A systematic literature review on deep learning approaches for pneumonia detection using chest X-ray images

• Published in: Multimedia tools and applications Vol. 83; no. 8; pp. 24101 - 24151
• Main Authors: Sharma, Shagun, Guleria, Kalpna
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.03.2024; Springer Nature B.V
• Subjects: Artificial neural networks; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Deep learning; Feature extraction; Literature reviews; Machine learning; Multimedia Information Systems; Neural networks; Pneumonia; Special Purpose and Application-Based Systems; Systematic review; Track 2: Medical Applications of Multimedia; Deep learning; Pre-trained models; Ensemble models; Pneumonia; Convolutional neural network; Machine learning
• ISSN: 1573-7721; 1380-7501; 1573-7721
• DOI: 10.1007/s11042-023-16419-1

As per World Health Organization, in 2019, 2.5 million deaths were reported due to pneumonia, of which 14% were observed among children between 0–5 years of age. Due to the increased mortality rate, it is essential to diagnose pneumonia to avoid the failure of the human body's functioning. Machine and deep learning techniques can be implemented for pneumonia prediction, but deep learning is preferred over machine learning due to its applicability of better performance outcomes along with an automatic feature extraction from the dataset. This systematic literature review meticulously discusses a wide range of techniques for detecting pneumonia using deep learning, including convolutional neural networks, pre-trained models, and ensemble models. The review provides an in-depth illustration of architecture and working process and evaluates the effectiveness of these models in solving various medical domain challenges. It presents a summarization and analytical discussion on convolutional neural networks-based, pre-trained, and ensemble models offering a deep insight into several factors, including performance measures, hyperparameters, and fine-tuning of the models. This meta-analysis also discusses the highly robust and outperforming ensemble pneumonia detection models. Furthermore, the review highlights various research gaps in the existing models, and probable solutions, enabling a deeper understanding of their performance and suitability for pneumonia detection tasks.