Kawasaki disease is an acute febrile vasculitis that predominantly affects young children and, if untreated, leads to coronary artery lesions in approximately 20 to 25 percent of patients, with the potential for luminal narrowing, myocardial ischemia, and tissue necrosis. High-dose intravenous immunoglobulin (IVIG) is the standard initial treatment, yet 10 to 20 percent of children are resistant to it and remain at heightened risk of severe cardiac complications. Early identification of these resistant cases is critical because intensified regimens, such as IVIG combined with corticosteroids or immunosuppressants, are recommended for high-risk patients. To meet this need, the research team built a support vector machine (SVM) model trained on routine clinical and laboratory data and tested it across multiple independent patient cohorts.

The study, published in World Journal of Pediatrics, drew on electronic medical records from Children’s Hospital of Soochow University in Suzhou, where a retrospective development cohort of 2,371 children with Kawasaki disease treated between December 2018 and June 2024 was assembled. External validation was performed on 443 patients from Fuzhou and 198 from Yangzhou, and a prospective cohort of 253 patients treated in Suzhou between July and December 2024 provided an additional real-world test. Among the development cohort, 12.9 percent were IVIG-resistant, consistent with previous reports from Chinese populations. The median age was 26.0 months, and 60.9 percent of patients were male.

The team evaluated twelve machine learning algorithms, including gradient boosting, adaptive boosting, random forest, neural networks, and regularized generalized linear models, using repeated 10-fold cross-validation for training and hyperparameter tuning. Gradient boosting achieved an internal AUC of 0.773, followed closely by AdaBoost at 0.765 and random forest at 0.756, with the SVM at 0.750. After considering both discrimination and balanced sensitivity and specificity, the researchers narrowed the field to three algorithms and then applied SHapley Additive exPlanations (SHAP) values, an explainable AI technique that quantifies each feature’s contribution to individual predictions, to guide stepwise feature elimination.

The support vector machine maintained the highest AUC throughout the feature reduction process and emerged as the final model, incorporating just eight predictors: coronary artery lesions, C-reactive protein (CRP), age at presentation, neutrophil percentage, white blood cell count, total cholesterol, platelet count, and serum albumin. The eight-feature model outperformed both a 14-feature version (AUC 0.782 versus 0.756) and a minimal two-feature model (0.782 versus 0.736), with differences confirmed by the DeLong test. In internal validation, the final model achieved an AUC of 0.782, and SHAP scatter plots clarified the direction of each predictor’s effect: CRP, neutrophil percentage, white blood cell count, and the presence of coronary artery lesions pushed predictions toward resistance, while older age, higher total cholesterol, higher platelet counts, and higher albumin were protective. For example, white blood cell counts above 21.15 ×10⁹/L and the presence of coronary lesions produced positive SHAP values, whereas patients older than 65 months showed predominantly negative values.

External and prospective validation demonstrated that the model generalizes beyond its original institution. The AUC reached 0.746 in the Fuzhou cohort, 0.759 in Yangzhou, and 0.799 in the prospective Suzhou cohort. Calibration curves showed good agreement between predicted and observed probabilities, with Brier scores of 0.099, 0.097, and 0.068 across the three validation settings. Decision curve analysis, which quantifies net clinical benefit across a range of risk thresholds, showed the model outperformed both the “treat-all” and “treat-none” default strategies across clinically relevant probability ranges in every cohort. When benchmarked against established scoring systems, the machine learning model clearly outperformed the Kobayashi, Egami, Sano, Formosa, and Li scores, whose AUCs ranged from 0.619 to 0.674, all well below the SVM’s 0.782, which also delivered balanced sensitivity of 0.735 and specificity of 0.721.

A key concern in clinical machine learning is the “black box” problem, in which opaque algorithms undermine clinician trust. The researchers addressed this with SHAP-based global and local explanations, showing for individual patients which factors drove their risk classification. Interaction analyses further revealed that albumin had the highest overall interaction strength, followed by CRP, platelets, and total cholesterol, indicating the model captured non-additive relationships among clinical variables rather than simple independent effects. To translate the model into practice, the team built an interactive web-based calculator on the Shiny platform that accepts the eight features and returns a color-coded, individualized risk estimate. The optimal classification threshold, set at 14.9 percent by the Youden index, yielded a sensitivity of 0.648 and specificity of 0.606, and the authors emphasize that the threshold should guide early risk stratification rather than serve as a strict treatment cutoff.

Because coronary artery lesions require echocardiography, which may not be immediately available, the team also reconstructed a model without that predictor. Among algorithms tested without coronary lesions, an SVM variant, LASSO, and elastic net performed comparably, and stepwise reduction identified a five-variable model with an AUC of 0.762 that outperformed both smaller and larger versions. Importantly, the performance difference between the model with coronary lesions (AUC 0.750) and the model without them (AUC 0.739) was not statistically significant, meaning clinicians facing a febrile child before echocardiography can still obtain a meaningful risk estimate from blood tests and age alone.

The biological signals captured by the model align with current understanding of Kawasaki disease pathogenesis. Elevated CRP and neutrophil percentage reflect intense systemic inflammation, and neutrophils are known to contribute to vascular injury by releasing reactive oxygen species, proteases, and proinflammatory cytokines, a mechanism supported by autopsy studies showing dense neutrophilic infiltration in coronary lesions. Thrombocytopenia may indicate ongoing platelet consumption at sites of coronary artery injury, where platelets adhere to exposed collagen and release vascular endothelial growth factor and matrix metalloproteinases that aggravate vascular remodeling. Hypoalbuminemia likely reflects increased vascular permeability and impaired hepatic synthesis under systemic inflammation, while the association between low total cholesterol and resistance is consistent with prior observations of lipid abnormalities in resistant patients. Younger age, an independent risk factor, is thought to relate to immune system immaturity and to atypical or incomplete presentations that delay diagnosis; previous work has reported resistance rates as high as 31.5 percent in infants under 12 months.

The analysis also revealed clear age-related heterogeneity in which biomarkers matter. Among infants aged 0 to 5 months, only neutrophil percentage and total cholesterol differed significantly between resistant and responsive patients, whereas in children aged 3 to 4 years, seven of the eight features showed significant associations. The authors note that modeling age as a single variable may not fully capture such effect modification and suggest that future work could benefit from age-stratified or interaction-based modeling approaches.

The study was reported in accordance with the TRIPOD+AI guidelines for transparent reporting of clinical prediction models using artificial intelligence, and methodological quality was assessed with the PROBAST+AI tool. That assessment identified a high risk of bias in the participants domain owing to the retrospective design of the primary cohort, in the predictors domain because blinding of predictor assessment could not be ensured, and in the analysis domain because sample size considerations were not formally assessed, while the outcome domain was rated low risk and applicability concerns were low across all domains. A conventional logistic regression model built on the same eight predictors achieved an AUC of 0.732, statistically indistinguishable from the SVM’s 0.782, a finding the authors interpret honestly, noting that in settings with moderate sample sizes and structured clinical data, traditional approaches can remain competitive with machine learning.

The authors acknowledge further limitations, including potential clinical subjectivity in Kawasaki disease diagnosis, reliance on structured electronic medical record data without raw echocardiographic images, and possible calibration drift from temporal differences between development and validation cohorts collected over different years. They call for prospective trials using propensity score matching to determine whether model-guided decision-making actually improves clinical outcomes, and they propose future integration of multi-omics and imaging data, as well as exploration of emerging tabular foundation models such as TabPFN. For now, the model and its freely accessible web calculator offer pediatricians a validated, interpretable, and immediately usable instrument for identifying, at the bedside and before the first infusion, which children with Kawasaki disease need escalated therapy to protect their coronary arteries.

Historically, pediatric growth studies have struggled with the scarcity of reliable markers of pubertal staging or hormone quantification in large-scale datasets, often defaulting to broad classifications by sex that fail to capture the gradual and nuanced aspects of development. The ESI model elegantly addresses this gap by deriving a continuous numeric score directly from standard ECG metrics using AI algorithms trained on adult populations but applied here without recalibration in children and adolescents. The innovation lies in decoding cardiovascular electrical patterns that subtly encode developmental biology, thereby providing a non-invasive biomarker that reflects maturation processes at a granular level.

In analyzing over 60,000 pediatric ECGs spanning ages from newborns up to 18 years, researchers observed fascinating patterns that correspond closely with known physiological growth milestones. During early childhood, ESI scores exhibited remarkable homogeneity with minimal differences between sexes, indicating that the cardiovascular electrical signature at this stage is generally uniform. As children transition into late childhood and adolescence, however, ESI values begin to diverge distinctly between males and females, reflecting the underlying hormonal and physiologic shifts heralding puberty. These divergences reached plateaus during mid-to-late adolescence, mirroring the completion of many developmental trajectories.

The study reveals that the ESI provides a continuous representation of biological maturation rather than forcing developmental stages into rigid categories, thereby offering researchers a finer resolution tool for examining growth pathways. This spectral approach captures the complexity of gender-specific physiological changes through the lens of cardiac bioelectrics, a domain previously under-explored for its developmental insights. The ability of ESI to adapt to a wide demographic is further demonstrated by consistent age-related patterns observed across different racial groups, suggesting robustness and broad applicability.

Importantly, the model’s accuracy improved steadily with the subjects’ age, paralleling the convergence of adolescent cardiovascular physiology toward adult normative parameters. The adult-trained ESI was projected onto pediatric ECG data without retraining, a strategy allowing investigators to assess the evolving relationship between childhood cardiac electrical signals and adult benchmarks. This cross-age applicability underscores the continuum between pediatric maturation and adult physiology, with ESI serving as a bridge for longitudinal understanding.

Beyond its theoretical contributions, the ESI opens exciting clinical and research applications by providing a practical biomarker for biological stage when conventional indicators, such as Tanner staging or direct hormone assays, are unavailable. In particular, large-scale epidemiological studies and clinical trials could leverage ESI to refine patient stratification, control for developmental confounders, and better understand how maturation influences cardiovascular risk profiles and therapeutic responses. This is especially valuable given the logistical and ethical challenges of hormone measurement and physical staging in pediatric populations.

From a technical standpoint, the success of ESI hinges on the integration of AI with high-dimensional ECG data to extract latent developmental signals embedded within cardiac waveforms. AI techniques, potentially including deep learning and pattern recognition, analyze temporal and spatial ECG features such as QRS morphology, T-wave dynamics, and heart rate variability that exhibit developmental modulation. The approach redefines the role of ECGs from purely cardiologic diagnostics to multidimensional phenotyping tools that incorporate growth and maturation signals.

Despite the promising findings, the authors emphasize the necessity of future longitudinal investigations that incorporate direct clinical measures such as Tanner staging, hormone level measurements, and follow-up cardiovascular outcomes. Such studies are crucial for validating the biological and clinical relevance of ESI and for translating its use into routine pediatric practice. Assessing how ESI correlates with established developmental milestones and predicts long-term health trajectories will illuminate its full potential.

The implications of this work extend beyond pediatric cardiology into broader domains of developmental biology and precision medicine. By unveiling a novel quantitative biomarker derived from widely available clinical data, the study pioneers a new paradigm in which AI-augmented diagnostics can non-invasively track complex biological processes. This advances our ability to monitor health and disease from early life stages and tailor interventions according to individual developmental status.

Wake Forest University School of Medicine and its partners in Advocate Health have laid the groundwork for this innovation within an infrastructure rich in clinical data and AI research expertise. Their interdisciplinary collaboration showcases how AI can maximize the informational yield of standard medical tests, transforming routine clinical procedures into powerful tools for biomedical discovery and personalized healthcare.

This evolutionary leap in pediatric assessment will likely catalyze further research into the interplay between cardiovascular development and systemic maturation. It offers a promising route to addressing longstanding challenges in pediatric medicine, including the heterogeneity of pubertal timing and its impact on health outcomes. By situating cardiovascular electrophysiology within the broader context of developmental science, the ESI exemplifies the future of integrative, AI-guided medicine.

In conclusion, the routine ECG, augmented with AI-driven analytics embodied by the Electrocardiographic Sex Index, may soon become an indispensable instrument for tracking the intricate process of childhood and adolescent development. This innovative approach promises to refine our understanding of biological growth, improve the precision of pediatric research, and ultimately enhance clinical care by integrating developmental maturity into assessments. As the next phases of research unfold, this sensory window into maturation holds vast potential to transform pediatric health paradigms worldwide.

Subject of Research: Biological Development and Cardiovascular Maturation in Children and Adolescents Using ECG and AI

Article Title: ECG Sex Index in Children and Adolescents

News Publication Date: 11-Apr-2026

Web References:

• Wake Forest University School of Medicine: https://school.wakehealth.edu/
• European Heart Journal – Digital Health: https://academic.oup.com/ehjdh/article/7/4/ztag058/8651698?login=false
• Advocate Health: https://www.advocatehealth.org/

References:
Karabayir I., Hayit T., et al. (2026). ECG Sex Index in Children and Adolescents. European Heart Journal – Digital Health. DOI: 10.1093/ehjdh/ztag058

Image Credits: Advocate Health

Keywords: Pediatric Development, Electrocardiogram, Biological Maturation, Artificial Intelligence, Cardiovascular Growth, Electrocardiographic Sex Index, Tanner Staging, Hormonal Changes, Pediatric Cardiology, AI in Medicine, Childhood Growth Tracking, Cardiovascular Electrophysiology

The complexity of congenital and acquired cardiac conditions in children makes prognosis and treatment planning exceptionally challenging. Traditionally, clinicians have depended on a mixture of clinical judgment, standard diagnostic tools, and historical data to estimate hospital duration and tailor therapies. However, the heterogeneity within pediatric cardiology cases poses a significant barrier to precise predictions, often leading to either prolonged hospitalization or premature discharge, both of which can jeopardize patient outcomes. This new research pivots on the hypothesis that machine learning algorithms can learn underlying patterns from multi-dimensional datasets to forecast hospital stay length and identify patients with similar clinical trajectories.

The team spearheading this innovation integrated an array of structured and unstructured clinical data, encompassing demographic details, diagnostic imaging reports, biochemical markers, and electronic health records from pediatric cardiology patients. By employing sophisticated preprocessing techniques, they harmonized these inputs into a comprehensive dataset suitable for advanced machine learning models. This step ensured the removal of noise, imputation of missing values, and normalization to circumvent biases stemming from inconsistent data entry or recording protocols.

Central to their approach was the development and validation of prediction algorithms rooted in ensemble learning methods, which combine multiple machine learning models to enhance robustness and accuracy. Models such as gradient boosting machines and random forests were meticulously tuned to anticipate the length of hospital admission, factoring in complex interactions among clinical variables, previous interventions, and comorbidities. The predictive performance was rigorously evaluated against traditional statistical baselines, demonstrating a remarkable improvement in precision and recall metrics.

Beyond single-patient prediction, the researchers introduced a novel patient similarity retrieval system designed to cluster patients with analogous profiles and anticipated clinical courses. By leveraging embedding techniques and distance metrics tailored for heterogeneous medical data, they created a dynamic repository of patient archetypes. This advancement empowers clinicians to retrieve historical cases that closely align with a current patient’s characteristics, thereby enriching clinical insights through analogical reasoning and evidence-based comparisons.

The study’s significance extends into resource management within pediatric care units. Accurate predictions of hospital stay durations enable healthcare providers to optimize bed allocations, staffing schedules, and post-discharge planning. Particularly in pediatric cardiology, where prolonged hospitalizations can be resource-intensive and emotionally taxing for families, effective forecasting serves as a cornerstone for cost-efficiency and quality improvement initiatives.

From a technical perspective, the researchers navigated substantial challenges inherent in medical machine learning, including class imbalance due to varying prevalence of cardiac conditions and interpretability of predictive models. To tackle these hurdles, they incorporated stratified sampling and explainability tools such as SHAP (SHapley Additive exPlanations), enabling transparent elucidation of model decisions for each prediction. This feature is especially critical in clinical environments where acceptance hinges on trust and comprehension among healthcare practitioners.

The fusion of machine learning with pediatric cardiology also opens avenues for identifying latent phenotypes within the patient population. By analyzing clusters defined through similarity retrieval, the team discovered subgroups exhibiting distinct risk profiles and response patterns, potentially guiding targeted therapeutic interventions. Such phenotyping aligns with the broader movement towards precision medicine, which aims to move beyond one-size-fits-all treatments towards data-informed personalization.

Furthermore, the system’s adaptability was demonstrated through its capacity to update continually with new patient data, maintaining predictive relevance as treatment protocols evolve and patient demographics shift. This adaptability ensures that the machine learning framework remains a practical, living tool within clinical workflows rather than an obsolete academic exercise.

Ethical considerations surrounding data security, privacy, and algorithmic bias were meticulously addressed throughout the research process. The team implemented rigorous de-identification protocols and equitable model training techniques to uphold patient confidentiality and minimize disparities in prediction accuracy across different demographic groups. These measures underscore the critical intersection of technology, trust, and medicine.

Another exciting aspect of this development is its potential interoperable integration with existing hospital information systems and clinical decision support tools. Seamless embedding into electronic health records could enable real-time predictions during patient admissions, thereby aiding clinicians at the point of care without adding burdensome manual input. The usability factor significantly elevates the chances of adoption and meaningful impact.

The research, published in Nature Communications in 2026, stands as a testament to the transformative potential of artificial intelligence in pediatric healthcare. It highlights the collaborative synergy between data scientists, cardiologists, and clinical informaticians aiming to harness technology for tangible, life-improving outcomes. This convergence not only advances cardiology but also sets a precedent for other pediatric specialties grappling with similar prognostic complexities.

While promising, the authors acknowledge limitations including the need for multi-center validation across diverse populations to ensure generalizability. Additionally, prospective clinical trials measuring the actual impact on patient outcomes and healthcare logistics remain essential future steps. Nonetheless, the framework’s foundational robustness indicates a trajectory steering towards routine clinical applicability.

In essence, this innovative application of machine learning to predict hospital stays and retrieve clinically analogous patients represents a paradigm shift in pediatric cardiology. By transforming voluminous and complex clinical data into actionable intelligence, it empowers clinicians with foresight and precision previously unattainable. As artificial intelligence continues to evolve, such integrative technologies promise to elevate pediatric care standards, reduce healthcare costs, and ultimately improve the lives of children battling cardiac diseases worldwide.

Subject of Research: Machine learning application for predicting hospital stay duration and patient similarity retrieval in pediatric cardiology.

Article Title: Clinically-applicable prediction of hospital stay and patient similarity retrieval in paediatric cardiology using machine learning.

Article References:
Rigny, L., Biggart, I., Zakka, K. et al. Clinically-applicable prediction of hospital stay and patient similarity retrieval in paediatric cardiology using machine learning. Nat Commun (2026). https://doi.org/10.1038/s41467-026-73021-3

Image Credits: AI Generated

Fontan surgery, developed for patients with single ventricular physiology, has provided hope for many, allowing them to lead relatively normal lives. However, it comes with its own set of complications, notably Fontan failure and liver disease. These conditions not only challenge the longevity of patients but also complicate their quality of life. The study highlights the urgent need for pioneering predictive tools that would allow clinicians to make proactive decisions, rather than reactive ones, regarding patient care.

Utilizing multi-parametric abdominal MRI, the researchers explored how radiomic features—essentially quantitative data mined from medical images—can serve as robust predictors of clinical outcomes. By applying sophisticated machine learning algorithms, the team was able to analyze vast amounts of data and identify correlations that would likely stay hidden under traditional analytical methods. This approach opens new avenues for early intervention and personalized treatment plans that could significantly impact patient outcomes over time.

The study incorporated a diverse cohort of patients who had undergone the Fontan procedure, emphasizing the importance of a well-rounded dataset. By examining imaging features in conjunction with clinical parameters, the researchers were able to develop models that more accurately reflect the multidimensional aspects of Fontan physiology. This dual focus on imaging and clinical data represents a paradigm shift in how clinicians can assess risk and determine treatment strategies for their patients.

One of the standout findings of Prasad and colleagues was the correlation between specific radiomic features and liver disease severity. In particular, the study noted that certain parameters could predict advanced liver disease long before traditional clinical markers would raise alarms. The implications of this discovery could be far-reaching, allowing for timely interventions that could prevent the progression of liver complications in vulnerable populations.

Moreover, the integration of machine learning has been highlighted as a game-changer in the field of pediatric imaging. The algorithms are not only capable of processing vast datasets but are also constantly refining their predictions as new data becomes available. This adaptability positions machine learning as an invaluable asset in clinical settings where rapid, informed decision-making is crucial.

As the research community delves deeper into this innovative approach, we can expect to see more institutions adopting machine learning as a standard practice for analyzing medical imaging. The potential for these techniques to enhance diagnostic accuracy and the precision of therapeutic interventions cannot be overstated. The traditional methods that have long dominated the field are now increasingly being recognized as insufficient in the face of rapid technological advancements.

In addition to its clinical implications, this research raises important questions regarding the future of personalized medicine. With machine learning algorithms capable of predicting patient-specific outcomes, the healthcare landscape may soon witness a shift towards treatments tailored to individual patient profiles. Such an evolution could democratize high-quality care, making it accessible to a broader spectrum of patients and allowing for more nuanced management of congenital heart diseases.

In a broader context, the collaboration between disciplines—merging imaging, data science, and clinical practice—illustrates the potential benefits of interdisciplinary approaches in healthcare. By fostering environments where specialists in different fields can work together, there is a greater likelihood that innovative solutions will emerge, addressing some of the most pressing challenges facing pediatric cardiology today.

As we await further developments stemming from this research, the findings bridge a significant gap in the current methodologies used in clinical settings. They suggest a future where predictive analytics will support clinicians in managing complex conditions more effectively. With continued research and advancements, the potential to transform the management of Fontan patients and mitigate associated risks appears more promising than ever.

In summary, Prasad et al.’s study illuminates a path forward in the prediction of Fontan failure and liver disease severity through machine learning and advanced imaging techniques. As the fields of artificial intelligence and medical imaging converge, the hope remains that patients’ lives will improve through earlier detection, tailored treatments, and better quality of care. The ongoing dialogue in this area signifies a commitment to accomplish what was previously deemed complex, with the ultimate goal of enhancing patient outcomes.

With continued emphasis on research initiatives and technology integration in clinical practices, the future of pediatric cardiology seems poised for remarkable advancements. The attention garnered by studies like this one highlights not only the significance of technological innovation but also the persistent need for clinical vigilance in the management of congenital heart disease.

As we look ahead, we can expect the impact of these findings to ripple through the healthcare landscape, encouraging a new generation of tools and practices designed to improve the lives of patients facing chronic conditions. The collaboration of technology with expert clinical insight is indeed a thrilling prospect, one that promises a brighter future for children living with congenital heart disease.

Subject of Research: Prediction of Fontan failure and correlates of Fontan-associated liver disease severity using machine learning and radiomic features.

Article Title: Prediction of Fontan failure and correlates of Fontan-associated liver disease severity using machine learning and radiomic features from multi-parametric abdominal MRI.

Article References:

Prasad, A., Opotowsky, A., Trout, A. et al. Prediction of Fontan failure and correlates of Fontan-associated liver disease severity using machine learning and radiomic features from multi-parametric abdominal MRI.
Pediatr Radiol (2026). https://doi.org/10.1007/s00247-025-06506-w

Image Credits: AI Generated

DOI: 03 February 2026

Keywords: Fontan surgery, machine learning, radiomics, pediatric cardiology, liver disease, predictive analytics, imaging techniques.