A Systematic Review of the Accuracy of Machine Learning Models for Diagnosing Pulmonary Tuberculosis: Implications for Nursing Practice and Implementation.

Abstract

This systematic review evaluates the application of machine learning (ML) models for diagnosing pulmonary tuberculosis and their potential to inform nursing practice and implementation strategies. Studies published between 2019 and 2024 were systematically identified through searches in Scopus, PubMed, Medline, ScienceDirect, CINAHL Plus with Full Text, Clinical Key, Ovid, EMBASE, and Web of Science. The review adhered to PRISMA guidelines, with rigorous inclusion and exclusion criteria applied. A total of 734 records were retrieved, with 18 duplicates removed, leaving 716 articles for screening. Of these, 699 did not meet the inclusion criteria. Full-text review of 17 articles excluded five studies, resulting in 12 studies included in the final analysis. The synthesis revealed five key diagnostic features commonly utilized in ML models: chest x-rays, computed tomography scans, sputum smear images, human exhaled breath, and personal information. Among 13 identified ML algorithms, convolutional neural networks were the most frequently employed due to their superior performance in analyzing imaging data. These findings emphasize the transformative potential of ML technologies to enhance early tuberculosis diagnosis, optimize nursing practice, and improve clinical outcomes.