Document Type
Thesis
Date of Award
2022
Degree Name
Master of Science (MS)
Department
Computer Science
First Advisor
KC Santosh
Abstract
The presence of non-biomedical foreign objects (NBFO) such as coins, buttons, jewelry, etc. and biomedical foreign objects (BFO) such as medical tubes, and devices in Chest X-Rays (CXRs) make accurate interpretation difficult as they do not indicate known biological abnormalities like excess fluids, Tuberculosis (TB) or cysts. Accurate diagnosis and screening, require these NBFO and BFO to be detected, categorized as either NBFO or BFO, and removed from CXR or highlighted in CXR for effective abnormality analysis. During an automated CXR screening process, NBFOs can adversely impact the process as typical machine learning algorithms would consider these objects to be biological abnormalities producing false-positive cases. In cases where the NBFO is not detected (metallic candy wrappers), the test would produce a false negative and could have a significant impact on the patient if overlooked. This paper examines detailed discussions on numerous clinical reports in addition to Computer-Aided Detection (CADe) with Diagnosis tools (CADx), where both shallow learning and deep learning algorithms are applied. Our discussion reflects the importance of accurately detecting, isolating, classifying, and either removing or highlighting NBFOs in CXRs by taking 28 peer-reviewed research reports and articles into account. In our work, we employ You Only Look Once (YOLOv4) algorithm – a Deep Neural Network – to detect foreign objects in CXR images. Considering its genericity, on a dataset of 400 publicly available CXR images hosted by LHNCBC, U.S National Library of Medicine (NLM), National Institutes of Health (NIH), we achieve the following performance scores: accuracy of 91.00%, precision of 85.00%, recall of 93.00% and f1-score of 89.00%. Unlike state-of-artworks, where they are limited to a specific type of foreign object (e.g., circle-like objects), this is the first time we report experimental results on all possible types of foreign objects.
Subject Categories
Computer Sciences
Keywords
AI-guided tools, Chest x-ray, foreign objects, pulmonary abnormality, YOLOv4
Number of Pages
81
Publisher
University of South Dakota
Recommended Citation
Roy, Shotabdi, "Automated Chest X-ray Analysis: Biomedical/Non-biomedical foreign object detection" (2022). Dissertations and Theses. 80.
https://red.library.usd.edu/diss-thesis/80