Texas Children's Hospital

This study aims to enhance antenatal detection of placenta accreta spectrum (PAS) and predict severe hemorrhage at delivery using machine learning by evaluating the association between antenatal hematologic index trends across trimesters, imaging markers, and patient history. We retrospectively analyzed 2017-2023 data from a PAS referral center, including demographics, laboratory results, ultrasounds, and outcomes. Patients with confirmed PAS (cases) were compared to those with antenatal risk but no histopathologic evidence of PAS (controls). Statistical analyses and machine learning models were developed to predict PAS. We also used machine learning to predict severe hemorrhage (>1500 mL) in the cases. A total of 186 PAS cases and 217 controls were identified, showing significant differences in body mass index, gravidity, parity, prior cesarean deliveries, gestational age at delivery, and PAS ultrasound findings. Logistic regression highlighted prior cesarean deliveries (odds ratio [OR] 1.8; 95% confidence interval [CI] 1.3-2.4) and second (OR 28.1; 95% CI 12.7-60.8) or third trimester ultrasound markers (OR 27.6; 95% CI 13.2-61.1) as strong predictors of PAS. Third trimester mean platelet volume was inversely associated with PAS (OR 0.55; 95% CI 0.39-0.78). Machine learning models achieved high accuracy. Model 1 predicted PAS with 90% accuracy. Model 2 predicted PAS with 88.8% accuracy using early gestational hematologic markers. Model 3 predicted severe hemorrhage (>1500 mL) with 74.3% accuracy. Machine learning models combining patient history, imaging, and hematologic markers detect PAS and predict hemorrhage with up to 90% accuracy. These tools improve antenatal diagnosis of PAS, which enhances maternal outcomes by enabling early identification and better resource allocation.

Contrast-enhanced echocardiography (CE-echo) is commonly used in adult patients to optimize endocardial border delineation (EBD) but there is limited experience in children. A prospective, open-label, non-randomized, multicenter study was performed to evaluate the efficacy of intravenous Optison for CE-echo in pediatric patients and determine the optimal dose of Optison. Subjects ≥ 9 and < 18years with suboptimal echocardiograms (defined as ≥ 2 contiguous segments that could not be adequately visualized) were enrolled. After a baseline non-contrast echocardiogram (NC-echo), CE-echo was performed with 2 dose levels of Optison based on patient weight. The primary endpoint was visualization of the LV wall segments (LVWS) in apical 2- and 4-chamber views. An EBD score was calculated by assigning a value from 0 (no visualization) to 3 (optimal visualization) for 12 LVWS. 37 subjects across 8 sites completed the study. 30 had complete echocardiograms at both dose levels. 2 patients had usable images at dose 2 only. Mean age was 13.4years (± 2.6) with 23 males (62%). Mean weight was 73.9kg (± 36.8) with wide distribution of weight (interquartile range of 42-102kg). There were 8 patients under 40kg. Most subjects had a prior medical/surgical history (n = 33, 89%). Both dose levels of Optison CE-echo improved the number of LVWS visualized from 3.7 ± 1.1 on NC-echo to 10.0 ± 0.1 with dose 1 (p < 0.001) and 10.5 ± 0.1 with dose 2 (p < 0.001). EBD score also improved at both dose levels from 11.9 ± 2.5 for NC-echo to 28.5 ± 1.7 for dose 1 (p < 0.001) and 30.2 ± 8.4 for dose 2 (p < 0.001). The number of suboptimal echocardiograms improved from 92.3% ± 10.7 in NC-echo to 26.2% ± 9 with dose 1 (p < 0.001) and 30.8% ± 4.5 with dose 2 (p < 0.001). CE-echo with intravenous Optison in pediatric patients improved the number of LVWS visualized, improved EBD score, and reduced the number of suboptimal echocardiograms at both dosages used in this study. Intravenous Optison can improve LV visualization in pediatric patients with suboptimal echocardiograms.

Left atrial strain (LAS) is a sensitive marker of early atrial dysfunction and left ventricular diastolic abnormalities in pediatric heart disease, yet clinical integration remains limited. Evidence from 57 studies (2015-2024) including >5,300 patients <21 years of age demonstrates that healthy pediatric patients have higher LAS values, particularly left atrial conduit strain, than neonates and adults. Reduced left atrial reservoir and conduit strain is consistently reported in congenital heart disease and is associated with adverse outcomes across single-ventricle palliation stages. In pediatric cardiomyopathy, LAS impairment parallels worsening diastolic function. Among heart transplant recipients, left atrial reservoir strain correlates more strongly with invasive filling pressures than conventional noninvasive metrics. LAS also predicts myocardial injury in multisystem inflammatory syndrome in children, ischemic risk in diabetes, and early left ventricular dysfunction from chemotherapy. Despite its potential for diastolic function assessment, variability in acquisition and analysis remains substantial. Standardized protocols and larger prospective studies are needed to establish normative values and define clinical thresholds.

Curve progression to surgical range in patients with adolescent idiopathic scoliosis (AIS) is strongly associated with the initial major curve angle and the degree of skeletal maturity as staged by the modified Proximal Humerus Ossification System (PHOS) and other systems. Our purpose was to (1) develop a prognostic model and risk score to estimate the probability of progression to surgical indications in untreated patients using the PHOS and (2) compare its performance to a model using the Risser stage. Patients from the BrAIST study and 2 other institutions were followed to either skeletal maturity (Risser 4+), a major curve angle of ≥45°, or spinal fusion. Candidate variables for the predictive models included age, sex, major curve angle, Scoliosis Research Society (SRS) curve classification, status of triradiate cartilage, Risser stage, and PHOS. Model calibration and discrimination were evaluated. A probability threshold was set, creating low- and high-risk groups to aid in clinical decision-making. Overall, 164 patients (77% female) were included. The mean age at presentation was 12.2±1.4 years (range: 10 to 16 years), and the mean maximum major curve angle was 24±9° (range: 10° to 43°). Fifty-six (34%) patients progressed to a surgical range or had a spinal fusion. The PHOS model included the major curve angle and the presence/absence of a thoracic apex. The model demonstrated strong discrimination (c-statistic = 0.89) and calibration (ICI = 0.02), performing similarly to one developed in this sample using the Risser stage. The sensitivity was 0.91, the specificity was 0.71, the PPV was 0.62, and the NPV was 0.94 at the probability cut-point of 0.22. This study derived a prognostic model estimating the baseline risk of progression to surgical indications in AIS patients using the PHOS. Estimates from this model can inform the shared decision-making process and motivate compliant bracing. Further validation in larger independent samples and exploration of the PHOS to predict bracing outcomes should be performed. Level II.

Background Children with posterior fossa tumors (PFTs) are at risk of long-term neurocognitive deficits, yet the impact of tumor grade and treatment intensity on cortical development remains poorly understood. The aim of this study was to determine whether cortical thickness trajectories differ between children with low- versus high-grade PFTs and whether radiation dose influences cortical development in medulloblastoma patients. We conducted a retrospective longitudinal study of 55 children (37 low-grade, 18 high-grade medulloblastoma) diagnosed between 2018 and 2020. A total of 406 T1-weighted MRI scans acquired up to five years post-diagnosis were processed using CAT12 in SPM12 to estimate cortical thickness. Vertex-wise linear mixed-effects models assessed cortical thickness changes over time, adjusting for age at diagnosis, sex, and hydrocephalus. Within the high-grade group, we evaluated the effect of craniospinal irradiation (CSI) dose (23.4 vs. ≥36 Gy) on cortical development. Children with low-grade tumors showed significant bilateral cortical thinning over time, consistent with typical neurodevelopment. In contrast, medulloblastoma patients treated with surgery, chemotherapy, and radiation therapy showed relative cortical stability across follow-up. Significant group-by-time interactions were observed, with low-grade patients demonstrating greater cortical thinning in parietal, temporal, and occipital regions. No significant differences in cortical trajectories were found between high- and standard-dose CSI subgroups. Cortical development trajectories in pediatric PFT survivors differ by tumor grade and treatment intensity. While low-grade patients exhibit expected maturational thinning, cortical stability in high-grade patients may reflect disrupted neurodevelopment. These findings underscore the need for longitudinal neuroimaging in pediatric neuro-oncology survivorship care.