A Newborn Screening System Based on Service-Oriented Architecture Embedded Support Vector Machine

Abstract

The clinical symptoms of metabolic disorders are rarely apparent during the neonatal period, and if they are not treated early, irreversible damage, such as mental retardation, or even death, may occur. Therefore, the practice of newborn screening is important to the prevention of permanent disabilities in newborns.In this paper, we propose a newborn screening system that uses Support Vector Machine (SVM) classification techniques to evaluate raw data on the concentrations of metabolic substances obtained from tandem mass spectrometry (MS/MS) to determine if a newborn has a metabolic disorder. As per the concept of Service-Oriented Architecture (SOA), we designed the system using web services techniques, which are suitable for integrating heterogeneous platforms, protocols, and applications. It can be integrated seamlessly into National Taiwan University Hospital Information Systems (HIS) or other related HIS .

Opportunities and challenges in machine learning‐based newborn screening—A systematic literature review

“…For parameter optimization, grid search is commonly applied 17,19 and iterates through a set of parameter combinations returning the combination with the best performance. To test the robustness of the methods and estimate the performance on different subsets, cross‐validation 17,21 or stratified cross‐validation 19 and evaluation of receiver operating characteristic curves 25 is applied. The classification performance is evaluated using classification sensitivity, specificity, and PPV, which are calculated using the amount of true positive (TP), false positive (FP), true negative (TN), and false negative (FN) predicted patients, …”

“…For parameter optimization, grid search is commonly applied 17,19 and iterates through a set of parameter combinations returning the combination with the best performance. To test the robustness of the methods and estimate the performance on different subsets, cross-validation 17,21 or stratified cross-validation 19 and evaluation of receiver operating characteristic curves 25 is applied.…”

“…The main reasons for dismissing papers were that they did not apply ML classification methods, 28 investigated other diseases from NBS programs such as hearing disabilities 29 or did not use data obtained from MS/MS. 30 Publications from different screening centers in Europe, [12][13][14][15][16]19 Asia, 7,17,18,20,21,23,24,26 and North America 22,25,27 are reviewed in this work.…”

“…Other methods such as K-nearest neighbor 7,14,16 consider the distance between data points and identify clusters of healthy and sick patients within the data. Support vector machines (SVMs) 7,14,[16][17][18]20,21,24 attempt to find the largest separating band between sick and healthy patients by transforming the features into a higher dimensional space with linear kernels, 7,16 radial basis function, 16,18,20,21 or polynomial (degree 2 or 3) 14,16 kernels. Linear discriminant analysis 7,14 is a linear classifier that aims to find a discriminant line (plane or hyperplane) by fitting weights of features that are optimal for maximizing the between-class variance.…”

A Data-Mining Framework for Transnational Healthcare System

“…Due to SOA [20], which was already adopted by NTUH [21], most of the system functions are implemented as web services. It becomes very convenient to expand the system modules and to integrate with other heterogeneous systems.…”