JMIR Publications

ABSTRACT

Background:

Tuberculosis (TB) was the leading infectious cause of mortality globally prior to COVID-19 and chest radiography has an important role in the detection, and subsequent diagnosis, of patients with this disease. The conventional experts (radiologists) reading has substantial within-and between-observer variability, indicating poor reliability of human readers. There has been substantial work in utilising various Artificial Intelligence (AI)-based algorithms to address the limitations of human reading of chest radiographs for diagnosing TB.

Objective:

This systematic literature review (SLR) aims to assess the role of Machine Learning (ML) and Deep Learning (DL) in the detection of tuberculosis (TB) using chest radiography (CXR).

Methods:

In conducting and reporting the SLR, we followed the Preferred Reporting Items for Systematic reviews and Meta-Analysis (PRISMA) Guidelines. A total of 309 records was identified from Scopus, PubMed, and IEEE. We independently screened, reviewed, and assessed all available records and included 47 studies that met the inclusion criteria in this SLR.

Results:

DL is more commonly used than ML in the included studies. Most studies used human radiologist’s report as the reference standard. Support Vector Machine (SVM), k-Nearest Neighbours (kNN), and Random Forest (RF) were the most popular ML approaches. Meanwhile, Convolutional Neural Networks (CNNs) were the most commonly used DL techniques, with the four most popular applications being ResNet-50, VGG-16, VGG-19, and AlexNet. Based on data from ten studies we estimated the pooled sensitivity of ML/DL methods was 0.9857 (95% CI: 0.9477-1.00) and a pooled specificity was 0.9805 (95% CI: 0.9255-1.00).

Conclusions:

Findings from this SLR confirm the high potential of both ML and DL for TB detection using CXR.