Emerging multifaceted application of artificial intelligence in chest radiography: a narrative review

Nwaiwu, Victor Chigbundu and Das, Sreemoy Kanti (2024) Emerging multifaceted application of artificial intelligence in chest radiography: a narrative review. Journal of Medical Artificial Intelligence, 7 (39). ISSN 2617-2496

Text Nwaiwu V 2024 JMAI full article.pdf - Published Version Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (255kB)

Abstract

Background and Objective: Chest radiography, otherwise known as chest X-ray (CXR) is the most in demand and widely performed investigation in radiology department; owing to the multiple combining effect of rise in prevalence of respiratory diseases globally and the growing need of heath assessment for pre-employment, pre-operative and migration purposes. However, this task is already proving overwhelming, placing an immeasurable burden of workload on radiographers, radiologist, and the entire health system; this has resulted in long waiting time, fatigue-based technical error, interpretation error, reporting delays and backlogs. To ameliorate this predicament, medical imaging has witnessed the introduction of artificial intelligence (AI). Thus, with the raid evolutionary trend in technology, this article seeks to review current state of evidence on AI use in CXR and level of progress made to minimize these errors and delays. In addition, point out challenges, as well as unfold areas for future research to better detection rates and improve overall clinical outcomes. Methods: A search for relevant literature that focuses on AI in CXR was conducted with the help of certain keywords [machine learning (ML), chest radiography, deep learning (DL), natural language processing (NLP), expert system (ES) and fuzzy logic (FL)]. Thereafter, a narrative logical approach to technically analysing and synthesizing findings across domains of AI (ML, DL, NLP, ES, FL) and robot technologies as it relates to CXR done. Key Content and Findings: A thorough evaluation of the substance of evidence these studies bring to enhance overall workflow and health outcomes show that ML is very useful in performing administrative and imaging tasks such as exam scheduling, worklist management and image acquisition. On the other hand, DL is better suited for classification tasks on a broad spectrum of chest anomalies in CXR. However, a hybrid approach involving ML-DL, FL-DL and NLP-DL/ML technologies seems to further improve reporting accuracy and offer more insights into CXR interpretation. Further studies on training and refining models for clinical use in this perspective is demanded. Conclusions: AI still in its early stages; this review to serve as road map to implementation and policy making, guide routine practice and improve clinical governance.

Item Type:	Article
Depositing User:	Bridget Roberts
Date Deposited:	07 Aug 2025 15:41
Last Modified:	11 Aug 2025 12:10
URI:	https://hsu.repository.guildhe.ac.uk/id/eprint/545

Actions (login required)

Edit Item