Keywords:
Infection, Computer Applications-General, Digital radiography, Conventional radiography, Respiratory system, Artificial Intelligence
Authors:
T. Fung1, J. W. Luo2, T. C. Lee3, B. Gallix2, J. J. R. Chong2; 1Montreal, Québec/CA, 2Montreal, QC/CA, 3Montreal/CA
DOI:
10.26044/ecr2019/C-3416
Aims and objectives
Tuberculosis (TB) is a leading cause of death worldwide and while there are effective ways to prevent,
detect and cure the disease,
it remains a significant global public health challenge that has sparked the concern of the World Health Organization [5].
The use of chest radiographs (CXR) has long been part of the triaging,
screening and diagnostic strategy for TB [6].
However,
despite improved screening methods,
detection technologies and curative treatment,
the burden of TB remains a huge issue.
While the development of advanced modern technologies such as digital CXRs helped shorten the processing time for the images,
inadequate human resources limit the time to interpretation [3].
Inadequate human and financial resources mean delayed detection and treatment of disease,
resulting in worse outcomes and further spread among low-resourced communities globally.
The integration of computer-aided diagnosis (CAD) in the diagnostic process can reduce the burden of interpretation and improve workflow efficiency.
A systematic review conducted by Pande et al.
(2016) analyzed five studies evaluating CAD4TB,
a commercially available TB screening software,
and its ability to detect TB on CXRs [4].
The area under the curve (AUC) of these studies ranged from 0.71 to 0.84 and demonstrate some of the earliest attempts to utilize artificial intelligence (AI) in initial radiographic TB detection.
Since then,
other studies have been designed yielding improved detection rates.
Hwang et al.
(2016) tested a TB screening system using a pre-trained neural networks and transfer learning which yielded AUCs ranging from 0.88 to 0.96 [2].
More recently,
Lakhani & Sundaram (2017) demonstrated that deep convolutional neural networks such as AlexNet and GoogLeNet may detect TB with an AUC of 0.99 [3].
These studies,
along with several others,
highlight the potentially significant role AI could play towards the global goal of TB eradication and show promising performance in this regard.
We propose a deep-learning based algorithm to detect and help triage radiographs that could improve the time to interpretation and warn both clinicians and radiologists to this critical finding,
and to evaluate objective and subjective performance of such a system.