Abstract
Objective
To evaluate the clinical value of bedside bimodal ultrasound in assessing early cerebral circulation in preterm infants from structural and functional perspectives, and to examine whether maternal recurrent spontaneous abortion (RSA), maternal active immunotherapy status, and ultrasound-derived markers were associated with neonatal cerebral hemodynamic trajectories and clinically significant intraventricular hemorrhage (IVH).
Methods
In this single-center prospective cohort study, 120 preterm infants born to mothers with RSA were enrolled and stratified by maternal active immunotherapy status. Bedside bimodal ultrasound, including transcranial Doppler (TCD) and optic nerve sheath diameter (ONSD) measurement, was performed on postnatal days 1, 3, 5, and 7. The coefficient of variation of the middle cerebral artery pulsatility index (PI-CV) was calculated. Generalized estimating equations (GEE) were used for longitudinal analyses, and multivariable logistic regression was used to evaluate associations with clinically significant IVH (Papile grade ≥II). Because the dataset was a limited tabular cohort, no random training/validation/test split and no class-balancing procedure were used; internal validation was performed using bootstrap resampling. Model performance was assessed using discrimination, calibration, clinical-utility, and likelihood-based model-quality metrics.
Results
PI-CV and ONSD decreased during the first postnatal week in both groups. GEE analysis using an AR(1) working correlation structure showed significant time effects and group × time interactions for PI-CV and ONSD (all P < 0.001), with a faster decline in the active immunotherapy group. ONSDmax and PI-CV at 24 h remained associated with clinically significant IVH after clinical adjustment. The bimodal model had an AUC of 0.759 (95% CI, 0.674–0.844), the clinical-ultrasound model had an AUC of 0.799 (95% CI, 0.721–0.877), and the expanded explanatory model had an AUC of 0.817 (95% CI, 0.743–0.892). For the expanded model, the likelihood-ratio chi-square was 40.915, AIC was 139.301, BIC was 161.601, McFadden pseudo-R2 was 0.249, and the apparent Brier score was 0.174. Bootstrap internal validation using 1,000 resamples yielded optimism-corrected AUC/Brier scores of 0.762/0.190 for the clinical-ultrasound model and 0.774/0.203 for the expanded explanatory model. These results support exploratory risk stratification rather than definitive clinical deployment.
Conclusion
Bedside bimodal ultrasound is a feasible noninvasive approach for dynamic assessment of early cerebral hemodynamics in preterm infants. The combination of ONSDmax and PI-CV may improve exploratory early IVH risk stratification. Maternal immunotherapy status was associated with distinct longitudinal cerebral hemodynamic and ONSD trajectories, reflected by faster modeled declines in PI-CV and ONSD during the first postnatal week. However, given the observational design and associational regression models, these findings do not establish a direct protective effect of maternal immunotherapy on IVH risk. External validation and longer-term outcome studies are required before clinical implementation.
Keywords: active immunotherapy, bedside ultrasound, cerebral hemodynamics, intraventricular hemorrhage, preterm infant, recurrent spontaneous abortion
1. Introduction
Intraventricular hemorrhage (IVH) is one of the most serious complications in preterm infants and is strongly associated with mortality and long-term neurodevelopmental impairment. Its pathogenesis has traditionally been attributed to postnatal cerebral hemodynamic instability and the fragility of the germinal matrix vasculature (1–3). However, increasing evidence suggests that prenatal factors may also play a critical role in shaping neonatal cerebrovascular development and regulatory capacity (4, 5).
Recurrent spontaneous abortion (RSA) is commonly associated with maternal immune dysregulation and chronic inflammatory activation (6–11). These conditions may adversely affect placental perfusion and fetal vascular endothelial development, thereby influencing fetal neurodevelopment. Previous studies have suggested that maternal immune imbalance may increase the vulnerability of the neonatal brain to injury (12, 13), but its specific impact on early postnatal cerebral hemodynamics remains insufficiently understood.
These prenatal considerations intersect with postnatal vulnerability, as early fluid status and cerebral blood-flow modeling have been associated with severe IVH or IVH-risk assessment in extremely preterm infants (14, 15). Meanwhile, immunomodulatory interventions for recurrent pregnancy loss have been increasingly evaluated, but their effects and safety remain heterogeneous across study designs and patient subgroups (16–19).
Bedside ultrasound provides a practical tool for evaluating neonatal cerebral circulation. Transcranial Doppler (TCD) enables the assessment of cerebral blood flow dynamics (20, 21), while optic nerve sheath diameter (ONSD) has been proposed as a noninvasive surrogate marker of intracranial pressure (22, 23). However, a single parameter may not fully capture the complexity of cerebral circulation. A combined assessment integrating functional and structural parameters may improve early risk stratification (24, 25).
Therefore, this study aimed to dynamically evaluate early postnatal cerebral hemodynamic changes in preterm infants born to mothers with RSA using a bimodal ultrasound approach, and to investigate the association between these parameters and the risk of clinically significant IVH.
2. Methods
2.1. Study design and participants
This was a single-center prospective cohort study conducted in the neonatal intensive care unit (NICU) of Cangzhou Central Hospital between 26 November 2025 and 10 April 2026. The cohort database was locked on 10 April 2026, which preceded the manuscript's initial journal receipt date of 12 April 2026; therefore, no post-initial-submission participants or outcome data were added to the analytic cohort. All enrolled infants and all data included in the final analysis were covered by the approved protocol (Approval No. 2025-403-02). The study was approved by the Ethics Committee of Cangzhou Central Hospital, and written informed consent was obtained from the parents or legal guardians of all participating preterm infants. The study complied with the Declaration of Helsinki, Good Clinical Practice principles, and relevant national ethical regulations.
Inclusion criteria were:
-
(1)
preterm infants with gestational age <37 weeks;
-
(2)
maternal history of recurrent spontaneous abortion (RSA);
-
(3)
admission to the NICU within 24 h after birth with completion of initial ultrasound assessment.
Exclusion criteria were:
-
(1)
major congenital malformations or chromosomal abnormalities;
-
(2)
significant congenital heart disease affecting hemodynamics;
-
(3)
severe perinatal asphyxia requiring resuscitation for more than 10 min;
-
(4)
incomplete clinical data or loss to follow-up.
Infants were stratified into an active immunotherapy group and a control group according to maternal treatment status before pregnancy.
2.2. Ultrasound assessment
Bedside bimodal ultrasound was performed according to standardized neonatal neurosonography and transcranial Doppler practice principles. All examinations were conducted by trained sonographers who were blinded to the final IVH classification. Device presets, depth, gain, sample volume, and insonation windows were kept constant whenever clinically feasible.
TCD was performed through the anterior fontanelle to obtain middle cerebral artery flow signals. For each examination, stable Doppler spectra from at least five consecutive cardiac cycles were recorded. The pulsatility index (PI) was calculated for each accepted cycle, and PI-CV was defined as the standard deviation divided by the mean PI multiplied by 100%, thereby quantifying short-term variability of cerebral vascular pulsatility (Figure 1 A, B).
Figure 1.
Bimodal ultrasound assessment and representative imaging findings. (A) Transfontanelle color Doppler ultrasound localization. The arrow indicates the M1 segment of the middle cerebral artery (MCA), which serves as the target vessel for functional assessment of cerebral hemodynamics. (B) Pulsed-wave Doppler spectrum of the middle cerebral artery. A stable cardiac cycle is demonstrated, allowing calculation of the pulsatility index (PI) and its coefficient of variation (PI-CV), which are used to quantify the resilience of cerebral autoregulation. (C) Measurement of optic nerve sheath diameter (ONSD). According to the CLOSED protocol, the diameter was measured 3 mm posterior to the globe (calipers shown), serving as a noninvasive surrogate marker of intracranial pressure. (D) Clinically significant IVH outcome phenotype. The neonatal sagittal section ultrasound showed grade II intraventricular hemorrhage, hyperechoic mass in the ventricle (indicated by asterisk) and normal ventricular size.
ONSD was measured using a transorbital approach in accordance with the CLOSED protocol and recent ONSD point-of-care ultrasonography quality criteria. The diameter was measured 3 mm posterior to the globe, avoiding direct globe pressure and excessive acoustic exposure. Bilateral measurements were repeated three times when image quality allowed; the mean bilateral value and maximum value (ONSDmax) were recorded (Figure 1C).
Ultrasound examinations were performed on postnatal days 1, 3, 5, and 7. For the IVH prediction model, ONSDmax was strictly defined as the maximum bilateral value obtained within the first 24 h after birth (postnatal day 1), rather than the maximum value selected across the entire first postnatal week, to preserve its interpretation as an early bedside predictor.
2.2.1. Outcome adjudication and expert panel
The expert panel supporting ultrasound image review, clinical annotation, and IVH outcome adjudication consisted of five specialists: one attending neonatologist from the NICU with 7 years of clinical experience, one attending sonographer with 25 years of ultrasound experience, one chief sonographer with 26 years of ultrasound experience, and two associate chief sonographers with 15 years of ultrasound experience each. The mean professional experience of the panel was 17.6 ± 7.9 years. This information is reported to characterize the clinical expertise underlying image annotation and outcome review.
2.3. Variables and definitions
The primary variables included PI-CV and ONSD, as well as the maximum ONSD (ONSDmax).
The primary outcome was clinically significant intraventricular hemorrhage (IVH), defined as Papile grade ≥II (Figure 1D).
Covariates considered for longitudinal and prediction analyses included gestational age, birth weight, mechanical ventilation, PaCO2 or PaCO2 variability, infection status, and maternal immunotherapy status. Variables retained in each model are reported explicitly below.
2.4. Sample size, data partitioning, and class distribution
The analytic cohort consisted of 120 infants, including 52 infants with clinically significant IVH events (43.3%) and 68 without events (56.7%). The endpoint distribution was therefore not severely imbalanced. The prediction component used structured tabular variables derived from bedside ultrasound and clinical data rather than an image-based machine-learning classifier. Given the limited sample size and event count, the cohort was not randomly divided into separate training, validation, and test sets, because such partitioning would have reduced the effective number of events for model estimation and produced unstable estimates. Instead, the full cohort was used for model fitting, and internal validation was performed using 1,000 bootstrap resamples to estimate optimism-corrected discrimination and calibration-related prediction error, reported as optimism-corrected AUC and Brier score. No undersampling, oversampling, SMOTE, class weighting, or other class-balancing strategy was applied. The prediction analysis should therefore be interpreted as exploratory internal model development and validation, not as a ready-to-use clinical decision rule.
2.5. Statistical analysis
Statistical analyses were performed using SPSS 29.0 and R/Python statistical software. The observational cohort component was cross-checked against STROBE reporting principles (26), and the prediction component followed TRIPOD + AI and PROBAST + AI principles where applicable for regression- or machine-learning-based prediction models (27, 28).
Continuous variables were expressed as mean ± standard deviation or median (interquartile range), depending on distribution. Normality was assessed using the Shapiro–Wilk test and inspection of Q–Q plots. Normally distributed continuous variables were compared using Student's t-test, non-normally distributed continuous variables and ordinal variables such as Apgar scores were compared using the Mann–Whitney U-test, and categorical variables were compared using the χ2 test or Fisher's exact test, as appropriate. The Z statistics reported for nonparametric comparisons represent standardized Mann–Whitney U-test statistics and should not be interpreted as parametric Z-tests.
Generalized estimating equations (GEE) with robust standard errors and a first-order autoregressive [AR(1)] working correlation structure were used to analyze repeated PI-CV and ONSD measurements. Time was coded as postnatal day (D1, D3, D5, and D7). Models included group, time, group × time interaction, gestational age, birth weight, mechanical ventilation, and simultaneously measured PaCO2. A sensitivity analysis using cubic spline terms for time and mechanical ventilation × spline-time interaction was performed to assess potential nonlinearity.
For IVH prediction, candidate variables included ONSDmax, PI-CV at 24 h, gestational age, birth weight, mechanical ventilation, PaCO2 variability, and maternal immunotherapy status. Variables were retained on the basis of prespecified clinical relevance, univariable screening (P < 0.10), collinearity diagnostics, and the available number of events. Model 1 was defined as the bedside clinical-ultrasound model and excluded maternal immunotherapy; Model 2 was defined as an expanded explanatory model that included maternal immunotherapy as a contextual upstream exposure. No random training/validation/test split and no class-balancing procedure were applied. Model quality was reported using convergence status, log-likelihood, likelihood-ratio chi-square, AIC, BIC, McFadden, Cox–Snell and Nagelkerke pseudo-R2, Brier score, Hosmer-Lemeshow calibration test, variance inflation factors, and bootstrap optimism correction using 1,000 resamples for both Model 1 and Model 2. Optimism-corrected AUC was used as the discrimination metric, and optimism-corrected Brier score was used as the calibration-related prediction-error metric. AUCs were compared using DeLong tests, and clinical utility was evaluated using decision curve analysis.
3. Results
3.1. Baseline characteristics
A total of 120 preterm infants born to mothers with recurrent spontaneous abortion (RSA) were included in this study, with 60 infants in the active immunotherapy group and 60 in the control group.
There were no significant differences between the two groups in maternal characteristics and perinatal factors, including maternal age, incidence of clinical chorioamnionitis, and antenatal corticosteroid use (all P > 0.05).
Regarding neonatal baseline demographic characteristics, no significant differences were observed in gestational age, birth weight, sex distribution, or mode of delivery between the two groups (all P > 0.05).
In terms of early postnatal clinical status, there were no significant differences in mean arterial pressure (MAP) at admission, incidence of patent ductus arteriosus (PDA), use of mechanical ventilation, or vasoactive drugs (all P > 0.05). In addition, the 5 min Apgar score was comparable between the two groups (P = 0.094).
No significant differences were found in baseline coagulation parameters, including platelet count (PLT) and activated partial thromboplastin time (APTT) (all P > 0.05).
These findings indicate that the two groups were comparable at baseline, providing a reliable basis for subsequent analyses (Table 1).
Table 1.
Baseline characteristics of preterm infants stratified by maternal immunotherapy.
| Variable | Active immunotherapy (n = 60) | Control (n = 60) | Statistic | P value |
|---|---|---|---|---|
| Sex, n (%) | χ2 = 0.541 | 0.462 | ||
| Male | 36 (60.0) | 31 (51.7) | ||
| Female | 24 (40.0) | 29 (48.3) | ||
| Maternal age, years | 30 (26–33) | 29 (27–31) | MWU Z = 0.262 | 0.795 |
| Apgar score (1 min) | 7 (6–7) | 6 (6–7) | MWU Z = 1.564 | 0.097 |
| Apgar score (5 min) | 8 (7–8) | 7 (7–8) | MWU Z = 1.580 | 0.094 |
| Vasoactive drug use, n (%) | 11 (18.3) | 17 (28.3) | χ2 = 1.165 | 0.281 |
| Clinical chorioamnionitis, n (%) | 5 (8.3) | 8 (13.3) | χ2 = 0.345 | 0.557 |
| Patent ductus arteriosus, n (%) | 22 (36.7) | 24 (40.0) | χ2 = 0.035 | 0.851 |
| Antenatal steroid use, n (%) | 33 (55.0) | 34 (56.7) | χ2 = 0.000 | 1.000 |
| Mean arterial pressure (mmHg) | 35 (33–38) | 34 (32–36) | MWU Z = 1.044 | 0.296 |
| Birth weight (g) | 1,255 (1,198–1,330) | 1,240 (1,165–1,325) | MWU Z = 0.541 | 0.590 |
| Mechanical ventilation, n (%) | 25 (41.7) | 29 (48.3) | χ2 = 0.303 | 0.582 |
| Gestational age (weeks) | 30.9 ± 1.4 | 30.6 ± 1.6 | t = 1.073 | 0.286 |
| Mode of delivery, n (%) | χ2 = 0.000 | 1.000 | ||
| Cesarean section | 37 (61.7) | 38 (63.3) | ||
| Vaginal delivery | 23 (38.3) | 22 (36.7) | ||
| Platelet count (×109/L) | 211 ± 27 | 202 ± 26 | t = 1.885 | 0.062 |
| APTT (s) | 56 (54–60) | 57 (54–59) | MWU Z = −0.276 | 0.784 |
MAP, mean arterial pressure; PLT, platelet count; APTT, activated partial thromboplastin time.
Data are presented as mean ± standard deviation, median (interquartile range), or number (percentage), as appropriate.
Comparisons between groups were performed using Student's t-test, Mann–Whitney U-test, χ2 test, or Fisher's exact test, as appropriate. Apgar scores were treated as ordinal variables and compared using the Mann–Whitney U-test.
3.2. Longitudinal changes in cerebral hemodynamic parameters
During serial assessments on postnatal days 1, 3, 5, and 7, both PI-CV and ONSD showed a decreasing trend in both groups.
Generalized estimating equations (GEE) analysis demonstrated a significant time effect for both PI-CV and ONSD (both P < 0.001), indicating a progressive decline in cerebral hemodynamic parameters over time.
A significant group effect was also observed, with lower overall levels of PI-CV and ONSD in the active immunotherapy group compared with the control group (both P < 0.001). In addition, a significant interaction between time and group was identified (interaction effect, P < 0.001), suggesting different temporal patterns between the two groups.
Further comparisons showed that the decline in PI-CV and ONSD was more pronounced in the active immunotherapy group.
These findings indicate dynamic changes in cerebral hemodynamics during early postnatal life, with distinct patterns associated with maternal immunotherapy status (Table 2, Figure 2).
Table 2.
Generalized estimating equation (GEE) analysis of longitudinal changes in PI-CV and ONSD in preterm infants.
| Variable | β | Robust SE | Wald χ2 | P value | 95% CI | Interpretation |
|---|---|---|---|---|---|---|
| PI-CV | ||||||
| Active immunotherapy group | −2.569 | 0.390 | 43.363 | <0.001 | −3.334 to −1.804 | Lower PI-CV level |
| Time, per postnatal day | −1.149 | 0.047 | 589.2 | <0.001 | −1.242 to −1.056 | Declining trajectory |
| Group × time | −0.225 | 0.057 | 15.468 | <0.001 | −0.336 to −0.113 | Faster decline |
| Gestational age, weeks | −0.153 | 0.071 | 4.657 | 0.031 | −0.293 to −0.014 | Lower PI-CV |
| Birth weight, per 100 g | 0.157 | 0.074 | 4.477 | 0.034 | 0.012 to 0.303 | Higher PI-CV |
| Mechanical ventilation | −0.641 | 0.236 | 7.371 | 0.007 | −1.103 to −0.178 | Lower short-term variability |
| PaCO2, mmHg | 0.035 | 0.019 | 3.341 | 0.067 | −0.003 to 0.073 | Borderline effect |
| ONSD | ||||||
| Active immunotherapy group | −1.137 | 0.215 | 27.96 | <0.001 | −1.559 to −0.715 | Lower ONSD level |
| Time, per postnatal day | −0.951 | 0.032 | 882.5 | <0.001 | −1.014 to −0.888 | Declining trajectory |
| Group × time | −0.143 | 0.041 | 12.16 | <0.001 | −0.224 to −0.062 | Faster decline |
| Gestational age, weeks | −0.082 | 0.061 | 1.80 | 0.180 | −0.202 to 0.038 | NS |
| Birth weight, per 100 g | 0.038 | 0.062 | 0.38 | 0.539 | −0.084 to 0.160 | NS |
| Mechanical ventilation | −0.411 | 0.226 | 3.31 | 0.069 | −0.854 to 0.032 | Borderline effect |
| PaCO2, mmHg | 0.008 | 0.011 | 0.53 | 0.467 | −0.014 to 0.030 | NS |
GEE models were fitted with robust standard errors and a first-order autoregressive [AR(1)] working correlation structure to account for within-infant repeated measurements across D1, D3, D5, and D7. The models included group, postnatal time, group × time interaction, gestational age, birth weight, mechanical ventilation, and simultaneously measured PaCO2; cubic spline sensitivity analyses were additionally performed to examine potential nonlinear time effects.
NS, not significant; PI-CV, coefficient of variation of pulsatility index; ONSD, optic nerve sheath diameter; PaCO2, partial pressure of arterial carbon dioxide.
Figure 2.
Longitudinal trajectories of cerebral hemodynamics in preterm infants. Lines show mean values and shaded areas indicate 95% confidence intervals. Postnatal time was displayed as D1, D3, D5, and D7 rather than as a continuous 120-day axis. Longitudinal changes were analyzed using GEE with robust standard errors and AR(1) working correlation. Significant group × time interactions were observed for both PI-CV and ONSD (P < 0.001).
3.3. Factors associated with clinically significant IVH
Multivariable logistic regression was performed with clinically significant IVH (Papile grade ≥II) as the dependent variable. Candidate variables and reasons for inclusion were reported explicitly: ONSDmax and PI-CV were prespecified ultrasound markers; gestational age, birth weight, mechanical ventilation, and PaCO2 variability were clinically relevant covariates; maternal immunotherapy status was evaluated only as a contextual upstream exposure rather than as a directly deployable bedside predictor.
In the clinical-ultrasound model excluding maternal treatment status, ONSDmax (OR = 1.201 per 0.1 mm increase, 95% CI: 1.074–1.342, P = 0.001), PI-CV (OR = 1.283, 95% CI: 1.046–1.575, P = 0.017), mechanical ventilation (OR = 2.483, 95% CI: 1.023–6.026, P = 0.044), and lower gestational age (OR = 0.736 per week, 95% CI: 0.549–0.988, P = 0.041) remained associated with clinically significant IVH after adjustment. These estimates should be interpreted as adjusted associations, not as proof of causal effects.
The expanded explanatory model including maternal immunotherapy status achieved slightly higher apparent discrimination but yielded an imprecise treatment-status coefficient after adjustment for ONSDmax and PI-CV. Given the observational design, limited sample size, and wide confidence interval, this coefficient was not interpreted as evidence of either a harmful or protective independent causal effect. Maternal immunotherapy was therefore discussed only as a contextual exposure associated with longitudinal hemodynamic trajectories.
Collinearity was low for all predictors (all variance inflation factors <2), and the expanded model converged appropriately.
These findings support the association of ONSDmax and PI-CV with clinically significant IVH, while estimates for mechanical ventilation and maternal immunotherapy require cautious interpretation because of their wide confidence intervals and the exploratory model setting (Table 3). The change in estimates after adding maternal immunotherapy to Model 2 indicates sensitivity to model specification, but this pattern does not establish mediation, suppression, or causality (Table 4).
Table 3.
Multivariable logistic regression analysis of factors associated with clinically significant IVH.
| Domain | Predictor | Model 1 OR (95% CI) | P value | Model 2 OR (95% CI) | P value |
|---|---|---|---|---|---|
| Ultrasound-derived | ONSDmax, per 0.1-mm increase | 1.201 (1.074–1.342) | 0.001 | 1.277 (1.118–1.459) | <0.001 |
| Ultrasound-derived | PI-CV at 24 h | 1.283 (1.046–1.575) | 0.017 | 1.451 (1.129–1.865) | 0.004 |
| Clinical | Mechanical ventilation, yes vs. no | 2.483 (1.023–6.026) | 0.044 | 2.781 (1.117–6.927) | 0.028 |
| Clinical | Gestational age, per week | 0.736 (0.549–0.988) | 0.041 | 0.730 (0.544–0.980) | 0.036 |
| Clinical | Birth weight, per 100 g | 0.944 (0.687–1.298) | 0.723 | 0.914 (0.661–1.265) | 0.589 |
| Clinical | PaCO2 variability | 0.983 (0.877–1.102) | 0.765 | 0.995 (0.887–1.116) | 0.929 |
| Contextual exposure | Maternal immunotherapy, yes vs. no | — | — | 3.413 (1.006–11.578) | 0.049 |
Model 1 was the bedside clinical–ultrasound prediction model and excluded maternal immunotherapy. Model 2 was an expanded explanatory model that included maternal immunotherapy as a contextual upstream exposure. The wide confidence intervals for mechanical ventilation and maternal immunotherapy indicate limited precision; therefore, these estimates should be interpreted as exploratory associations rather than definitive effects.
IVH, intraventricular hemorrhage; ONSD, optic nerve sheath diameter; PI-CV, coefficient of variation of pulsatility index; OR, odds ratio; CI, confidence interval. The dependent variable was clinically significant IVH (Papile grade ≥II). A two-sided P value < 0.05 was considered statistically significant. PI-CV was measured at 24 h after birth (postnatal day 1).
Table 4.
Logistic regression model-quality metrics and internal validation results.
| Metric | Clinical–ultrasound model (Model 1) | Expanded explanatory model (Model 2) |
|---|---|---|
| Number of infants | 120 | 120 |
| Events/non-events | 52/68 | 52/68 |
| Candidate predictors retained | 6 | 7 |
| Convergence | Yes | Yes |
| Maximum VIF | <2 | <2 |
| Log-likelihood | −63.832 | −61.651 |
| Null log-likelihood | −82.108 | −82.108 |
| −2 log-likelihood | 127.665 | 123.301 |
| Likelihood-ratio χ2 | 36.551 (df = 6) | 40.915 (df = 7) |
| Likelihood-ratio test P value | <0.001 | <0.001 |
| AIC | 141.665 | 139.301 |
| BIC | 161.177 | 161.601 |
| McFadden pseudo-R2 | 0.223 | 0.249 |
| Cox–Snell pseudo-R2 | 0.263 | 0.289 |
| Nagelkerke pseudo-R2 | 0.352 | 0.388 |
| Apparent Brier score | 0.181 | 0.174 |
| Hosmer–Lemeshow test | χ2 = 6.460, df = 8, P = 0.596 | χ2 = 7.939, df = 8, P = 0.439 |
| Bootstrap resamples | 1,000 | 1,000 |
| Bootstrap optimism-corrected AUC | 0.762 | 0.774 |
| Bootstrap optimism-corrected Brier score | 0.190 | 0.203 |
Model-quality metrics were calculated from the final fitted logistic regression models using the same variables and scaling as Table 3. The apparent Brier score was calculated as the mean squared difference between observed IVH status and model-predicted probability. Hosmer-Lemeshow calibration testing used 10 risk groups. AIC, Akaike information criterion; BIC, Bayesian information criterion; VIF, variance inflation factor; AUC, area under the receiver operating characteristic curve.
3.4. Model performance and validation
3.4.1. Discrimination and internal validation
ROC analysis showed moderate discrimination for ONSDmax (AUC = 0.733, 95% CI: 0.644–0.822) and PI-CV (AUC = 0.714, 95% CI: 0.621–0.807). At the Youden cutoff, ONSDmax had a sensitivity of 65.4% and specificity of 72.1%; this threshold should therefore be treated as an exploratory risk-stratification point rather than a stand-alone diagnostic cutoff.
The bimodal model combining ONSDmax and PI-CV improved apparent discrimination (AUC = 0.759, 95% CI: 0.674–0.844). The clinical-ultrasound model excluding treatment status achieved an AUC of 0.799 (95% CI: 0.721–0.877), and the expanded explanatory model (Table 5, Figure 3) achieved an AUC of 0.817 (95% CI: 0.743–0.892). In apparent DeLong comparisons, the expanded model showed higher discrimination than the bimodal model (P = 0.048), ONSDmax alone (P = 0.016), and PI-CV alone (P = 0.021). Bootstrap internal validation (1,000 resamples) was performed for both models. For Model 1, the optimism-corrected AUC was 0.762 and the optimism-corrected Brier score was 0.190. For Model 2, the optimism-corrected AUC was 0.774 and the optimism-corrected Brier score was 0.203. These internally validated estimates indicate modest optimism in the apparent performance and reinforce that the models should be interpreted as exploratory risk-stratification tools requiring external validation.
Table 5.
Discriminative performance of different models for predicting clinically significant IVH.
| Model | AUC | 95% CI | Sensitivity | Specificity |
|---|---|---|---|---|
| ONSDmax | 0.733 | 0.644–0.822 | 0.654 | 0.721 |
| PI-CV | 0.714 | 0.621–0.807 | 0.635 | 0.706 |
| Bimodal model | 0.759 | 0.674–0.844 | 0.885 | 0.559 |
| Clinical-ultrasound model | 0.799 | 0.721–0.877 | 0.596 | 0.853 |
| Expanded model | 0.817 | 0.743–0.892 | 0.750 | 0.794 |
IVH, intraventricular hemorrhage; AUC, area under the receiver operating characteristic curve; CI, confidence interval; ONSDmax, maximum optic nerve sheath diameter; PI-CV, coefficient of variation of pulsatility index.
Clinically significant IVH was defined as Papile grade ≥ II.
The bimodal model included ONSDmax and PI-CV measured at 24 h after birth.
The expanded model incorporated ultrasound parameters, clinical covariates, and maternal immunotherapy status.
Sensitivity and specificity were calculated at the optimal cutoff determined by the Youden index.
Figure 3.
Receiver operating characteristic (ROC) curves for prediction of intraventricular hemorrhage. The ROC curve compares the bimodal ultrasound model and the expanded explanatory model. The expanded model shows higher apparent discriminative performance in this cohort. The diagonal dashed line represents the chance level.
3.4.2. Calibration
Calibration plots showed acceptable apparent agreement between predicted and observed probabilities. The Hosmer-Lemeshow test did not indicate significant lack of fit for either the clinical-ultrasound model (chi-square = 6.460, df = 8, P = 0.596) or the expanded explanatory model (chi-square = 7.939, df = 8, P = 0.439), suggesting adequate apparent calibration within this exploratory cohort (Figure 4).
Figure 4.
Calibration curve of the expanded explanatory model. Calibration plot comparing predicted probabilities with observed outcomes. The dashed diagonal line represents perfect calibration. The plot suggests acceptable apparent agreement between predicted and observed risks in the development cohort.
3.4.3. Decision curve analysis
Decision curve analysis (DCA) suggested that both the bimodal model and the expanded explanatory model provided higher net benefit than treat-all or treat-none strategies across a range of clinically plausible threshold probabilities (Figure 5).
Figure 5.
Decision curve analysis (DCA) of predictive models. Decision curve analysis showing net benefit across threshold probabilities. The bimodal and expanded explanatory models provided greater apparent clinical net benefit within clinically relevant threshold ranges. The dashed line represents the treat-all strategy, and the horizontal line represents the treat-none strategy.
Taken together, the discrimination, calibration, Brier score, and decision-curve findings indicate that ultrasound-based models incorporating ONSDmax and PI-CV provide clinically interpretable exploratory risk-stratification information. However, the modest sample size, optimism-corrected performance, and absence of external validation mean that the model should not be used for routine clinical decision-making before independent validation.
4. Discussion
4.1. Principal findings
In this prospective cohort study, we dynamically evaluated early postnatal cerebral hemodynamic changes in preterm infants born to mothers with recurrent spontaneous abortion (RSA) using a bimodal ultrasound approach. Both PI-CV and ONSD showed a decreasing trend during the first postnatal week, suggesting gradual stabilization of cerebral circulation, which is consistent with previous observations of postnatal hemodynamic adaptation in preterm infants (29, 30).
Compared with the control group, infants in the active immunotherapy group exhibited a more pronounced decline in PI-CV and ONSD, suggesting a faster pattern of early hemodynamic stabilization. In addition, ONSDmax and PI-CV measured at 24 h after birth remained associated with clinically significant IVH after adjustment, and their combination improved apparent predictive performance, consistent with prior studies linking cerebral blood flow variability and structural indicators with IVH risk (31, 32).
These findings suggest that combined structural and functional bedside ultrasound assessment may enhance early, noninvasive IVH risk stratification when interpreted together with clinical context.
4.2. Association between maternal immunotherapy and neonatal cerebral hemodynamics
RSA is commonly associated with maternal immune dysregulation and chronic inflammation, which may influence placental function and fetal vascular development (6, 7). Previous studies have suggested that adverse intrauterine environments may increase the susceptibility of neonates to brain injury (10, 12).
In this study, maternal active immunotherapy was associated with faster longitudinal declines in PI-CV and ONSD. In contrast, the treatment-status coefficient in the expanded IVH model was imprecise after adjustment for proximal ultrasound markers. This pattern is best interpreted as uncertainty related to model specification and residual confounding within an observational dataset, not as evidence that maternal immunotherapy directly increases or decreases IVH risk. Accordingly, we avoid describing maternal immunotherapy as an independent direct protective factor for IVH.
Because this was an observational cohort study, causality cannot be established from logistic regression alone. We did not perform formal causal modeling, Granger causality, transfer entropy, or mediation analysis; such analyses would require larger datasets, clearer temporal ordering of exposures and candidate mediators, and prespecified causal assumptions. The biological mechanisms therefore remain uncertain and may involve immune modulation, placental perfusion, fetal vascular development, and postnatal cerebral autoregulatory adaptation.
4.3. Clinical relevance of PI-CV and ONSD
In this study, PI-CV and ONSD were used to reflect functional and structural aspects of cerebral circulation, respectively. PI-CV represents the variability of cerebral blood flow and may reflect the stability of cerebral autoregulation, while ONSD serves as a surrogate marker of intracranial pressure, as supported by previous neurocritical care studies (22, 23).
Both parameters were associated with IVH, suggesting that functional instability and structural changes may jointly identify infants with greater vulnerability to early brain injury, consistent with current understanding of impaired cerebral autoregulation in preterm infants (29).
The combined model showed better apparent discrimination than either single marker, supporting a pragmatic multimodal approach to early bedside risk stratification rather than reliance on a single cutoff (25).
4.4. Interpretation of mechanical ventilation and nonlinear sensitivity findings
Mechanical ventilation showed different associations across longitudinal and binary-outcome models. In GEE analysis, ventilated infants had lower short-term PI-CV after adjustment for PaCO2, whereas mechanical ventilation was associated with higher odds of clinically significant IVH in the logistic model.
To address the possibility of nonlinearity, cubic spline terms for postnatal time and mechanical ventilation × spline-time interactions were added as sensitivity analyses. These nonlinear interaction terms were not statistically significant for PI-CV or ONSD, and the group × time findings remained directionally unchanged. The apparent discrepancy is therefore more plausibly explained by differences in analytic target: short-term physiologic stabilization under ventilatory support in the repeated-measures model vs. overall illness severity in the IVH outcome model.
Therefore, PI-CV should be interpreted cautiously in ventilated infants, and structural indicators such as ONSD should be considered alongside clinical severity rather than in isolation.
4.5. Clinical implications
Bimodal ultrasound is noninvasive, repeatable, and feasible in neonatal intensive care settings, and is increasingly recognized as an important bedside neuromonitoring tool (21, 22).
Measurements obtained within the first 24 h after birth may provide clinically useful information for early risk stratification. The combined use of PI-CV and ONSD may help identify infants who warrant closer neurosonographic surveillance and individualized neuroprotective attention, although the present findings do not justify treatment decisions based solely on model-predicted risk (33, 34).
The bimodal model prioritized sensitivity (0.885) at the cost of lower specificity (0.559), which may be clinically acceptable for an early screening tool where missed high-risk infants are a major concern. By contrast, the clinical-ultrasound and expanded models showed a more balanced sensitivity-specificity profile. Decision curve analysis suggested greater apparent net benefit than treat-all or treat-none strategies across clinically plausible threshold ranges. Given the event rate of 43.3% and the single-center setting, this finding should be interpreted as exploratory decision-support evidence rather than a ready-to-use clinical rule.
4.6. Limitations
This study has several limitations. First, it was a single-center prospective cohort study with 120 infants and 52 events; no independent training, validation, or test set was available, and bootstrap validation cannot replace external validation. Second, although the endpoint distribution was not severely imbalanced, the limited sample size produced wide confidence intervals for several predictors, particularly mechanical ventilation and maternal immunotherapy. Third, maternal immunotherapy was observational rather than randomized, leaving room for residual confounding and indication bias; no causal modeling was performed. Fourth, biological markers, placental pathology, and direct measures of intracranial pressure or cerebral autoregulation were not included, limiting mechanistic inference. Fifth, although the expert panel was characterized in terms of composition and experience (17.6 ± 7.9 years), single-center image annotation and outcome review may still limit generalizability. Finally, long-term neurodevelopmental outcomes were not assessed. These limitations are consistent with domains emphasized in PROBAST + AI for prediction-model risk-of-bias and applicability assessment (28). The need for external validation is also consistent with broader perinatal risk-prediction literature (35).
4.7. Future directions
Future studies should include multicenter cohorts, prespecified external validation, clearly separated training/validation/test datasets when sample size permits, biological and placental markers, standardized ultrasound acquisition audits, transparent expert-annotation reporting, and long-term neurodevelopmental follow-up. Larger datasets would also permit prespecified causal or mediation analyses to evaluate whether maternal immune modulation, cerebral hemodynamic markers, and IVH risk are connected through biologically plausible pathways.
5. Conclusion
This study shows that PI-CV and ONSD change dynamically during the first postnatal week and are associated with clinically significant IVH in preterm infants born to mothers with RSA. A bimodal ultrasound approach combining structural and functional information improved apparent early risk stratification compared with single parameters.
Maternal immunotherapy status was associated with distinct longitudinal cerebral hemodynamic and ONSD trajectories, reflected by faster modeled declines in PI-CV and ONSD during the first postnatal week. However, given the observational design and associational regression models, these findings do not establish a direct protective effect of maternal immunotherapy on IVH risk. The findings should therefore be interpreted as hypothesis-generating.
The proposed ultrasound-based model may serve as a practical exploratory tool for early identification of high-risk infants, pending external validation, calibration assessment in independent cohorts, and long-term outcome evaluation.
Further multicenter studies are warranted to validate these findings and to explore the underlying mechanisms.
Funding Statement
The author(s) declared that financial support was received for this work. This research was supported by the Fundamental Research Project of Cangzhou (Science and Technology Bureau of Cangzhou) (20251102041).
Footnotes
Edited by: José Javier Reyes-Lagos, Cinvestav Unidad Zacatenco, Mexico
Reviewed by: Omar Piña-Ramírez, Autonomous Metropolitan University, Mexico
Aldo Rodrigo Mejía Rodríguez, Autonomous University of San Luis Potosí, Mexico
Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.
Ethics statement
The studies involving humans were approved by Ethics Committee of Cangzhou Central Hospital. The studies were conducted in accordance with the local legislation and institutional requirements. Written informed consent for participation in this study was provided by the participants' legal guardians/next of kin. Written informed consent was obtained from the individual(s), and minor(s)' legal guardian/next of kin, for the publication of any potentially identifiable images or data included in this article.
Author contributions
NW: Data curation, Investigation, Methodology, Visualization, Writing – original draft. RC: Conceptualization, Data curation, Investigation, Supervision, Writing – review & editing.
Conflict of interest
The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declared that generative AI was not used in the creation of this manuscript.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
References
- 1.Christensen R, Cizmeci MN, de Vries LS. Preterm hemorrhagic brain injury: recent advances on evaluation and management. Clin Perinatol. (2025) 52(2):307–19. 10.1016/j.clp.2025.02.007 [DOI] [PubMed] [Google Scholar]
- 2.Ducatez F, Tebani A, Abily-Donval L, Snanoudj S, Pilon C, Plichet T, et al. New insights and potential biomarkers for intraventricular hemorrhage in extremely premature infants: a case-control study. Pediatr Res. (2024) 96(2):395–401. 10.1038/s41390-024-03111-9 [DOI] [PubMed] [Google Scholar]
- 3.Razak A, Patel W, Durrani NUR, Pullattayil AK. Interventions to reduce severe brain injury risk in preterm neonates: a systematic review and meta-analysis. JAMA Netw Open. (2023) 6(4):e237473. 10.1001/jamanetworkopen.2023.7473 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Kwon HK, Choi GB, Huh JR. Maternal inflammation and its ramifications on fetal neurodevelopment. Trends Immunol. (2022) 43(3):230–44. 10.1016/j.it.2022.01.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Lammertink F, Benders MJNL, Hermans EJ, Tataranno ML, Dudink J, Vinkers CH, et al. Vulnerability of the neonatal connectome following postnatal stress. J Neurosci. (2022) 42(48):8948–59. 10.1523/JNEUROSCI.0176-22.2022 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Li Q-H, Zhao Q-Y, Yang W-J, Jiang A-F, Ren C-E, Meng Y-H. Beyond immune balance: the pivotal role of decidual regulatory T cells in unexplained recurrent spontaneous abortion. J Inflamm Res. (2024) 17:2697–710. 10.2147/JIR.S459263 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Garmendia JV, De Sanctis CV, Hajdúch M, De Sanctis JB. Exploring the immunological aspects and treatments of recurrent pregnancy loss and recurrent implantation failure. Int J Mol Sci. (2025) 26(3):1295. 10.3390/ijms26031295 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Bender Atik R, Christiansen OB, Elson J, Kolte AM, Lewis S, Middeldorp S, et al. ESHRE Guideline: recurrent pregnancy loss—an update in 2022. Hum Reprod Open. (2023) 2023(1):hoad002. 10.1093/hropen/hoad002 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Obstetrics Subgroup. Chinese Society of obstetrics and gynecology; Chinese medical association. Chinese expert consensus on diagnosis and management of recurrent spontaneous abortion (2022). Zhonghua Fu Chan Ke Za Zhi. (2022) 57(9):653–67. 10.3760/cma.j.cn112141-20220421-00259 [DOI] [PubMed] [Google Scholar]
- 10.Osman HC, Moreno R, Rose D, Rowland ME, Ciernia AV, Ashwood P. Impact of maternal immune activation and sex on placental and fetal brain cytokine and gene expression profiles. J Neuroinflammation. (2024) 21(1):118. 10.1186/s12974-024-03106-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Abad C, Farina M, Damiano AE, Marín R. Maternal–fetal interface: new insight in placenta research. Front Endocrinol. (2025) 16:1642059. 10.3389/fendo.2025.1642059 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Wang H, Li F, Miao M, Yu Y, Ji H, Liu H, et al. Maternal spontaneous abortion and the risk of ADHD in offspring. Hum Reprod. (2020) 35(5):1211–21. 10.1093/humrep/deaa035 [DOI] [PubMed] [Google Scholar]
- 13.Campbell KSJ, Brain U, Hanley GE, Oberlander TF, Lim KI. Maternal depressed mood and antidepressant exposure and neonatal neurobehaviour. Early Hum Dev. (2024) 198:106129. 10.1016/j.earlhumdev.2024.106129 [DOI] [PubMed] [Google Scholar]
- 14.Weaver LJ, Gentle SJ, Nakhmani A, Rahman F, Ambalavanan N, Shukla VV, et al. Early fluid status and severe IVH or death in extremely preterm infants. Pediatr Nephrol. (2026) 41(1):239–47. 10.1007/s00467-025-06962-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Sidorenko I, Brodkorb S, Felderhoff-Müser U, Rieger-Fackeldey E, Krüger M, Feddahi N, et al. Assessment of IVH risk using simulated cerebral blood flow. Front Neurol. (2024) 15:1465440. 10.3389/fneur.2024.1465440 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Xin W, Mu F, Wang K, Wang F. Intravenous immunoglobulin improves live birth outcomes in RPL. Front Immunol. (2026) 17:1689166. 10.3389/fimmu.2026.1689166 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Nørgaard-Pedersen C, Kesmodel US, Jørgensen MM, Christiansen O. IVIG And prednisolone in recurrent pregnancy loss: rCT. BMJ Open. (2025) 15(11):e106024. 10.1136/bmjopen-2025-106024 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Han JW, Park JS, Kim JS, Lee SK. IVIG Efficacy in recurrent pregnancy loss with immune abnormalities. Front Endocrinol. (2025) 16:1546602. 10.3389/fendo.2025.1546602 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Cavalcante MB, Harrity C, Luu T, Fatunbi J, Nakagawa K, Zhang Y, et al. ASRI Guidelines for recurrent pregnancy loss. Am J Reprod Immunol. (2025) 93(6):e70099. 10.1111/aji.70099 [DOI] [PubMed] [Google Scholar]
- 20.Vinci F, Tiseo M, Colosimo D, Calandrino A, Ramenghi LA, Biasucci DG. Point-of-care brain ultrasound and Doppler in neonates. Eur J Pediatr. (2024) 183(3):1059–72. 10.1007/s00431-023-05388-0 [DOI] [PubMed] [Google Scholar]
- 21.Benedetti GM, Vesoulis ZA. Updates in neonatal neuromonitoring. Clin Perinatol. (2025) 52(2):215–36. 10.1016/j.clp.2025.02.003 [DOI] [PubMed] [Google Scholar]
- 22.Cannata G, Pezzato S, Esposito S, Moscatelli A. Optic nerve sheath diameter ultrasound. Diagnostics (Basel). (2022) 12(3):767. 10.3390/diagnostics12030767 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Truong DM-T, Nguyen MHN, Nguyen HQ, Vo LT, Nguyen TT. ONSD Trajectories and mortality. World J Crit Care Med. (2025) 14(4):110669. 10.5492/wjccm.v14.i4.110669 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Zhang Z, Zhang C, Zhang X, Xu Y, Dou W, Li M, et al. Dynamic structure–function coupling in neonatal brain networks. Commun Biol. (2025) 9(1):35. 10.1038/s42003-025-09300-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Jiang L, Yu Q, Wang F, Wu M, Liu F, Fu M, et al. Blood pressure variability and cerebral blood flow in IVH prediction. Front Pediatr. (2023) 11:1241809. 10.3389/fped.2023.1241809 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche PC, Vandenbroucke JP, et al. The strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. Ann Intern Med. (2007) 147(8):573–7. 10.7326/0003-4819-147-8-200710160-00010 [DOI] [PubMed] [Google Scholar]
- 27.Collins GS, Moons KGM, Dhiman P, Riley RD, Beam AL, Van Calster B, et al. TRIPOD + AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. Br Med J. (2024) 385:e078378. 10.1136/bmj-2023-078378 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Moons KGM, Damen JAA, Kaul T, Hooft L, Andaur Navarro C, Dhiman P, et al. PROBAST + AI: an updated quality, risk of bias, and applicability assessment tool for prediction models using regression or artificial intelligence methods. Br Med J. (2025) 388:e082505. 10.1136/bmj-2024-082505 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Pal A, Stewart D, Ojha K, Kriss V, Ziegler C, Fischer H. Cerebral resistive indices and IVH. BMC Pediatr. (2025) 25(1):766. 10.1186/s12887-025-06119-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Tanaka K, Matsumoto S, Minamitani Y, Imamura H, Yoshimatsu H, Nakamura T, et al. Internal cerebral vein pulsation and IVH. Am J Perinatol. (2024) 41(S01):e37–45. 10.1055/a-1817-6638 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Pfurtscheller D, Baik-Schneditz N, Schwaberger B, Urlesberger B, Pichler G. Neonatal cerebral autoregulation: systematic review. Children. (2023) 10(8):1304. 10.3390/children10081304 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Zou M, Yu L, Lin R, Feng J, Zhang M, Ning S, et al. Cerebral autoregulation and brain injury. J Am Heart Assoc. (2023) 12(12):e028147. 10.1161/JAHA.122.028147 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Qin C, Zhao X, Shen Y, Lu Y, Li S, Zhang C, et al. Cerebral perfusion in IVH using ASL MRI. Pediatr Radiol. (2024) 54(5):776–86. 10.1007/s00247-024-05865-0 [DOI] [PubMed] [Google Scholar]
- 34.Pellicer A, de Boode W, Dempsey E, Greisen G, Mintzer J, Naulaers G, et al. Cerebral near-infrared spectroscopy guided neonatal intensive care management for the preterm infant. Pediatr Res. (2024). 10.1038/s41390-024-03649-8 [DOI] [PubMed] [Google Scholar]
- 35.van de Meent M, Bel EW, Ganzevoort W, Gordijn SJ, Groenendaal F, Kooi EMW, et al. Perinatal risk prediction model. Ultrasound Obstet Gynecol. (2025) 66(2):175–85. 10.1002/uog.29265 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.