socioeconomic factors, to enhance pre-dictive comprehensiveness and accuracy. By employing advanced data preprocessing techniques, such as outlier detection, missing value imputation, categorical variable encoding, and data standardization, we improved data quality, thereby bolstering model robustness. Six ML algorithms were rigorously evaluated: TabPFN, Bagging, Multilayer Perceptron (MLP), SVM, XGBoost, and RF, with exhaustive hyperparameter tuning. Results demonstrated that the XGBoost model achieved superior performance in predictive accuracy and reliability, while Bagging and RF also exhibited strong effi-cacy. SHAP value analysis provided interpretable insights into model predictions, elu-cidating the influence of key features. Despite these achievements, we acknowledge limitations in data standardization, algorithm optimization, and clinical translation. Future efforts will focus on establishing regional medical data consortiums to enhance data diversity; adopting distributed learning technologies to address data silos; conducting diagnostic-therapeutic hybrid trials to evaluate real-world clinical impact and improving clinical data standardization and healthcare interoperability. This study offers novel perspectives and tools for neonatal asphyxia risk prediction, with significant clinical implications for improving infant outcomes and optimizing healthcare resource allocation. We anticipate these findings will serve as a foundation for further research and practical applications in this field.\u003c/p\u003e\u003cp\u003eIn conclusion, this work provides innovative methodologies and directions for related research, while charting pathways for future advancements. Through continued research and technological innovation, we aim to refine predictive models further, delivering stronger decision-support tools for clinical practice.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003e\u003cimg src=\"data:image/png;base64,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\" width=\"415\" height=\"586\"\u003e\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthor Contributions:\u0026nbsp;\u003c/strong\u003eConceptualization, Sisi Yi and Xinfen XU; methodology, Xieli SHI; software, Xieli SHI; validation, Qiufang LI and Xinfen XU; formal analysis, Xieli SHI; investigation, Sisi Yi; resources, Yanping TENG; data curation, Sisi Yi; writing---original draft preparation, Sisi Yi; writing---review and editing, Qiufang LI; supervision, Qiufang LI and Xinfen XU. All authors have read and agreed to the published version of the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding:\u0026nbsp;\u003c/strong\u003eThis research received no external funding.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthical Approval and Consent to Participate declarations:\u0026nbsp;\u003c/strong\u003eThis study was approved by the Institutional Review Board (IRB) of the Women's Hospital, School of Medicine, Zhejiang University (ethics number: PRO2023-3395). Informed consent was waived due to the patients have been discharged and are difficult to contact. This study will not bring additional adverse effects to the patients. The researchers will strictly adhere to the confidentiality principle. Relevant research information is only accessible to the investigators or the Ethics Committee.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent to Publish declaration:\u0026nbsp;\u003c/strong\u003eThe corresponding author confirms that all authors have read and agreed to the publication of this manuscript. The manuscript does not contain any personal data that could lead to the identification of any individual.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability Statement:\u0026nbsp;\u003c/strong\u003eThe corresponding author will share the data upon request due to legal and ethical reasons..\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflicts of Interest:\u003c/strong\u003e The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements:\u003c/strong\u003e Not applicable\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eOrganization WH. Maternal mortality measurement: guidance to improve national reporting. 2022.\u003c/li\u003e\n\u003cli\u003eFottrell E, Osrin D, et al. Cause-specific neonatal mortality: analysis of 3772 neonatal deaths in Nepal, Bangladesh, Malawi and India. Archives of Disease in Childhood: Fetal and Neonatal Edition. 2015;100(5):F439-F47. \u003c/li\u003e\n\u003cli\u003eChina Maternal and Child Health Association. Report on the development of maternal and child health in China (2019) (Part 1) (in Chinese). Chinese Journal of Women and Child Health. 2019; (5): 1-8.\u003c/li\u003e\n\u003cli\u003eChen ZL, Liu J. Interpretation of \u0026quot;Recommended standards for diagnosis and classification of neonatal asphyxia\u0026quot; (in Chinese). Chinese Journal of Contemporary Pediatrics. 2013.\u003c/li\u003e\n\u003cli\u003eDouglas-Escobar M, Weiss MD. Hypoxic-Ischemic Encephalopathy: A Review for the Clinician. JAMA Pediatrics. 2015; 169(4).\u003c/li\u003e\n\u003cli\u003eTechane MA, Alemu TG, Wubneh CA, et al. The effect of gestational age, low birth weight and parity on birth asphyxia among neonates in sub-Saharan Africa: systematic review and meta-analysis. Italian Journal of Pediatrics. 2021; 2022.\u003c/li\u003e\n\u003cli\u003eWojcieszek AM, Portela A. WHO recommendations on maternal and newborn care for a positive postnatal experience: strengthening the maternal and newborn care continuum. Global Health. 2023.\u003c/li\u003e\n\u003cli\u003ePahnabi A. APGAR Scores in Cesarean Deliveries: Effects of General and Spinal Anesthesia: A Systematic Review. Journal of Pediatrics. 2025.\u003c/li\u003e\n\u003cli\u003eJepson HA, Tichy AM. The Apgar score: evolution, limitations, and scoring guidelines. Birth: Issues in Perinatal Care. 1991.\u003c/li\u003e\n\u003cli\u003eAmerican Academy of Pediatrics. The Apgar score. Pediatrics. 2015.\u003c/li\u003e\n\u003cli\u003eHarrell FE, Harrell J. Cox proportional hazards regression model. In: Regression Modeling Strategies. Springer; 2015.\u003c/li\u003e\n\u003cli\u003eBarnova K, Kahankova R. Artificial intelligence and machine learning in electronic fetal monitoring. Methods in Engineering. 2024.\u003c/li\u003e\n\u003cli\u003eApgar V. A proposal for a new method of evaluation of the newborn infant. Anesthesia \u0026amp; Analgesia. 1953.\u003c/li\u003e\n\u003cli\u003eChen JL, Liu J. Interpretation of the Experts\u0026apos; Consensus on the criteria for the diagnosis and grading of neonatal asphyxia in China. Translational Pediatrics. 2013.\u003c/li\u003e\n\u003cli\u003eZhang YJ. Analysis of clinical risk factors and construction of a prediction model for neonatal asphyxia (in Chinese) [Master\u0026rsquo;s thesis, Anhui Medical University]. 2023.\u003c/li\u003e\n\u003cli\u003eDarsareh F, Ranjbar A, Farashah MV, Mehrnoush V, Shekari M, Jahromi MS. Application of machine learning to identify risk factors of birth asphyxia. BMC Pregnancy and Childbirth. 2023; 23(1): 156.\u003c/li\u003e\n\u003cli\u003eWeiner GM. Updates for the neonatal resuscitation program and resuscitation guidelines. NeoReviews. 2022.\u003c/li\u003e\n\u003cli\u003eHollmann N, Eggensperger K. TabPFN: A transformer that solves small tabular classification problems in a second. arXiv preprint arXiv:2022.\u003c/li\u003e\n\u003cli\u003eCortes C, Vapnik V. Support-vector networks. Machine Learning. 1995; 20(3): 273-297.\u003c/li\u003e\n\u003cli\u003eChen T, Guestrin C. XGBoost: A scalable tree boosting system. In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM; 2016.\u003c/li\u003e\n\u003cli\u003eBreiman L. Random forests. Machine Learning. 2001; 45(1): 5-32.\u003c/li\u003e\n\u003cli\u003eBreiman L. Bagging predictors. Machine Learning. 1996; 24(2): 123-140.\u003c/li\u003e\n\u003cli\u003eBisong E. The multilayer perceptron (MLP). In: Building Machine Learning and Deep Learning Models on Google Cloud Platform. Packt Publishing; 2019.\u003c/li\u003e\n\u003cli\u003eLundberg SM, Lee SI. A Unified Approach to Interpreting Model Predictions. In: Advances in Neural Information Processing Systems. 2017.\u003c/li\u003e\n\u003cli\u003eFawcett T. An introduction to ROC analysis. Pattern Recognition Letters. 2005; 27(8): 861-874.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"italian-journal-of-pediatrics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"itjp","sideBox":"Learn more about [Italian Journal of Pediatrics](http://ijponline.biomedcentral.com)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/ITJP/default.aspx","title":"Italian Journal of Pediatrics","twitterHandle":"@BioMedCentral","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Neonatal asphyxia, Machine learning (ML), Risk prediction, Multi-source data integration, Meta-analysis","lastPublishedDoi":"10.21203/rs.3.rs-7660175/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7660175/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eBackground: Neonatal asphyxia is a leading cause of neonatal mortality and long-term neurodevelopmental impairment, particularly in low-resource settings, underscoring the need for improved predictive tools.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eMethods: This study developed a machine learning (ML) model using multi-source clinical data from 250 neonates (including mild/moderate asphyxia cases) at Zhejiang University Women’s Hospital. After meta-analysis identified 11 key risk factors, data preprocessing involved Random Forest imputation, standardization, and SMOTE for class balancing. Six ML algorithms (XGBoost, RF, Bagging, SVM, MLP, TabPFN) were evaluated, with SHAP analysis for interpretability.\u003c/p\u003e\n\u003cp\u003eResults: XGBoost demonstrated superior performance (recall=0.82, precision=0.82, F1-score=0.82), with nuchal cord, assisted delivery, and prolonged labor emerging as top predictors. Ensemble methods (RF, Bagging) followed, while traditional models (SVM, MLP) and TabPFN showed lower efficacy.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eConclusions: This study presents a validated ML framework for neonatal asphyxia prediction, offering clinical utility for early risk stratification and informed decision-making, particularly in resource-constrained environments.\u003c/p\u003e","manuscriptTitle":"Construction and validation of neonatal asphyxia risk prediction model based on machine learning","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-04 01:18:15","doi":"10.21203/rs.3.rs-7660175/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"","date":"2026-03-09T18:25:27+00:00","index":0,"fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-10-21T16:36:39+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-10-15T22:10:44+00:00","index":"","fulltext":""},{"type":"submitted","content":"Italian Journal of Pediatrics","date":"2025-10-11T09:53:11+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"italian-journal-of-pediatrics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"itjp","sideBox":"Learn more about [Italian Journal of Pediatrics](http://ijponline.biomedcentral.com)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/ITJP/default.aspx","title":"Italian Journal of Pediatrics","twitterHandle":"@BioMedCentral","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"e0a5815f-a5b2-47b7-85dd-ccb10628a986","owner":[],"postedDate":"November 4th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-11-04T01:18:15+00:00","versionOfRecord":[],"versionCreatedAt":"2025-11-04 01:18:15","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7660175","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7660175","identity":"rs-7660175","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}