An Entropy-Based Architecture for Detection of Sepsis in Newborn Cry Diagnostic Systems

Abstract

The acoustic characteristics of cries are an exhibition of an infant’s health condition and these characteristics have been acknowledged as indicators for various pathologies. This study focused on the detection of infants suffering from sepsis by developing a simplified design using acoustic features and conventional classifiers. The features for the proposed framework were Mel-frequency Cepstral Coefficients (MFCC), Spectral Entropy Cepstral Coefficients (SENCC) and Spectral Centroid Cepstral Coefficients (SCCC), which were classified through K-nearest Neighborhood (KNN) and Support Vector Machine (SVM) classification methods. The performance of the different combinations of the feature sets was also evaluated based on several measures such as accuracy, F1-score and Matthews Correlation Coefficient (MCC). Bayesian Hyperparameter Optimization (BHPO) was employed to tailor the classifiers uniquely to fit each experiment. The proposed methodology was tested on two datasets of expiratory cries (EXP) and voiced inspiratory cries (INSV). The highest accuracy and F-score were 89.99% and 89.70%, respectively. This framework also implemented a novel feature selection method based on Fuzzy Entropy (FE) as a final experiment. By employing FE, the number of features was reduced by more than 40%, whereas the evaluation measures were not hindered for the EXP dataset and were even enhanced for the INSV dataset. Therefore, it was deduced through these experiments that an entropy-based framework is successful for identifying sepsis in neonates and has the advantage of achieving high performance with conventional machine learning (ML) approaches, which makes it a reliable means for the early diagnosis of sepsis in deprived areas of the world.

Newborn Cry-Based Diagnostic System to Distinguish between Sepsis and Respiratory Distress Syndrome Using Combined Acoustic Features

“…The study presented by Matikolaie et al [ 18 ] investigated the role of prosodical characterization of the cry signal in detecting sepsis which accomplished 86% as their best F-score. Furthermore, Khalilzad et al [ 19 ] explored the potential of a NCDS in diagnosing sepsis by incorporating entropy-based features and fuzzy entropy feature selection, which attained 89.70% as their best F-score for the expiration cry segments. We believed that with sepsis being one of the globally leading post-partum mortality causes, there is a need for more in-depth studies that probe other perspectives of this pathology.…”

“…The GFCC feature set was also used as a more robust alternative to the MFCCs that are the most prevalent in the field of audio processing applications. It was shown in [ 18 , 19 ] that the combination of short-term and spectral features provide better classification performance for the study of RDS and sepsis. Furthermore, feature fusion was shown to enhance the performance of the diagnostic system designs for depression data [ 53 ] and artifact rejection in neuroimaging data [ 54 ] by playing a significant role in enhancing the linear separability through constructing the apropos feature set.…”

“…The accuracy owes its prevalence to simplicity in calculation and understanding, but it is not informative in terms of class assessment and missed cases; therefore, other measures were introduced and studied. Table 2 presents a number of these measures used in this study [ 19 ].…”

“…One can see that researchers continue to apply new machine learning methods to classify infant cry signals into normal and pathological records; for example, see the recent works in [ 11 , 12 ]. However, some of the current research works include identifying pathologies such as hypo-acoustic [ 13 ], asphyxia [ 14 , 15 , 16 ], hypothyroidism [ 17 ], septic [ 18 , 19 ], RDS [ 20 ], and autism spectrum disorder (ASD) [ 21 ]; additionally the authors in [ 8 , 22 , 23 , 24 ] have investigated different infant pathologies. In particular, the asphyxiated infant crying signals have been identified using different ML methods, including a deep feedforward neural network (DFNN) model [ 14 ], a support vector machine (SVM) model [ 15 ], and a convolutional neural network (CNN) approach [ 16 ], and achieved accuracy rates of 96.74%, 98.5%, and 92.8%, respectively.…”

“…In addition, hypothyroidism has been studied in [ 17 ] using a Multilayer Perceptron (MLP) classifier, and achieved a classification accuracy of 88.12%. Two groups of authors investigated sepsis in newborns recently; the authors in [ 18 , 19 ] have developed a machine learning-based CDS for identifying septic newborns and reached an accuracy of 83.9% using majority voting, while the authors in [ 19 ] attained 89.99% using entropy-based features. Furthermore, ASD in [ 21 ] and RDS in [ 20 ] have been identified based on a SVM and reached accuracies of 96% and 73.8%, respectively.…”