Machine learning-based approaches for distinguishing viral and bacterial pneumonia in paediatrics: A scoping review DOI Creative Commons

Diana G. Rickard,

Muhammad Ashad Kabir, Nusrat Homaira

et al.

Computer Methods and Programs in Biomedicine, Journal Year: 2025, Volume and Issue: 268, P. 108802 - 108802

Published: May 8, 2025

Pneumonia is the leading cause of hospitalisation and mortality among children under five, particularly in low-resource settings. Accurate differentiation between viral bacterial pneumonia essential for guiding appropriate treatment, yet it remains challenging due to overlapping clinical radiographic features. Advances machine learning (ML), deep (DL), have shown promise classifying using chest X-ray (CXR) images. This scoping review summarises evidence on ML techniques CXR images paediatric patients. was conducted following Joanna Briggs Institute methodology PRISMA-ScR guidelines. A comprehensive search performed PubMed, Embase, Scopus identify studies involving (0-18 years) with diagnosed through CXR, models binary or multiclass classification. Data extraction included models, dataset characteristics, performance metrics. total 35 studies, published 2018 2025, were this review. Of these, 31 used publicly available Kermany dataset, raising concerns about overfitting limited generalisability broader, real-world populations. Most (n=33) convolutional neural networks (CNNs) While many demonstrated promising performance, significant variability observed differences methodologies, sizes, validation strategies, complicating direct comparisons. For classification (viral vs pneumonia), a median accuracy 92.3% (range: 80.8% 97.9%) reported. (healthy, pneumonia, 91.8% 76.8% 99.7%). Current constrained by predominant reliance single which limit applicability findings. To address these limitations, future research should focus developing diverse representative datasets while adhering standardised reporting Such efforts are improve reliability, reproducibility, translational potential

Language: Английский

Machine learning-based approaches for distinguishing viral and bacterial pneumonia in paediatrics: A scoping review DOI Creative Commons

Diana G. Rickard,

Muhammad Ashad Kabir, Nusrat Homaira

et al.

Computer Methods and Programs in Biomedicine, Journal Year: 2025, Volume and Issue: 268, P. 108802 - 108802

Published: May 8, 2025

Pneumonia is the leading cause of hospitalisation and mortality among children under five, particularly in low-resource settings. Accurate differentiation between viral bacterial pneumonia essential for guiding appropriate treatment, yet it remains challenging due to overlapping clinical radiographic features. Advances machine learning (ML), deep (DL), have shown promise classifying using chest X-ray (CXR) images. This scoping review summarises evidence on ML techniques CXR images paediatric patients. was conducted following Joanna Briggs Institute methodology PRISMA-ScR guidelines. A comprehensive search performed PubMed, Embase, Scopus identify studies involving (0-18 years) with diagnosed through CXR, models binary or multiclass classification. Data extraction included models, dataset characteristics, performance metrics. total 35 studies, published 2018 2025, were this review. Of these, 31 used publicly available Kermany dataset, raising concerns about overfitting limited generalisability broader, real-world populations. Most (n=33) convolutional neural networks (CNNs) While many demonstrated promising performance, significant variability observed differences methodologies, sizes, validation strategies, complicating direct comparisons. For classification (viral vs pneumonia), a median accuracy 92.3% (range: 80.8% 97.9%) reported. (healthy, pneumonia, 91.8% 76.8% 99.7%). Current constrained by predominant reliance single which limit applicability findings. To address these limitations, future research should focus developing diverse representative datasets while adhering standardised reporting Such efforts are improve reliability, reproducibility, translational potential

Language: Английский

Citations

0