Complete Exploratory Analysis of Phases of Respiration with 3D Chest CT Scans using Deep Learning
Keywords
Loading...
Authors
Issue Date
2022-06-14
Language
en
Document type
Journal Title
Journal ISSN
Volume Title
Publisher
Title
ISSN
Volume
Issue
Startpage
Endpage
DOI
Abstract
Thoracic CT scans are used for detection of various pulmonary diseases. Two types of scans are used
to detect problems with the lungs, scans in the inspiration phase and in the expiration phase. However,
often, the phase information is missing or misclassified in these scans because of human error. This thesis
introduces a 3D convolutional neural network (CNN) architecture to classify between these respiration
scans. Themodel takes 3D full volumes of the scans as an input to the 3D model, which reaches a final test
classification accuracy of 96.5%. The most important biomarkers or features which the model pays attention
to are also visualised, providing information about how the model reaches the classification results. These
visualisations can be explained well, that is, a trustworthy AI has been constructed. Due to a gap in literature,
the proposed model acts as a benchmark for classification between phases of respiration using deep learning
methods.
Keywords: thoracic CT scans, inspiration, expiration, 3D Convolutional Networks
Description
Citation
Faculty
Faculteit der Sociale Wetenschappen