Predictive modelling of hypoxic ischaemic encephalopathy risk following perinatal asphyxia

Current progress and future prospects of machine learning in the diagnosis of neonatal encephalopathy: a narrative review

Background and Objective

Neonatal encephalopathy (NE) can cause permanent neurological damage in newborns. NE greatly increases the burden of care placed on families. It also places a tremendous economic strain on the social health system. Currently, NE is mostly diagnosed by imaging and blood gas analysis. However, current diagnostic methods mostly lag behind the disease, leading to a lag in medical interventions for NE. In recent years, machine learning (ML) techniques have been applied to medicine, including in the early diagnosis and screening of diseases. This study aimed to provide an overview of existing research on the application of ML to NE and to offer insights for future investigations.

Methods

A full library search in fuzzy matching mode was performed to retrieve articles from the Web of Science database published between January 1, 2008, and August 31, 2024 using the following search strategy: (neonatal encephalopathy * machine learning) (where NE comprised all the relevant diseases, and ML comprised the main algorithms), and the key information was filtered.

Key Content and Findings

A total of 159 documents were retrieved, and 23 relevant documents were identified based on the topic, keywords and content. The relevant content showed that the included articles on NE and ML had issues in terms of study standardization, dichotomous study outcomes, and clinical usefulness.

Conclusions

To date, most studies on the application of ML to NE have not comprehensively considered the aspects of experimental design, data processing, model building, and evaluation. It is hoped that such models will provide effective decision-making tools for clinical practice in the future, and thus improve the healthy life span of newborns.

Validation of a machine learning algorithm for identifying infants at risk of hypoxic ischaemic encephalopathy in a large unseen data set

OBJECTIVE

To validate a hypoxic ischaemic encephalopathy (HIE) prediction algorithm to identify infants at risk of HIE immediately after birth using readily available clinical data.

DESIGN

Secondary review of electronic health record data of term deliveries from January 2017 to December 2021.

SETTING

A tertiary maternity hospital.

PATIENTS

Infants >36 weeks' gestation with the following clinical variables available: Apgar Score at 1 min and 5 min, postnatal pH, base deficit, and lactate values taken within 1 hour of birth INTERVENTIONS: Previously trained open-source logistic regression and random forest (RF) prediction algorithms were used to calculate a probability index (PI) for each infant for the occurrence of HIE.

MAIN OUTCOME

Validation of a machine learning algorithm to identify infants at risk of HIE in the immediate postnatal period.

RESULTS

1081 had a complete data set available within 1 hour of birth: 76 (6.95%) with HIE and 1005 non-HIE. Of the 76 infants with HIE, 37 were classified as mild, 29 moderate and 10 severe. The best overall accuracy was seen with the RF model. Median (IQR) PI in the HIE group was 0.70 (0.53-0.86) vs 0.05 (0.02-0.15), (p<0.001) in the non-HIE group. The area under the receiver operating characteristics curve for prediction of HIE=0.926 (0.893-0.959, p<0.001). Using a PI cut-off to optimise sensitivity of 0.30, 936 of the 1081 (86.5%) infants were correctly classified.

CONCLUSION

In a large unseen data set an open-source algorithm could identify infants at risk of HIE in the immediate postnatal period. This may aid focused clinical examination, transfer to tertiary care (if necessary) and timely intervention.

Advancing Obstetric Care Through Artificial Intelligence-Enhanced Clinical Decision Support Systems: A Systematic Review

Although artificial intelligence (AI) has grown over the past 10 years and clinical decision support systems (CDSS) have begun to be used in obstetric care, little is known about how AI functions in obstetric care-specific CDSS. We conducted a systematic review based on research studies that looked at AI-augmented CDSS in obstetric care to identify and synthesize CDSS functionality, AI techniques, clinical implementation, and AI-augmented CDSS in obstetric care. We searched four different databases (Scopus, PubMed, Web of Science, and IEEE Xplore) for relevant studies, and we found 354 studies. The studies were evaluated for eligibility based on predefined inclusion and exclusion criteria. The systematic review incorporated 30 studies after conducting an eligibility assessment of all studies. We used the Newcastle Ottawa Scale for risk bias assessment of all included studies. Medical prediction, therapeutic recommendations, diagnostic support, and knowledge dissemination constitute the key features of CDSS service offerings. The current research on CDSS included findings about early fetal anomaly detection, economical surveillance, prenatal ultrasonography assistance, and ontology development methodologies according to our study findings.

Longitudinal Analysis of Amplitude-Integrated Electroencephalography for Outcome Prediction in Infants with Hypoxic-Ischemic Encephalopathy: A Validation Study

OBJECTIVES

To validate the prognostic accuracy of a previously published tool (HOPE calculator) using longitudinal analysis of amplitude-integrated electroencephalography (aEEG) background activity and sleep-wake cycling to predict favorable or adverse 2-year neurodevelopmental outcome in infants with hypoxic-ischemic encephalopathy (HIE) undergoing therapeutic hypothermia (TH), and to evaluate the predictive value for outcome at 5-8 years of age.

STUDY DESIGN

Single-center retrospective cohort study in 117 infants who underwent TH for HIE between 2008 and 2022. We scored 2-channel aEEG BGPs, sleep-wake cycling, and seizure activity at 6-hour intervals for 84 hours. Neurodevelopmental outcome at 2 years was evaluated using the Bayley Scales of Infant Development-III, defining adverse outcome as death, cerebral palsy, and/or cognitive/motor scores of <85. Adverse outcome at 5-8 years was defined as a total IQ score of <85, a Movement-ABC-2 score of less than p15, cerebral palsy, severe sensory impairment, or death.

RESULTS

The prediction model showed an area under the curve of 0.90 (95% CI, 0.83-0.95) at 2 years and 0.83 (95% CI, 0.73-0.92) at 5-8 years. Mean predicted probability of favorable outcome was 74.5% (95% CI, 69.4-79.6) in the favorable outcome group compared with 32.8% (95% CI, 23.5-42.2) in the adverse outcome group (P < .001) at 2 years (n = 115) and 76.85% (95% CI, 70.0-83.4) compared with 40.7% (95% CI, 30.0-51.4) at 5-8 years (n = 68).

CONCLUSIONS

Our study provided external validation of the HOPE calculator, assessing longitudinal aEEG background activity during TH in infants with HIE. The results suggest that this method can predict favorable or adverse outcomes accurately not only at 2 but also at 5-8 years of age.

Artificial Intelligence-Augmented Clinical Decision Support Systems for Pregnancy Care: Systematic Review

BACKGROUND

Despite the emerging application of clinical decision support systems (CDSS) in pregnancy care and the proliferation of artificial intelligence (AI) over the last decade, it remains understudied regarding the role of AI in CDSS specialized for pregnancy care.

OBJECTIVE

To identify and synthesize AI-augmented CDSS in pregnancy care, CDSS functionality, AI methodologies, and clinical implementation, we reported a systematic review based on empirical studies that examined AI-augmented CDSS in pregnancy care.

METHODS

We retrieved studies that examined AI-augmented CDSS in pregnancy care using database queries involved with titles, abstracts, keywords, and MeSH (Medical Subject Headings) terms. Bibliographic records from their inception to 2022 were retrieved from PubMed/MEDLINE (n=206), Embase (n=101), and ACM Digital Library (n=377), followed by eligibility screening and literature review. The eligibility criteria include empirical studies that (1) developed or tested AI methods, (2) developed or tested CDSS or CDSS components, and (3) focused on pregnancy care. Data of studies used for review and appraisal include title, abstract, keywords, MeSH terms, full text, and supplements. Publications with ancillary information or overlapping outcomes were synthesized as one single study. Reviewers independently reviewed and assessed the quality of selected studies.

RESULTS

We identified 30 distinct studies of 684 studies from their inception to 2022. Topics of clinical applications covered AI-augmented CDSS from prenatal, early pregnancy, obstetric care, and postpartum care. Topics of CDSS functions include diagnostic support, clinical prediction, therapeutics recommendation, and knowledge base.

CONCLUSIONS

Our review acknowledged recent advances in CDSS studies including early diagnosis of prenatal abnormalities, cost-effective surveillance, prenatal ultrasound support, and ontology development. To recommend future directions, we also noted key gaps from existing studies, including (1) decision support in current childbirth deliveries without using observational data from consequential fetal or maternal outcomes in future pregnancies; (2) scarcity of studies in identifying several high-profile biases from CDSS, including social determinants of health highlighted by the American College of Obstetricians and Gynecologists; and (3) chasm between internally validated CDSS models, external validity, and clinical implementation.

Dried Blood Spot Metabolome Features of Ischemic-Hypoxic Encephalopathy: A Neonatal Rat Model

Hypoxic-ischemic encephalopathy (HIE) is a severe neurological disorder caused by perinatal asphyxia with significant consequences. Early recognition and intervention are crucial, with therapeutic hypothermia (TH) being the primary treatment, but its efficacy depends on early initiation of treatment. Accurately assessing the HIE severity in neonatal care poses challenges, but omics approaches have made significant contribution to understanding its complex pathophysiology. Our study further explores the impact of HIE on the blood metabolome over time and investigated changes associated with hypothermia's therapeutic effects. Using a rat model of hypoxic-ischemic brain injury, we comprehensively analyzed dried blood spot samples for fat-soluble compounds using HPLC-MS. Our research shows significant changes in the blood metabolome after HIE, with a particularly rapid recovery of lipid metabolism observed. Significant changes in lipid metabolites were observed after 3 h of HIE, including increases in ceramides, carnitines, certain fatty acids, phosphocholines, and phosphoethanolamines, while sphingomyelins and N-acylethanolamines (NAEs) decreased (p < 0.05). Furthermore, NAEs were found to be significant features in the OPLS-DA model for HIE diagnosis, with an area under the curve of 0.812. TH showed a notable association with decreased concentrations of ceramides. Enrichment analysis further corroborated these observations, showing modulation in several key metabolic pathways, including arachidonic acid oxylipin metabolism, eicosanoid metabolism via lipooxygenases, and leukotriene C4 synthesis deficiency. Our study reveals dynamic changes in the blood metabolome after HIE and the therapeutic effects of hypothermia, which improves our understanding of the pathophysiology of HIE and could lead to the development of new rapid diagnostic approaches for neonatal HIE.

Metabolite Biomarkers for Early Ischemic-Hypoxic Encephalopathy: An Experimental Study Using the NeoBase 2 MSMS Kit in a Rat Model

Hypoxic-ischemic encephalopathy (HIE) is one of the most common causes of childhood disability. Hypothermic therapy is currently the only approved neuroprotective approach. However, early diagnosis of HIE can be challenging, especially in the first hours after birth when the decision to use hypothermic therapy is critical. Distinguishing HIE from other neonatal conditions, such as sepsis, becomes a significant problem in diagnosis. This study explored the utility of a metabolomic-based approach employing the NeoBase 2 MSMS kit to diagnose HIE using dry blood stains in a Rice-Vannucci model of HIE in rats. We evaluated the diagnostic fidelity of this approach in a range between 3 and 6 h after the onset of HIE, including in the context of systemic inflammation and concomitant hypothermic therapy. Discriminant analysis revealed several metabolite patterns associated with HIE. A logistic regression model using glycine levels achieved high diagnostic fidelity with areas under the receiver operating characteristic curve of 0.94 at 3 h and 0.96 at 6 h after the onset of HIE. In addition, orthogonal partial least squares discriminant analysis, which included five metabolites, achieved 100% sensitivity and 80% specificity within 3 h of HIE. These results highlight the significant potential of the NeoBase 2 MSMS kit for the early diagnosis of HIE and could improve patient management and outcomes in this serious illness.

The past, current, and future of neonatal intensive care units with artificial intelligence: a systematic review

Machine learning and deep learning are two subsets of artificial intelligence that involve teaching computers to learn and make decisions from any sort of data. Most recent developments in artificial intelligence are coming from deep learning, which has proven revolutionary in almost all fields, from computer vision to health sciences. The effects of deep learning in medicine have changed the conventional ways of clinical application significantly. Although some sub-fields of medicine, such as pediatrics, have been relatively slow in receiving the critical benefits of deep learning, related research in pediatrics has started to accumulate to a significant level, too. Hence, in this paper, we review recently developed machine learning and deep learning-based solutions for neonatology applications. We systematically evaluate the roles of both classical machine learning and deep learning in neonatology applications, define the methodologies, including algorithmic developments, and describe the remaining challenges in the assessment of neonatal diseases by using PRISMA 2020 guidelines. To date, the primary areas of focus in neonatology regarding AI applications have included survival analysis, neuroimaging, analysis of vital parameters and biosignals, and retinopathy of prematurity diagnosis. We have categorically summarized 106 research articles from 1996 to 2022 and discussed their pros and cons, respectively. In this systematic review, we aimed to further enhance the comprehensiveness of the study. We also discuss possible directions for new AI models and the future of neonatology with the rising power of AI, suggesting roadmaps for the integration of AI into neonatal intensive care units.

Maternal and Fetal Risk Factors for Neonatal Hypoxic-Ischemic Encephalopathy: A Retrospective Study

Background

Neonatal hypoxic-ischemic encephalopathy (HIE) leads to different degree of neurological sequelae. The incidence of HIE is relatively high, and the risk factors associated with HIE are still controversial. It is necessary to identify the risk factors associated with HIE.

Methods

A total of 258 neonates (110 HIE patients and 148 controls) were enrolled in this study. The characteristics of pregnant women and fetuses during pregnancy and delivery were compared between HIE patients and controls, and the risk factors of HIE were analyzed.

Results

The proportions of premature infants, low-birth-weight infants and the levels of 1-minute Apgar score, 5-minute Apgar score in HIE group were significantly lower than those in control group, while the proportion of amniotic fluid contamination in the HIE group was significantly higher than those of the controls. When HIE was taken as the end point of 1-minute Apgar score, and 5-minute Apgar score, the cut-off value of 1-minute Apgar score was 3, and 5-minute Apgar score was 7 by receiver operating characteristic (ROC) curve analysis. The results of multivariate logistic regression analysis showed that low birth weight (<2.5 kg) (OR 1.780, 95% CI: 0.124-25.463, P=0.016), amniotic fluid contamination (OR 3.223, 95% CI: 1.049-9.901, P=0.041), low 1-minute Apgar score (≤3) (OR 92.425, 95% CI: 15.522-550.343, P<0.001), and low 5-minute Apgar score (≤7) (OR 12.641, 95% CI: 2.894-55.227, P=0.001) may increase risk of HIE. In addition, amniotic fluid contamination, low 1-minute Apgar score (≤3), and low 5-minute Apgar score (≤7) may increase risk of HIE among newborns born to women without previous childbearing history, but not in newborns born to women with previous childbearing history.

Conclusion

Low birth weight (<2.5 kg), amniotic fluid contamination, low 1-minute Apgar score (≤3), and 5-minute Apgar score (≤7) may increase risk of HIE.

Longitudinal perturbations of plasma nuclear magnetic resonance profiles in neonatal encephalopathy

BACKGROUND

Neonatal encephalopathy (NE) is a major cause of mortality and severe neurological disability in the neonatal period and beyond. We hypothesized that the degree of brain injury is reflected in the molecular composition of peripheral blood samples.

METHODS

A sub-cohort of 28 newborns included in the HYPOTOP trial was studied. Brain injury was assessed by magnetic resonance imaging (MRI) once per patient and neurodevelopment at 24 months of age was evaluated using the Bayley III Scales of Infant and Toddler Development. The nuclear magnetic resonance (NMR) profile of 60 plasma samples collected before, during, and after cooling was recorded.

RESULTS

In total, 249 molecular features were quantitated in plasma samples from newborns and postnatal age showed to affect detected NMR profiles. Lactate, beta-hydroxybutyrate, pyruvate, and three triglyceride biomarkers showed the ability to discern between different degrees of brain injury according to MRI scores. The prediction performance of lactate was superior as compared to other clinical and biochemical parameters.

CONCLUSIONS

This is the first longitudinal study of an ample compound panel recorded by NMR spectroscopy in plasma from NE infants. The serial determination of lactate confirms its solid position as reliable candidate biomarker for predicting the severity of brain injury.

IMPACT

The use of nuclear magnetic resonance (NMR) spectroscopy enables the simultaneous quantitation of 249 compounds in a small volume (i.e., 100 μL) of plasma. Longitudinal perturbations of plasma NMR profiles were linked to magnetic resonance imaging (MRI) outcomes of infants with neonatal encephalopathy (NE). Lactate, beta-hydroxybutyrate, pyruvate, and three triglyceride biomarkers showed the ability to discern between different degrees of brain injury according to MRI scores. Lactate is a minimally invasive candidate biomarker for early staging of MRI brain injury in NE infants that might be readily implemented in clinical guidelines for NE outcome prediction.