Pakzad, Ashkan;
(2023)
Deep Learning for Morphometric Analysis of Airways in Computed Tomography.
Doctoral thesis (Ph.D), UCL (University College London).
Preview |
Text
Pakzad_PhD_Thesis_deposit.pdf - Other Download (38MB) | Preview |
Abstract
Chronic lung diseases are often characterised by dilatation, wall thickening, contortion and plugging of airways. Lung function tests are the gold-standard measure to assess disease severity and progression, but are associated with a degree of measurement variability. Computed tomography (CT) scans of the lungs can visualise lung damage, including airway damage. However visual CT evaluation is time-consuming and prone to intra and inter-observer variability. We proposed a novel airway skeletonisation method based on region-growing exploration with minimum cost-path search. To facilitate analysis, we built and released an open-source airway analysis framework, AirQuant which defines an airway segment as the distance between airway bifurcations and endpoints. We also propose a synthetic supervised deep-learning based model, termed airway transfer network (ATN) that transforms simulated airways to mimic target CT airway appearances by perceptual losses. We finally consider our proposed technical methods in a clinical analysis of CT of patients with idiopathic pulmonary fibrosis (IPF), a chronic fibrosing lung disease characterised by abnormal airway dilatation. Our novel skeletonisation method outperformed a state-of-the-art technique on an open-source airway dataset. IPF airways demonstrated significantly lower inter-segmental tapering and significantly greater tortuosity compared to healthy never-smokers. Our ATN framework also outperformed a state-of-the-art generative airway measurement model, clinically evaluated against an objective endpoint of mortality. Median global inter-segmental tapering was an independent predictor of survival alongside lung function tests across two independent cohorts of IPF patients. AirQuant represents a versatile automated airway analysis framework which could be applied to cohort enrichment in clinical trial recruitment with potential applications beyond IPF such as cystic fibrosis and chronic obstructive pulmonary disease. AirQuant also has potential to analyse other macro-tubular branching structures such as blood vessels.
| Type: | Thesis (Doctoral) |
|---|---|
| Qualification: | Ph.D |
| Title: | Deep Learning for Morphometric Analysis of Airways in Computed Tomography |
| Open access status: | An open access version is available from UCL Discovery |
| Language: | English |
| Additional information: | Copyright © The Author 2023. Original content in this thesis is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) Licence (https://creativecommons.org/licenses/by-nc/4.0/). Any third-party copyright material present remains the property of its respective owner(s) and is licensed under its existing terms. Access may initially be restricted at the author’s request. |
| Keywords: | Deep Learning, Computed Tomography, Lung, Airways, Idiopathic Pulmonary Fibrosis, Fibrosis, synthetic data generation, skeletonisation, x-ray CT, machine learning |
| UCL classification: | UCL UCL > Provost and Vice Provost Offices > UCL BEAMS UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science > Dept of Med Phys and Biomedical Eng |
| URI: | https://discovery-pp.ucl.ac.uk/id/eprint/10174666 |
Archive Staff Only
 |
View Item |
