Prediction of periventricular leukomalacia occurrence in neonates after heart surgery

High-Resolution and Multidimensional Phenotypes Can Complement Genomics Data to Diagnose Diseases in the Neonatal Population

Advances in genomic medicine have greatly improved our understanding of human diseases. However, phenome is not well understood. High-resolution and multidimensional phenotypes have shed light on the mechanisms underlying neonatal diseases in greater details and have the potential to optimize clinical strategies. In this review, we first highlight the value of analyzing traditional phenotypes using a data science approach in the neonatal population. We then discuss recent research on high-resolution, multidimensional, and structured phenotypes in neonatal critical diseases. Finally, we briefly introduce current technologies available for the analysis of multidimensional data and the value that can be provided by integrating these data into clinical practice. In summary, a time series of multidimensional phenome can improve our understanding of disease mechanisms and diagnostic decision-making, stratify patients, and provide clinicians with optimized strategies for therapeutic intervention; however, the available technologies for collecting multidimensional data and the best platform for connecting multiple modalities should be considered.

A machine learning approach to predict quality of life changes in patients with Parkinson's Disease

OBJECTIVE

Parkinson disease (PD) is a progressive neurodegenerative disorder with an annual incidence of approximately 0.1%. While primarily considered a motor disorder, increasing emphasis is being placed on its non-motor features. Both manifestations of the disease affect quality of life (QoL), which is captured in part II of the Unified Parkinson's Disease Rating Scale (UPDRS-II). While useful in the management of patients, it remains challenging to predict how QoL will change over time in PD. The goal of this work is to explore the feasibility of a machine learning algorithm to predict QoL changes in PD patients.

METHODS

In this retrospective cohort study, patients with at least 12 months of follow-up were identified from the Parkinson's Progression Markers Initiative database (N = 630) and divided into two groups: those with and without clinically significant worsening in UPDRS-II (n = 404 and n = 226, respectively). We developed an artificial neural network using only UPDRS-II scores, to predict whether a patient would clinically worsen or not at 12 months from follow-up.

RESULTS

Using UPDRS-II at baseline, at 2 months, and at 4 months, the algorithm achieved 90% specificity and 56% sensitivity.

INTERPRETATION

A learning model has the potential to rule in patients who may exhibit clinically significant worsening in QoL at 12 months. These patients may require further testing and increased focus.

A Novel Embedded Feature Selection and Dimensionality Reduction Method for an SVM Type Classifier to Predict Periventricular Leukomalacia (PVL) in Neonates

This paper is concerned with the prediction of the occurrence of periventricular leukomalacia (PVL) in neonates after heart surgery. Our prior work shows that the Support Vector Machine (SVM) classifier can be a powerful tool in predicting clinical outcomes of such complicated and uncommon diseases, even when the number of data samples is low. In the presented work, we first illustrate and discuss the shortcomings of the traditional automatic machine learning (aML) approach. Consequently, we describe our methodology for addressing these shortcomings, while utilizing the designed interactive ML (iML) algorithm. Finally, we conclude with a discussion of the developed method and the results obtained. In sum, by adding an additional (Genetic Algorithm) optimization step in the SVM learning framework, we were able to (a) reduce the dimensionality of an SVM model from 248 to 53 features, (b) increase generalization that was confirmed by a 100% accuracy assessed on an unseen testing set, and (c) improve the overall SVM model's performance from 65% to 100% testing accuracy, utilizing the proposed iML method.

Machine Learning Applied to Registry Data: Development of a Patient-Specific Prediction Model for Blood Transfusion Requirements During Craniofacial Surgery Using the Pediatric Craniofacial Perioperative Registry Dataset

BACKGROUND

Craniosynostosis is the premature fusion of ≥1 cranial sutures and often requires surgical intervention. Surgery may involve extensive osteotomies, which can lead to substantial blood loss. Currently, there are no consensus recommendations for guiding blood conservation or transfusion in this patient population. The aim of this study is to develop a machine-learning model to predict blood product transfusion requirements for individual pediatric patients undergoing craniofacial surgery.

METHODS

Using data from 2143 patients in the Pediatric Craniofacial Surgery Perioperative Registry, we assessed 6 machine-learning classification and regression models based on random forest, adaptive boosting (AdaBoost), neural network, gradient boosting machine (GBM), support vector machine, and elastic net methods with inputs from 22 demographic and preoperative features. We developed classification models to predict an individual's overall need for transfusion and regression models to predict the number of blood product units to be ordered preoperatively. The study is reported according to the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) checklist for prediction model development.

RESULTS

The GBM performed best in both domains, with an area under receiver operating characteristic curve of 0.87 ± 0.03 (95% confidence interval) and F-score of 0.91 ± 0.04 for classification, and a mean squared error of 1.15 ± 0.12, R-squared (R) of 0.73 ± 0.02, and root mean squared error of 1.05 ± 0.06 for regression. GBM feature ranking determined that the following variables held the most information for prediction: platelet count, weight, preoperative hematocrit, surgical volume per institution, age, and preoperative hemoglobin. We then produced a calculator to show the number of units of blood that should be ordered preoperatively for an individual patient.

CONCLUSIONS

Anesthesiologists and surgeons can use this continually evolving predictive model to improve clinical care of patients presenting for craniosynostosis surgery.

Deep Learning for Improved Risk Prediction in Surgical Outcomes

The Norwood surgical procedure restores functional systemic circulation in neonatal patients with single ventricle congenital heart defects, but this complex procedure carries a high mortality rate. In this study we address the need to provide an accurate patient specific risk prediction for one-year postoperative mortality or cardiac transplantation and prolonged length of hospital stay with the purpose of assisting clinicians and patients' families in the preoperative decision making process. Currently available risk prediction models either do not provide patient specific risk factors or only predict in-hospital mortality rates. We apply machine learning models to predict and calculate individual patient risk for mortality and prolonged length of stay using the Pediatric Heart Network Single Ventricle Reconstruction trial dataset. We applied a Markov Chain Monte-Carlo simulation method to impute missing data and then fed the selected variables to multiple machine learning models. The individual risk of mortality or cardiac transplantation calculation produced by our deep neural network model demonstrated 89 ± 4% accuracy and 0.95 ± 0.02 area under the receiver operating characteristic curve (AUROC). The C-statistics results for prediction of prolonged length of stay were 85 ± 3% accuracy and AUROC 0.94 ± 0.04. These predictive models and calculator may help to inform clinical and organizational decision making.

Neural Network Classifier for Automatic Detection of Invasive Versus Noninvasive Airway Management Technique Based on Respiratory Monitoring Parameters in a Pediatric Anesthesia

Children undergoing general anesthesia require airway monitoring by an anesthesia provider. The airway may be supported with noninvasive devices such as face mask or invasive devices such as a laryngeal mask airway or an endotracheal tube. The physiologic data stored provides an opportunity to apply machine learning algorithms distinguish between these modes based on pattern recognition. We retrieved three data sets from patients receiving general anesthesia in 2015 with either mask, laryngeal mask airway or endotracheal tube. Patients underwent myringotomy, tonsillectomy, adenoidectomy or inguinal hernia repair procedures. We retrieved measurements for end-tidal carbon dioxide, tidal volume, and peak inspiratory pressure and calculated statistical features for each data element per patient. We applied machine learning algorithms (decision tree, support vector machine, and neural network) to classify patients into noninvasive or invasive airway device support. We identified 300 patients per group (mask, laryngeal mask airway, and endotracheal tube) for a total of 900 patients. The neural network classifier performed better than the boosted trees and support vector machine classifiers based on the test data sets. The sensitivity, specificity, and accuracy for neural network classification are 97.5%, 96.3%, and 95.8%. In contrast, the sensitivity, specificity, and accuracy of support vector machine are 89.1%, 92.3%, and 88.3% and with the boosted tree classifier they are 93.8%, 92.1%, and 91.4%. We describe a method to automatically distinguish between noninvasive and invasive airway device support in a pediatric surgical setting based on respiratory monitoring parameters. The results show that the neural network classifier algorithm can accurately classify noninvasive and invasive airway device support.

Automatic Detection of Endotracheal Intubation During the Anesthesia Procedure

This paper is concerned with the mathematical modeling and detection of endotracheal (ET) intubation in children under general anesthesia during surgery. In major pediatric surgeries, the airway is often secured with an endotracheal tube (ETT) followed by initiation of mechanical ventilation. Clinicians utilize auscultation of breath sounds and capnography to verify correct ETT placement. However, anesthesia providers often delay timely charting of ET intubation. This latency in event documentation results in decreased efficacy of clinical decision support systems. In order to target this problem, we collected real inpatient data and designed an algorithm to accurately detect the intubation time within the clinically valid range; the results show that we are able to achieve high accuracy in more than 96% of the cases. Automatic detection of ET intubation time would thus enhance better real-time data capture to support future improvement in clinical decision support systems.

Advanced analytics for outcome prediction in intensive care units

In this paper we present a new expert knowledge based clinical decision support system for prediction of intensive care units outcome based on the physiological measurements collected during the first 48 hours of the patient's admission to the ICU. The developed CDSS algorithm is composed of several stages. First, we categorize the collected data based on the physiological organ that they represent. We then extract clinically relevant features from each data category and then rank these features based on their mutual information with the outcome. Then, we design an artificial neural network to serve as a classifier to detect patients at high risk of critical deterioration. We use the eight-fold cross validation method to test the developed CDSS classifier. The results from the classification show that the newly designed CDSS outperforms the widely used acuity scoring systems, SOFA and SAPS-III. The F-score classification result of our developed algorithms is 42% while the F-score results for SOFA and SAPS-III are 26% and 29% respectively.

Application of Mathematical Modeling for Simulation and Analysis of Hypoplastic Left Heart Syndrome (HLHS) in Pre- and Postsurgery Conditions

This paper is concerned with the mathematical modeling of a severe and common congenital defect called hypoplastic left heart syndrome (HLHS). Surgical approaches are utilized for palliating this heart condition; however, a brain white matter injury called periventricular leukomalacia (PVL) occurs with high prevalence at or around the time of surgery, the exact cause of which is not known presently. Our main goal in this paper is to study the hemodynamic conditions under which HLHS physiology may lead to the occurrence of PVL. A lumped parameter model of the HLHS circulation has been developed integrating diffusion modeling of oxygen and carbon dioxide concentrations in order to study hemodynamic variables such as pressure, flow, and blood gas concentration. Results presented include calculations of blood pressures and flow rates in different parts of the circulation. Simulations also show changes in the ratio of pulmonary to systemic blood flow rates when the sizes of the patent ductus arteriosus and atrial septal defect are varied. These changes lead to unbalanced blood circulations and, when combined with low oxygen and carbon dioxide concentrations in arteries, result in poor oxygen delivery to the brain. We stipulate that PVL occurs as a consequence.

Scoring system for periventricular leukomalacia in infants with congenital heart disease

BACKGROUND

Currently two magnetic resonance imaging (MRI) methods have been used to assess periventricular leukomalacia (PVL) severity in infants with congenital heart disease: manual volumetric lesion segmentation and an observational categorical scale. Volumetric classification is labor intensive and the categorical scale is quick but unreliable. We propose the quartered point system (QPS) as a novel, intuitive, time-efficient metric with high interrater agreement.

METHODS

QPS is an observational scale that asks the rater to score MRIs on the basis of lesion size, number, and distribution. Pre- and postoperative brain MRIs were obtained on term congenital heart disease infants. Three independent observers scored PVL severity using all three methods: volumetric segmentation, categorical scale, and QPS.

RESULTS

One-hundred and thirty-five MRIs were obtained from 72 infants; PVL was seen in 48 MRIs. Volumetric measurements among the three raters were highly concordant (ρc = 0.94-0.96). Categorical scale severity scores were in poor agreement between observers (κ = 0.17) and fair agreement with volumetrically determined severity (κ = 0.26). QPS scores were in very good agreement between observers (κ = 0.82) and with volumetric severity (κ = 0.81).

CONCLUSION

QPS minimizes training and sophisticated radiologic analysis and increases interrater reliability. QPS offers greater sensitivity to stratify PVL severity and has the potential to more accurately correlate with neurodevelopmental outcomes.

Computerized Clinical Decision Support: Contributions from 2014

OBJECTIVE

To summarize recent research and propose a selection of best papers published in 2014 in the field of computerized clinical decision support for the Decision Support section of the IMIA yearbook.

METHOD

A literature review was performed by searching two bibliographic databases for papers related to clinical decision support systems (CDSSs) and computerized provider order entry systems in order to select a list of candidate best papers to be then peer-reviewed by external reviewers. A consensus meeting between the two section editors and the editorial team was finally organized to conclude on the selection of best papers.

RESULTS

Among the 1,254 returned papers published in 2014, the full review process selected four best papers. The first one is an experimental contribution to a better understanding of unintended uses of CDSSs. The second paper describes the effective use of previously collected data to tailor and adapt a CDSS. The third paper presents an innovative application that uses pharmacogenomic information to support personalized medicine. The fourth paper reports on the long-term effect of the routine use of a CDSS for antibiotic therapy.

CONCLUSIONS

As health information technologies spread more and more meaningfully, CDSSs are improving to answer users' needs more accurately. The exploitation of previously collected data and the use of genomic data for decision support has started to materialize. However, more work is still needed to address issues related to the correct usage of such technologies, and to assess their effective impact in the long term.