Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.791971
Title: Detection and description of pulmonary nodules through 2D and 3D clustering
Author: Al-Funjan, A.
ISNI:       0000 0004 8504 5274
Awarding Body: University of Salford
Current Institution: University of Salford
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
Precise 3D automated detection, description and classification of pulmonary nodules offer the potential for early diagnosis of cancer and greater efficiency in the reading of computerised tomography (CT) images. CT scan centres are currently experiencing high loads and experts shortage, especially in developing countries such as Iraq where the results of the current research will be used. This motivates the researchers to address these problems and challenges by developing automated processes for the early detection and efficient description of cancer cases. This research attempts to reduce workloads, enhance the patient throughput and improve the diagnosis performance. To achieve this goal, the study selects techniques for segmentation, classification, detection and implements the best candidates alongside a novel automated approach. Techniques for each stage in the process are quantitatively evaluated to select the best performance against standard data for lung cancer. In addition, the ideal approach is identified by comparing them against other works in detecting and describing pulmonary nodules. This work detects and describes the nodules and their characteristics in several stages: automated lung segmentation from the background, automated 2D and 3D clustering of vessels and nodules, applying shape and textures features, classification and automatic measurement of nodule characteristics. This work is tested on standard CT lung image data and shows promising results, matching or close to experts' diagnosis in the nodules number and their features (size/volume, location) and in terms the accuracy and automation. It also achieved a classification accuracy of 98% and efficient results in measuring the nodules' volume automatically.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.791971  DOI: Not available
Share: